CN116069897A - Query error correction methods, devices, computer equipment, storage media and program products - Google Patents

Abstract

The application provides a query error correction method, a query error correction device, a computer device, a storage medium and a program product, and relates to the technical fields of artificial intelligence, cloud technology, intelligent traffic, auxiliary driving and the like. Determining first error indication information of query information based on a target model matched with a user, so that personalized error judgment is performed based on the target model of the user, and the judgment result is more matched with the actual interest requirement of the user; when the query information is wrong, determining at least one piece of candidate information based on target error correction data matched with a user, and subsequently, further screening correction information based on second error indication information of the candidate information determined by the target model; and the target error correction data includes at least interest data of the user; therefore, personalized error correction is carried out on the query information based on the interest angle of the user, the error correction requirement of thousands of people and thousands of faces can be met, the actual accuracy of error judgment and correction is improved, and the actual accuracy of query error correction is further improved.

Description

Query error correction methods, devices, computer equipment, storage media and program products

technical field

The present application relates to technical fields such as artificial intelligence, cloud technology, intelligent transportation, and assisted driving. The present application relates to a query error correction method, device, computer equipment, storage medium, and program product.

Background technique

Currently, search engines are one of the most important ways for people to obtain information. Users only need to input keywords in the search box to find relevant information they are interested in. Since the keywords entered by users are often wrong, query error correction came into being. Query error correction is to identify and correct errors entered by users, so that the search engine can return the query results expected by users based on the corrected keywords.

In related technologies, a unified vocabulary of search engines is usually used for query error correction, and the vocabulary can collect correct words that appear frequently in the search engine; for example, for any user input keyword "direct dial", unified vocabulary can be used Corrected "direct dial" to "live".

The essence of the above error correction process is to use a unified error correction standard to measure the query intent behind each user's keyword. However, there are still some users whose corrected keywords do not match the user's corresponding query intention, so that the expected results cannot be obtained based on the corrected keywords. Therefore, the actual error correction accuracy rate of the above-mentioned query error correction is low.

Contents of the invention

The present application provides a query error correction method, device, computer equipment, storage medium and program product, which can solve the problem of low accuracy of actual error correction in related technologies. Described technical scheme is as follows:

On the one hand, a query error correction method is provided, the method includes:

In response to receiving query information entered into the search box by the user, determining first error indication information of the query information based on a target model matched with the user, the target model being used to base the user's interest data in the target application determining whether the information input into the target model is wrong;

In response to the first error indication information indicating that the query information is wrong, determining at least one candidate information associated with the query information based on the target error correction data matched with the user and the query information, the target correction The error data at least includes the interest data of the user in the target application;

determining second error indication information for the at least one candidate information based on the target model;

Correction information for the query information is determined from the at least one candidate information based on the second error indication information.

In another aspect, a query error correction device is provided, which includes:

The first determination module is configured to determine first error indication information of the query information based on a target model matched with the user in response to receiving query information input by the user into the search box, and the target model is used to determine the first error indication information of the query information based on the target model The user's interest data in the target application determines whether the information input into the target model is wrong;

A candidate determination module, configured to determine at least one candidate associated with the query information based on the target error correction data matching the user and the query information in response to the first error indication information indicating that the query information is wrong Information, the target error correction data at least includes the interest data of the user in the target application;

A second determining module, configured to determine second error indication information of the at least one candidate information based on the target model;

A correction module, configured to determine correction information of the query information from the at least one candidate information based on the second error indication information.

In another aspect, a computer device is provided, including a memory, a processor, and a computer program stored on the memory, and the processor executes the computer program to implement the above query error correction method.

In another aspect, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the above query error correction method is implemented.

In another aspect, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the above query error correction method is implemented.

The beneficial effects brought by the technical solution provided by the application are:

By determining the first error indication information of the query information based on the target model matched with the user, the target model is used to determine whether the information input into the model is wrong based on the user's interest data, so that the user can be personalized and wrongly judged based on the user's interest data , so that the judgment result more matches the user's actual interest needs; and when the query information is wrong, based on the target error correction data that matches the user, determine at least one candidate information associated with the query information, and subsequently determine the candidate information based on the target model The second error indication information is used to further filter out the correction information; and the target error correction data includes at least the user's interest data; thus, the query information can be personalized error correction based on the user's interest angle, ensuring that the correction information obtained after correction is more accurate Matching the user's real query intent can meet the error correction needs of thousands of people, improve the actual accuracy of the error judgment process and correction process, and then improve the actual accuracy of query error correction.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings that need to be used in the description of the embodiments of the present application.

FIG. 1 is a schematic diagram of an implementation environment of a query error correction method provided by an embodiment of the present application;

FIG. 2a is a schematic flowchart of a query error correction method provided by an embodiment of the present application;

Fig. 2b is a schematic flowchart of a query error correction method provided by the embodiment of the present application;

FIG. 3 is a schematic structural diagram of a user group model provided in an embodiment of the present application;

FIG. 4 is a schematic diagram of a query error correction process provided by an embodiment of the present application;

FIG. 5 is a schematic structural diagram of a query error correction device provided in an embodiment of the present application;

FIG. 6 is a schematic structural diagram of a computer device provided by an embodiment of the present application.

Detailed ways

Embodiments of the present application are described below with reference to the drawings in the present application. It should be understood that the implementation manner described below in conjunction with the accompanying drawings is an exemplary description for explaining the technical solutions of the embodiments of the present application, and does not limit the technical solutions of the embodiments of the present application.

Those skilled in the art will understand that unless otherwise stated, the singular forms "a", "an", "said" and "the" used herein may also include plural forms. It should be further understood that the terms "comprising" and "comprising" used in the embodiments of the present application mean that the corresponding features can be implemented as the presented features, information, data, steps, operations, elements and/or components, but do not exclude The realization is other features, information, data, steps, operations, elements, components and/or their combinations etc. supported by the technical field. It should be understood that when we say that an element is "connected" or "coupled" to another element, the one element can be directly connected or coupled to the other element, or it can mean that the one element and another element pass through intermediate elements. Establish a connection relationship. Additionally, "connected" or "coupled" as used herein may include wireless connection or wireless coupling. The term "and/or" used herein indicates at least one of the items defined by the term, for example, "A and/or B" indicates implementation as "A", or implementation as "A", or implementation as "A and B ".

In order to make the purpose, technical solution and advantages of the present application clearer, the implementation manners of the present application will be further described in detail below in conjunction with the accompanying drawings.

The query error correction method provided by this application involves the following technologies such as cloud computing, cloud storage, and big data. Exemplarily, computing devices included in the cloud computing resource pool can be used to call a model that matches the user to input data to the user. For example, cloud storage technology can be used to store the error correction data of each user, the user group model of each user group, the general model of the target application, and the general error correction data; Invoke the model, matching error correction data, etc. when querying the request. Exemplarily, the way of distributed database in big data can also be used to store a large amount of data such as interest data of a large number of users and interest data sets of user groups in a distributed manner.

Cloud computing is a computing model that distributes computing tasks on a resource pool composed of a large number of computers, enabling various application systems to obtain computing power, storage space and information services as needed. The network that provides resources is called a "cloud". From the user's point of view, the resources in the "cloud" can be infinitely expanded, and can be obtained at any time, used on demand, expanded at any time, and paid according to use.

As a basic capability provider of cloud computing, a cloud computing resource pool (referred to as a cloud platform, generally called an IaaS (Infrastructure as a Service, infrastructure as a service) platform will be established, and various types of virtual resources will be deployed in the resource pool for supply. External customers choose to use. The cloud computing resource pool mainly includes: computing equipment (a virtualized machine, including an operating system), storage equipment, and network equipment.

According to the division of logical functions, the PaaS (Platform as a Service) layer can be deployed on the IaaS (Infrastructure as a Service) layer, and the SaaS (Software as a Service, software as a service) layer can be deployed on the PaaS layer. Service) layer, or directly deploy SaaS on IaaS. PaaS is a platform on which software runs, such as databases, web containers, etc. SaaS is a variety of business software, such as web portals, SMS group senders, etc. Generally speaking, SaaS and PaaS are the upper layer relative to IaaS.

Cloud storage (cloud storage) is a new concept extended and developed from the concept of cloud computing. Distributed cloud storage system (hereinafter referred to as storage system) refers to the functions of cluster application, grid technology and distributed storage file system. A storage system that integrates a large number of different types of storage devices (storage devices are also called storage nodes) in the network to work together through application software or application interfaces to jointly provide data storage and service access functions.

Big data refers to a collection of data that cannot be captured, managed and processed by conventional software tools within a certain period of time. , high growth rates and diverse information assets. With the advent of the cloud era, big data has also attracted more and more attention, and big data requires special techniques to effectively process large amounts of data that tolerate elapsed time. Technologies applicable to big data, including massively parallel processing databases, data mining, distributed file systems, distributed databases, cloud computing platforms, the Internet, and scalable storage systems.

FIG. 1 is a schematic diagram of an implementation environment of a query error correction method provided by the present application. As shown in FIG. 1 , the implementation environment includes: a server 101 and a terminal 102. The server 101 may be a background server of an application program. The terminal 102 is installed with an application program, and the terminal 102 and the server 102 can perform data interaction based on the application program.

The application program may be configured with a search function, and the search function refers to returning search results related to the query information based on the query information input by the user. In this application, the application program can also be configured with a query error correction function. Query error correction refers to the error determination and correction of the query information input by the user, so that the server 101 returns the search result expected by the user based on the corrected information. In this application, personalized query correction can be performed on the query information based on the user's interest data, so that the query correction function of the application program can meet the personalized needs of thousands of people.

Personalized query error correction means that different users have different error types and error correction goals. This application can realize personalized error correction for user input query information, meet the personalized error correction needs of users with different interests, and improve video quality. Platform search user experience. For example, the user mistakenly entered the query information "Li Qianming Scene Collection", and the user who is interested in "Celebrating More Than Years" corrected it to "Li Qinming Scene Collection" is more appropriate. It is more reasonable for the user to correct it as "Li Qian". For another example, the user enters the query information "Surprise Story". For users who like to watch horror videos, the query input by the user is no problem, but for fans who tend to watch gangster movies and Jackie Chan, it is more reasonable and appropriate to correct it to "Police Story". intelligent.

In a possible scenario, the terminal 102 may send the query information input by the user to the server 101, and the server 101 will judge and correct the query information to obtain corrected information. The server 101 continues searching based on the corrected information. In another possible scenario, the terminal 102 acquires the query information input by the user and makes an error judgment and correction on the query information. The terminal 102 sends the corrected information to the server 101, and the server searches based on the corrected information.

Of course, the error determination and correction process of the query information can also be realized through interaction between the server 101 and the terminal 102; The result of the judgment is sent to the server 101, and the server 101 corrects the query information based on the result of the wrong judgment; or, both the server 101 and the terminal 102 can perform wrong judgment and correction on the query information, that is, the terminal 102 makes the wrong judgment and The corrected information is sent to the server 101, and the server 101 performs error determination and correction on the information sent by the terminal 102 again. Of course, in this embodiment of the application, the query error correction process and the various steps included in the query error correction process can be executed by any computer device such as a terminal or server, or between multiple computer devices. There is no restriction on the specific execution device of the error.

The embodiments of the present application can be applied to various scenarios, including but not limited to big data search, cloud technology, artificial intelligence, smart transportation, assisted driving, etc. The application can be any application that supports the query error correction function, for example, a video application, a live broadcast application, a content interaction application, a social application, a search engine, a shopping application, a game application, an enterprise communication and office tool platform, and the like. For example, for a scenario where a user enters query information in the search box of a video application to perform a query, the application may first correct the query information, so as to return the desired video list to the user based on the corrected information. The application program can be an independent application, or a program plug-in installed in the independent application, for example, an independent video application or a video playback applet installed in a social application, a video website based on a web page in a browser, and the like.

The server 101 can be an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, and can also provide cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, Cloud servers or server clusters for basic cloud computing services such as middleware services, domain name services, security services, CDN (Content Delivery Network, content distribution network), and big data and artificial intelligence platforms. The above-mentioned network may include but not limited to: wired network, wireless network, wherein, the wired network includes: local area network, metropolitan area network and wide area network, and the wireless network includes: bluetooth, Wi-Fi and other networks that realize wireless communication. The terminal 102 can be a mobile phone, an intelligent voice interaction device, a smart home appliance, a vehicle-mounted terminal (such as a vehicle-mounted navigation terminal, a vehicle-mounted computer, etc.), a tablet computer, a notebook computer, a digital broadcast receiver, a MID (Mobile Internet Devices, mobile Internet device), a desktop Computers, smart speakers, smart watches, etc. The terminal 102 and the server 101 may be connected directly or indirectly through wired or wireless communication, but is not limited thereto. In addition, the number of terminals 102 and servers 101 may be one or more, and the embodiment of the present application does not limit the specific values of the number of terminals and the number of servers. The details can also be determined based on actual application scenario requirements, and are not limited here.

Fig. 2a is a schematic flowchart of a query error correction method provided by an embodiment of the present application. The execution body of the method may be a computer device, and the computer device may be a server or a terminal. As shown in Figure 2, the method includes the following steps.

Step 201: In response to receiving query information entered into a search box by a user, the computer device determines first error indication information of the query information based on a target model matched with the user.

Wherein, the target model is used to determine whether the information input into the target model is wrong based on the interest data of the user in the target application. The computer device inputs the query information into the target model, and outputs first error indication information of the query information through the target model's error determination process of the query information. The first error indication information is at least used to indicate whether the query information is incorrect. Exemplarily, the target model matching the user may be different if the user maturity is different; the computer device may be configured with an association between the user maturity and the error judgment model; the computer device may be based on the association, The error judgment model corresponding to the user maturity of the user is used as the target model. Exemplarily, the first error indication information includes at least a correct probability of whether the query information is correct.

Step 202: The computer device determines at least one candidate information associated with the query information based on the target error correction data matching the user and the query information in response to the first error indication information indicating that the query information is wrong.

Wherein, the target error correction data at least includes the user's interest data in the target application. In the present application, the computer device may search the target error correction data for similar information similar to the query information based on the query information, and use the similar information whose similarity exceeds the target similarity threshold as the at least one candidate information. Certainly, the candidate information is information searched from the target error correction data, therefore, the candidate information also includes the user's interest data in the target application.

Exemplarily, the computer device may use at least one item of the user's error correction data or the user group's error correction data as the target error correction data. Wherein, the error correction data of the user group includes the interest data set of the user group, and the error correction data of the user includes the interest data of the user in the target application. Certainly, the target error correction data may also include general error correction data, and the general error correction data includes interest data sets of global users of the target application.

Step 203, the computer device determines second error indication information of the at least one candidate information based on the target model.

The computer device inputs each candidate information into the target model to obtain second error indication information for each candidate information, the second error indication information including the correct probability of the candidate information.

Step 204, the computer device determines correction information of the query information from the at least one candidate information based on the second error indication information.

The computer device can filter corrected information satisfying a specified probability condition from at least one candidate information, for example, the specified probability condition can be that the correct probability of the candidate information is greater than 0.8, the correct probability of the candidate information is greater than the correct probability of the query information, and the like.

By determining the first error indication information of the query information based on the target model matched with the user, the target model is used to determine whether the information input into the model is wrong based on the user's interest data, so that the user can be personalized and wrongly judged based on the user's interest data , so that the judgment result more matches the user's actual interest needs; and when the query information is wrong, based on the target error correction data that matches the user, determine at least one candidate information associated with the query information, and subsequently determine the candidate information based on the target model The second error indication information is used to further filter out the correction information; and the target error correction data includes at least the user's interest data; thus, the query information can be personalized error correction based on the user's interest angle, ensuring that the correction information obtained after correction is more accurate Matching the user's real query intent can meet the error correction needs of thousands of people, improve the actual accuracy of the error judgment process and correction process, and then improve the actual accuracy of query error correction.

Fig. 2b is a schematic flowchart of a query error correction method provided by the embodiment of the present application. The execution body of the method may be a computer device, and the computer device may be a server or a terminal. As shown in Figure 2b, the method includes the following steps.

Step 301, the computer device receives the query information input by the user into the search box.

The user can input query information based on the search box of the target application; if the computer device is a terminal, the terminal acquires the query information when it detects that the search box includes the query information, or when the terminal detects a search instruction triggered by the user, it acquires the information in the search box. search information. If the computer device is a server, the terminal sends the query information to the server, and the server receives the query information sent by the terminal.

The query information may include at least one of a query character string input by a user, a character string extracted from a query image, or text description information of an emoticon. Exemplarily, the user may input a character string in the search box. Alternatively, the user can also input images, emoticons, etc. based on the search box, and the computer device can obtain character strings corresponding to the images, emoticons, and the like. Correspondingly, the implementation manner for the computer device to receive the query information may include at least the four manners shown in the following manners 1 to 4.

Mode 1: The computer device receives the query string input by the user in the search box of the target application, and uses the query string as the query information.

Wherein, the character string may include but not limited to words, pinyin, symbols, numbers, punctuation marks, operators and so on. The page of the target application may include a search box, and the search box may include a character string input plug-in, and the user may trigger the character string input plug-in to input a query string in the search box, and the computer device receives the query string input by the user.

Method 2: The computer device receives the query image input by the user through the search box, extracts a character string included in the query image, and uses the character string extracted from the query image as the query information.

The search box may also include an image input plug-in, the user can trigger the image input plug-in to input a query image, and the computer device can perform character string extraction on the query image to obtain a character string included in the query image. For example, the query image may be a film and television drama cover image including a film and television drama title, and the computer device may extract the film and television drama title from the film and television drama cover image. Of course, the computer device can also perform image recognition on the query image to recognize the text information corresponding to the image content, for example, based on the recognized celebrity's head portrait in the image, and further obtain the celebrity's name.

Method 3: The computer device receives the emoticon input by the user through the search box, obtains the text description information of the emoticon, and uses the text description information of the emoticon as the query information.

The search box can also include an emoticon input plug-in, and the user can trigger the emoticon input plug-in to input an emoticon, and the computer device can obtain the text description information corresponding to the emoticon input by the user from the association between the emoticon icon and the text description information. Exemplarily, the user triggers the emoticon image input plug-in of the search box on the page of the target application, multiple candidate emoticons may be displayed on the page, and the computer device may acquire the emoticon selected by the user among the multiple candidate emoticons. For example, candidate emoticons may include emoticons corresponding to various roles in multiple film and television dramas, for example, a user may select an emoticon corresponding to an interested heroine of a film and television drama to query.

Mode 4: The computer device receives the query string and the emoticon input by the user through the search box, and obtains the text description information of the emoticon, and uses the query string and the text description information of the emoticon as the query information.

The user may input two types of information, character strings and emoticons, and the query information may include query strings and text description information of emoticons. Wherein, the manner in which the computer device obtains the text description information of the emoticon is the same as the implementation manner of the third manner above, and will not be repeated here.

It should be noted that the computer device can obtain query information by combining any two or three of the above methods 1 to 3. This application only uses the combination of user input strings and emoticons in the above method 4 for illustration. The query information may be obtained by combining the user input of characters and images, or the query information may be obtained by combining the user input of characters, images, and emoticons. The embodiment of the present application does not limit the specific combination method.

In this step, by configuring the search box, the character string entered by the user in the search box can be accurately obtained as the query information, and subsequent query error correction is performed based on the string entered by the user, which improves the accuracy of obtaining the query information; The text description information of character strings or emoticons extracted from the query image is used as query information, which supports query error correction of information input by users in various search ways, and improves the applicability of the query error correction method in this application.

Step 302: In response to receiving the query information entered by the user into the search box, the computer device determines a target model matching the user based on the user maturity of the user.

The user maturity is used to indicate the user's activity level in the target application; the target model is used to determine whether the information input into the target model is wrong based on the user's interest data in the target application. Depending on the user maturity, the target model matching the user may also be different; in this step, the computer device may determine the user maturity of the user, and then determine the target model based on the user maturity. The computer device may be configured with an association relationship between the user maturity and the error judgment model; based on the association relationship, the computer device may use the error judgment model corresponding to the user maturity of the user as the target model.

In a possible implementation manner, this step may include the following steps 3021-3022 (not shown in the figure).

Step 3021, the computer device determines the user maturity of the user.

The computer device may determine user maturity based on the user's active time, interaction conditions, and the like in the target application. The computer device may determine the maturity of the user based on at least one of the user's active time in the target application or the number of user interest tags. In a possible implementation manner, the user maturity may be the maturity of user interest profiles accumulated by the user during the active time of the target application. The process for the computer device to determine the user's maturity may include: the computer device determines the user's active time in the target application; the computer device calculates the user's active time within the active time based on the operation object associated label of the user's interactive operation the number of interest tags; the computer device determines the maturity of the user based on the active time and the number of interest tags. Exemplarily, the number of interest tags may be the total number of multiple interest tags included in the user interest profile. Exemplarily, the computer device may use the tag associated with the operation object of the interactive operation as the user's interest tag, and count the number of the user's interest tags based on the number of interactive operations. For example, each time an interest tag is added, the number of interest tags of the interest tag is configured as 1, and each time an interactive operation of the interest tag is added, the number of interest tags is accumulated by 1. Exemplarily, the active time may include the duration of the user's interactive operations on the target application. For example, the interactive operations include but are not limited to: playing videos, giving likes, sending barrage messages, posting comments, posting videos, and browsing in the community. Any operation related to the target app, such as posting, downloading videos, interacting with other users on the target app, browsing product detail pages or product transactions.

Based on the active time and the number of interest tags, the computer device determines the user maturity. The process may include: when the active time does not exceed the target time threshold, the computer device determines that the user maturity is the first maturity; when the When the active time exceeds the target time threshold, if the number of interest tags does not exceed the target tag number threshold, the computer device determines that the user maturity is the second maturity; if the active time exceeds the target time threshold and the interest tag number exceeds the target tag number threshold, the The computer device determines the user maturity as a third maturity. Exemplarily, the target time threshold and the target tag count threshold can be configured based on needs, which are not specifically limited in this embodiment of the present application. For example, the target time threshold may be 5 days, 10 days, etc.; the target label number threshold may be 300, 500, etc.

In a possible example, the computer device may be configured with three user maturity levels, which are arranged in ascending order according to the level of user maturity, and the three user maturity levels may include newness, immature user interest portrait, and mature user interest portrait; example Specifically, the maturity level of the user can be determined through the following (1)-(3) process:

(1) Determine whether it is a new cold user: whether the active time of the user is less than T (such as T is 5 days), if it is less than T, it is a new cold user;

(2) Determine whether the user’s interest profile is not mature enough: if it is not a new user, continue to determine whether the number of user interest tags is less than or equal to K (for example, K is 300), if it is less than or equal to K, the user’s interest profile is not mature enough;

(3) Determine whether it is a mature user of user interest portrait: if it is not a new user and the number of user interest tags is greater than K, it is a mature user of user interest portrait.

In a possible example, the associated tags of the operation object may include, but not limited to: tags associated with the operation object, tags associated with objects associated with the operation object, interest tags configured by the user through tag configuration operations, etc. . For example, the associated object of the operation object may be: a certain film and television drama, a certain creator, etc. associated with the post published or replied by the user. Exemplarily, the target application may be a video application, and the video application is configured with a query error correction function. The user's interactive operations on the video application may include, but are not limited to: video playback operations, video interactive operations, interactive operations on posts in the community, interactive operations on creators in the video application, label configuration operations, etc. Exemplarily, the computer device can count the number of tags of interest through the following processes (1)-(3):

(1) The user completes an effective playback operation of a video (the playback completion degree is greater than a certain threshold; for example, the playback time exceeds 30 minutes), and interacts with a video (like, forward, comment, barrage, share, etc.), then Accumulate the tags of the video to the user's user interest profile; if the user interest profile does not have the same interest tag, configure the number of interest tags of the interest tag as 1; if the user profile already has the same interest tag, the user interest profile The number of interest tags of the same interest tag in .

(2) Accumulation of user interest in post interaction operations in the community: If a post or reply post in the community is associated with a certain character, a certain film and television drama, or a certain film and television character, it will correspond to the actors, roles, and characters of the film and television drama. , feature description tags and other related tags are accumulated to the user profile; if the user interest profile does not have the same interest tag, the number of interest tags of the interest tag is configured as 1; if the user interest profile already has the same interest tag, the user interest profile The number of interest tags of the same interest tag in .

(3) The user completes an interactive operation on a creator (following operation, like operation, sharing operation of the creator's work, sharing operation of the creator's homepage, etc.), and the tags of all the creator's videos are accumulated to the user portrait, if If there is no same interest tag in the user interest profile, configure the number of interest tags of the interest tag as 1; if the same interest tag already exists in the user interest profile, add 1 to the number of interest tags of the same interest tag in the user interest profile.

In a possible example, the interest tags in the user interest profile can also be attenuated; for example, if the number of interest tags of a certain interest tag of the user does not increase within time D (such as one week, one month, half a year, etc.), the Attenuate the number of interest tags of this interest tag. For example, the half-life decay method can be used to decay the number of interest tags to half of the previous time D. For example, if the original number of interest tags is 100 and the time D is one month, then the first After one month, the number of interest tags decays to 50, after the second month, the number of interest tags decays to 25, and so on.

Step 3022: When the computer device responds that the user's maturity exceeds the maturity threshold, the computer device determines at least one target model that matches the user based on user groups that have similar interests to the user.

The target model at least includes a user group model corresponding to a user group with similar interests to the user, and the user group model is used to determine whether the information input to the user group model is wrong based on the interest data set of the user group, and the user group includes the user group The user has at least two users with similar interests; the interest data set of the user group includes the interest data of the user in the target application. In a possible example, this step may include: when the user maturity exceeds a maturity threshold, the computer device may determine, based on a user group with similar interests to the user, that the user group model corresponding to the user group is the same as The target model that this user matched. In another possible example, the target model may include a user group model and a general model, and the general model is used to determine whether the information input into the general model is wrong based on the global user data set of the target application; correspondingly, this step may Including: when the maturity of the user exceeds the maturity threshold, the computer device may also determine the general model and the user group model as target models matching the user based on user groups with similar interests to the user. Exemplarily, the maturity threshold may be the first maturity, that is, when the user maturity is the second maturity or the third maturity, the user maturity exceeds the maturity threshold.

It should be noted that the computer device may divide multiple users of the target application into multiple user groups based on user interest tags, and each user group includes multiple users with similar interests. In a possible example, the computer device may cluster users whose user maturity is the third maturity, for example, cluster users with mature interests according to interest similarity to construct a user group with similar interests. Wherein, the interest similarity between two users is: the quotient of the cumulative count of repeated interest tags included in the interest tag sets of the two users divided by the number of interest tags of the two users. For example, the KMeans (k-means clustering algorithm, k-means clustering algorithm) clustering method can be used to cluster users with mature interests to obtain multiple user groups with similar interests and the cluster center of each user group with similar interests. Wherein, the clustering center of each user group may be a center point of the user group, for example, the clustering center may include the number of interest tags of each interest tag corresponding to the center point. Exemplarily, the goal of using the KMeans clustering method is to make the interests of users in the same cluster similar, and the interests of users in different clusters are far apart. When using KMeans clustering, the number of target clusters can be C, where C can be the quotient of the number of users with mature interests divided by the number of standard users U, for example, U can be configured as 100,000.

By building user groups with similar interests, it is possible to assist users with immature interests in query error correction. For example, by mapping immature users to user groups with mature interests, the richness of personalized data for users with immature interests can be improved. In addition, it is also possible to avoid the need for a separate error judgment model for each user, avoiding the high cost of error judgment model training and storage, so that there is no need to build a separate model for each user, and by building an error judgment model for the user group, both training and storage The cost is controllable, and at the same time, it meets the user's personalized error correction needs.

Exemplarily, after clustering the users of the third maturity level to obtain multiple user groups, for users whose user maturity is the third maturity level, the user group with similar interests to the user is the user group to which the user belongs . For users whose user maturity is the second maturity, the similarity between the user and each user group can be calculated based on the cluster center of each user group and the number of each interest tag of the second maturity user, The user group corresponding to the maximum similarity is taken as the user group whose interests are similar to the user.

In a possible example, when the user maturity does not exceed the maturity threshold, for example, for a new cold user whose user maturity is the first maturity, the computer device may determine the general model as the target matching the user Model.

It should be noted that, through the above method of accumulating user interest tags and judging method of user maturity, users can be divided into different levels of personalization, divided from multiple dimensions such as active time and number of interest tags, which improves user maturity. The accuracy of the degree division. Furthermore, different error judgment models are matched to users for different maturity levels, and different error correction data are subsequently used, so as to facilitate subsequent personalized query error correction for different maturity levels and improve the accuracy of query error correction.

Step 303, the computer device determines first error indication information of the query information based on the target model matched with the user.

The computer device inputs the query information into the target model, and outputs the first error indication information of the query information through the error judgment process of the target model on the query information, and the first error indication information is at least used to indicate whether the query information is mistake. Exemplarily, the first error indication information includes at least a correct probability of whether the query information is correct.

The first error indication information is used to indicate whether the query information is wrong and an error type of the query information. In a possible implementation manner, the first error indication information includes a correct probability and an error type probability of the query information, and the error type probability is used to indicate the possibility that the query information has at least one type of error; then the step The process of 303 may include: the computer device extracts at least one type of description information of the query information through the target model, and the at least one type of description information includes pinyin, strokes, position codes, and keywords included in the query information. Or at least one of the phrases included in the query information; the computer device makes an error judgment on the query information based on at least one description information of the query information, and obtains the correct probability and error type probability of the query information.

Example 1, for the case where the target model includes any model in the user group model or the general model, for example, when the user maturity level is the second maturity level or the third maturity level, the query information can be output through the user group model probability of being correct. When the user maturity is the first maturity, the correct probability of query information can be output through the general model.

Example 2, when the target model includes a user group model and a general model, the correct probability of the query information is the average of the two correct probabilities respectively output by the two models; exemplary, when the user maturity is the second maturity or At the third maturity level, the average value between the first correct probability output by the user group model and the second correct probability output by the general model is determined as the correct probability of the query information.

It should be noted that the method of outputting the probability of the wrong type based on the target model is the same as the process of obtaining the correct probability shown in the above-mentioned examples 1 and 2. The same reasoning as the example can be used to obtain the probability of the wrong type output by a target model; Alternatively, as in Example 2, the average value between the first error type probability output by the user group model and the second error type probability output by the general model can be obtained as the error type probability of the query information, which will not be repeated here. A repeat.

The target model can be a BERT (Bidirectional Encoder Representations from Transformers, bidirectional encoding representation based on the converter) model, as shown in Figure 3, the query information can be expressed as a user Query (query), and the user Query is input into the trained BERT model Through the BERT model, multiple descriptive information such as the word representation in the user query, the character representation in the user query, the position code of the character in the user query, the stroke representation of the characters in the user query, and the pinyin representation of the characters in the user query are obtained; in Figure 3, CLS represents a placeholder mark , the CLS position corresponding output is used to represent the context features of the whole sentence, which can be placed at the first place of the sentence. The BERT model is based on the features of each position obtained by using multiple description information, and outputs the depth features corresponding to each position, for example, it can be a 768-dimensional feature vector; then based on the feature vectors corresponding to the output depth features of multiple positions, the maximum value The pooling operation (maxpool) obtains the maximum value of the depth corresponding to each position, splices the vector representing the maximum depth of each position and the representation context feature vector output by the CLS bit, and passes the fully connected layer based on the vector obtained by splicing, and After the activation function of the output layer (such as the softmax normalized exponential function), etc., the correct probability and the wrong type probability are output.

In a possible implementation, the training process of the user group model includes: a computer device obtains historical query information of at least two users in the user group, and generates noise data including The negative sample data of the negative sample data; the computer device predicts the negative sample data through the initial model to obtain the predicted sample data; the computer device calculates the initial model based on the similarity between the predicted sample data and the positive sample data Adjust the model parameters until the initial model meets the model training conditions and stop adjusting to obtain the user group model. Wherein, the noise generation operation includes at least one of replacing a word segment operation, deleting a word segment operation or adding a word segment operation, and the replacing word segment operation includes replacing with a phonetic similar word segment, replacing with a font similar word segment Or replace it with at least one item in the semantically similar word segment.

Exemplarily, for the user group model of any user group, the historical query information set (such as Query data) input by multiple users in the user group can be obtained, and a part of the input historical query information set is randomly selected, and the The filtered part of the historical query information set is used for noise generation to automatically construct negative sample data including noise data. Exemplarily, noise generation can be performed in the following ways (1)-(3):

(1) Generate replacement-type error noise data by replacing word fragment operations, and randomly replace words in the query data with other words, including the following four methods (a)-(d):

(a) Randomly replace words in certain positions with other words;

(b) Randomly replace words in certain positions with other words with similar shapes;

(c) Randomly replace words in certain positions with other words with similar sound;

(d) Randomly replace words in certain positions with other words with similar word content.

(2) Generate missing noise data by deleting word fragments, and randomly delete words in certain positions in the Query data.

(3) Generate redundant error noise data by adding word fragment operations, and randomly add some other words at certain positions in the Query data.

Among them, the random noise data generated by the above (1)-(3) method is used as the noise data included in the wrong negative sample Query data, and the original form of the Query data can be used as the correct positive sample Query data after a certain amount of manual verification. The initial error decision model is then trained on the basis of positive and negative samples. In a possible example, the computer device can predict the correct form of each Query data in the negative sample, for example, predict the correct data corresponding to each Query data in the negative sample to obtain the predicted sample data; based on the predicted sample data and the positive sample The similarity between the data is used to train the initial error judgment model, and after the initial error judgment model training converges, the user group model is obtained. In yet another possible example, the computer device can also predict the sample indication information of each Query data in the negative sample, and the sample indication information can include the probability of the correct sample, the probability of the wrong type of the sample, and each Query obtained during the noise generation process of the negative sample. The true value label of the data, the true value label can include whether each Query data in the negative sample is wrong and the true value of the error type; based on the similarity between the sample indication information and the true value label, the initial error judgment model is trained, After the initial error judgment model training converges, the user group model is obtained.

Exemplarily, for a general model, the global historical query information set (such as Query data) input by all users on the target application can be obtained, and a part of the input historical query information set is randomly selected, and the filtered part of the historical query information Query information set for noise generation to automatically construct negative sample data including noisy data. Positive sample data can be a collection of global historical query information that has undergone certain manual verification. Wherein, the generation method of the negative sample data using the global historical query information set is the same as the generation method of the negative sample data corresponding to the user group model above, and will not be repeated here. Wherein, the computer device can predict the negative samples through the initial general model, based on the similarity between the predicted prediction sample data and the positive sample data, train the initial general model, and after the initial general model training converges, the general model is obtained . Wherein, the training method of the general model is the same as the training method of the above-mentioned user group model, and will not be repeated here.

In this step, a part of the historical Query data input by users in the user group is randomly selected, noise generation is performed on this part of the data, Query error data is automatically constructed, and the user group model is obtained based on positive and negative sample training, which is convenient for follow-up Based on the user group model, personalized error judgments are made for users with different interests, which improves the accuracy of error judgments and further improves the accuracy of query error correction.

Further, when the first error indication information indicates that the query information is wrong, based on the following step 304, error correction is performed on the query information; when the first error indication information indicates that there is no error in the query information, subsequent steps can be directly performed based on the query information Search to return search results, and the subsequent step 304 may not be performed.

Step 304: In response to the first error indication information indicating that the query information is wrong, the computer device determines target error correction data that matches the user based on the maturity of the user.

The target error correction data includes at least the user's interest data in the target application; for example, the computer device may use at least one of the user's error correction data or the user group's error correction data as the target error correction data data. Wherein, the error correction data of the user group includes the interest data set of the user group, and the error correction data of the user includes the interest data of the user in the target application.

In a possible implementation manner, the target error correction data may also include general error correction data, and the general error correction data includes the interest data set of global users of the target application; based on several possible situations of the target error correction data, this The implementation of the steps may include the following four methods.

In the first manner, the computer device determines the error correction data of the user group as the target error correction data matching the user based on the user group of the user.

In a possible example, when the user maturity level is the second maturity level, the error correction data of user groups with similar interests to the user may be used as the user's target error correction data. The error correction data of the user group may include the interest data set of the user group in the target application.

Wherein, the interest data set of the user group may include the behavior data of the user group in the target application. In a possible example, the behavior data of the user group in the target application includes, but is not limited to: video titles and video descriptions of the creators that the user group has effectively watched or followed, video comments and bullet chats of the user group, and the user group’s content in the community. Comments on posts or replies published in the website, user history query information of user groups, etc. It should be noted that the behavior data can be configured based on needs, and this application only uses the above-mentioned several types of data as examples for illustration. Of course, the behavior data may also include the product name of the user’s favorite or purchased product, the title of the article favorited or forwarded, the title of the live room when watching the live broadcast, the video title of the downloaded video, etc. The behavior data in this embodiment of the application may include data that does not Do limited.

In a possible implementation manner, the target error correction data further includes a query index of the interest data, and the query index is used to index similar segments in the interest data. The computer device may generate error correction data of the user group based on the interest data set of the user group, and the process may include: the computer device performs word segmentation on the interest data set of the user group, and based on the similarity between multiple segments obtained by word segmentation degree, generating a query index for indexing similar segments; and generating error correction data for the user group based on the plurality of segments and the query index of similar segments. In an example, the computer device performs word segmentation on the interest data set, and may filter out segments whose occurrence frequency exceeds a target frequency threshold in the word segmentation results, and generate a query index based on the similarity between the screened out segments. Wherein, the target frequency threshold may be configured based on needs, which is not limited in this embodiment of the present application. For example, the target frequency threshold may be 5 times, 3 times, 10 times, and so on.

In a possible example, the error correction data can be constructed based on the similarity of pinyin between multiple segments, then the query index includes a pinyin query index, and the pinyin query index is used to index word segments with similar pinyin in the interest data. In yet another possible example, error correction data can also be constructed based on the font similarity between multiple segments, then the query index includes a stroke query index, and the stroke query index is used to index word segments with similar strokes in the interest data; In yet another possible example, error correction data can also be constructed based on the semantic similarity between multiple segments, then the query index includes a semantic query index, and the semantic query index is used to index words with similar semantics in the interest data fragment. Correspondingly, the following (1)-(3) respectively show the ways of constructing error correction data based on pinyin similarity, font similarity or semantic similarity.

(1) Construct error correction data based on pinyin similarity: perform word segmentation on the interest data set, build a user group vocabulary based on multiple word segments whose frequency of occurrence exceeds the target frequency threshold, and perform pinyin notation on the word segments in the user vocabulary; And based on the pinyin of the multiple word segments, the pinyin-like word segments among the multiple word segments are obtained, and the pinyin is used as the pinyin index of the pinyin-like word segment, and the error correction data of the user group including the multiple word segments and the pinyin index is obtained. For example, use the pinyin of an entry as the query index for similar retrieval in Elasticsearch (used for distributed full-text search); then input the pinyin of a certain entry to obtain one or more similar entries that are similar to the pinyin of the entry , and the pinyin similarity between the entry and similar entries can also be obtained.

(2) Constructing error correction data based on font similarity: dismantling the strokes of the word fragments in the constructed user group vocabulary; The stroke sequence is used as a grapheme index of the grapheme-similar word segment, and error correction data of a user group including multiple word segments and grapheme indexes are obtained. For example, use the stroke sequence of an entry as the query index for Elasticsearch similarity retrieval; then input the stroke sequence of a certain entry to obtain one or more similar entries that are similar in shape to the entry, and you can also obtain the Glyph similarity between an entry and similar entries.

(3) Construct error correction data based on semantic similarity: split the word fragments in the constructed user group vocabulary according to word granularity, and based on the split keywords and similar keywords semantically similar to the keyword, determine including multiple semantically similar word fragments of similar keywords; and construct a semantic query index for multiple semantically similar word fragments. For example, use semantically similar word fragments as a semantic query index; for example, semantic similarity can be that the word content of the keywords included in the word fragments is similar, and it can be realized to enter a certain query entry and return other entries with similar word content, such as Enter "Manchengquandaijinjia" and search for "Manchengjinjia" with similar word content according to word similarity; you can also get the word content similarity between this entry and similar entries.

Wherein, the method of constructing the user group vocabulary in (2) and (3) above is the same as that in (1), and will not be repeated in (2) and (3).

In the second manner, the computer device determines the error correction data of the user and the error correction data of the user group as the target error correction data based on the user group of the user.

In a possible example, when the user maturity level is the third maturity level, the error correction data of user groups with similar interests to the user may be used as the user's target error correction data. The user's error correction data may include the user's interest data in the target application.

Wherein, the user's error correction data may include a query index, and the computer device may also obtain the error correction data including multiple segments and query indexes based on word segmentation of the user's interest data. The user's interest data may also include the user's behavior data in the target application, and the user's behavior data may include an illustration of the behavior data shown in the first manner. In addition, the query index may include at least one of a pinyin index, a font index, or a semantic index. Among them, the method of constructing the user's error correction data based on the user's interest data, and the specific method of constructing the user's error correction data based on the similarity of pinyin, font similarity or semantic similarity, are similar to the construction of user error correction data in the first method above. The process of group error correction data is the same. The behavior data and the specific implementation of constructing the user's error correction data will not be repeated here.

In a third manner, the computer device determines the error correction data and general error correction data of the user group as the target error correction data based on the user group of the user.

In a possible example, when the user maturity is the second maturity, the general error correction number and the error correction data of user groups with similar interests to the user may be used as the user's target error correction data.

Wherein, the general error correction data includes at least the interest data set of the global users of the target application, that is, the interest data of all users in the target application. Of course, the error correction data of the global user may also include a query index, and the computer device may also obtain the error correction data including multiple segments and the query index based on word segmentation of the interest data set of the global user. The global user interest data set may also include behavior data of all users in the target application; the query index may include at least one of a pinyin index, a font index, or a semantic index. Among them, the method of constructing general error correction data based on the interest data set of global users, and the method of constructing general error correction data based on pinyin similarity, font similarity or semantic The process of group error correction data is the same. Here, the specific implementation manner of constructing the general error correction data will not be repeated one by one.

In a fourth manner, the computer device determines the error correction data of the user, the error correction data of the user group, and the general error correction data as the target error correction data based on the user group of the user.

In a possible example, when the user maturity is the third maturity, the general error correction number, the user's error correction data, and the error correction data of user groups similar to the user's interests can be used as the user's target correction data. wrong data. Among them, the construction methods of the user's error correction data, the general error correction data and the user group's error correction data are the same as the above three methods, and will not be repeated here.

In this step, when the query information is wrong, the target error correction data that matches the user is obtained based on the user's maturity. The target error correction data includes the user's interest data, which is convenient for subsequent follow-up based on the user's interest data, with the user's interest direction as the Correct the target to correct the query information, meet the personalized error correction needs of users with different interests, further improve the accuracy of the actual error correction of the query error correction, and improve the user search experience.

Step 305, the computer device determines at least one candidate information associated with the query information based on the target error correction data matched with the user and the query information.

The target error correction data at least includes the user's interest data in the target application. In the present application, the computer device may search the target error correction data for similar information similar to the query information based on the query information, and use the similar information whose similarity exceeds the target similarity threshold as the at least one candidate information. Certainly, the candidate information is information searched from the target error correction data, therefore, the candidate information also includes the user's interest data in the target application.

In a possible implementation manner, the query information includes a query string, and associated word segments may be searched based on the word segments included in the string, so as to construct candidate information based on the associated word segments. Correspondingly, this step may include: the computer device obtains at least one associated word segment corresponding to the at least one word segment from the target error correction data based on the at least one word segment included in the query string; the computer device obtains at least one associated word segment corresponding to the at least one word segment based on the at least one Associating word segments, replacing at least one corresponding word segment in the query information, to obtain the at least one candidate information. For example, for the first word segment "AB" in the string "ABXXX", the associated word segments associated with "AB" in the target error correction data include "CD" and "EF". After replacing "AB", we get " ABXXX" candidate information "CDXXX", "EFXXX"; where "AB" can represent English words, Chinese phrases, one or more Chinese characters, symbols and other forms.

In a possible implementation, the target error correction data also includes a query index of the interest data, and for each word segment included in the query string, the computer device searches the target error correction data for the word segment based on the query index. At least one associated word segment that the segments are similar to. The associated word segment may include at least one of a pinyin-like word segment, a shape-like word segment, or a semantically similar word segment. Based on the pinyin query index, grapheme query index and phonetic query index shown in the above step 304, in this step, for each word segment, the computer device can obtain the word segment by at least one of the following methods 1 to 3 associated word fragments.

Mode 1. The query index includes a pinyin query index, and the computer device queries at least one pinyin-like word segment of the word segment from the target error correction data based on the pinyin of the word segment and the pinyin query index.

The computer device can extract the pinyin of the word segment, and query at least one pinyin-like word segment indexed by the pinyin from the target error correction data. For example, for a word segment "coupon" in the user input information, the corresponding pinyin "you hui juan" can be input, and the pinyin similar word segment "coupon" found based on the pinyin index can be output.

Mode 2. The query index includes a stroke query index, and the computer device searches the target error correction data for at least one similar word segment of the word segment based on the stroke sequence of the word segment and the stroke query index.

The computer device can extract the stroke sequence of the word segment, and query at least one word segment with similar font shape indexed by the stroke sequence from the target error correction data. For example, for a certain word segment "hook fish" in the information input by the user, the corresponding stroke sequence can be input, and the word segment "fishing" with a similar shape can be output based on the stroke sequence index query.

Method 3: The query index includes a semantic query index, and the computer device searches at least one semantically similar word segment of the word segment from the target error correction data based on the word segment and the semantic query index.

The computer device can query the target error correction data for at least one semantically similar word segment indexed by the word segment according to the word segment and the semantic query index. For example, if you input "Full City with Golden Armor" according to the similarity of words, you can search for "Full City with Golden Armor" with similar entry words; for example, enter "Suddenly", and output the semantically similar word segment "Suddenly" based on the query of "Suddenly". ","suddenly".

It should be noted that, corresponding to the four ways of the target error correction data shown in step 304, correspondingly, this step can be based on one or more types of error correction data included in the target error correction data, from the target error correction data The at least one associated word segment is queried in each type of error correction data included. Taking the target error correction data as an example including the error correction data of the user group and the user error correction data, for the first method in step 305, the process for the computer device to obtain at least one pinyin-like word segment may include: the computer device may respectively base on the word segment and the pinyin index of the error correction data of the user group, query at least one first pinyin similar word segment of the word segment from the error correction data of the user group; and, based on the pinyin index of the error correction data of the word segment and the user, from At least one second pinyin-like word segment of the word segment is queried in the user's error correction data. Of course, when the target error correction data includes one kind of error correction data or three kinds of error correction data or other situations in two kinds of error correction data, and for situations such as font-like word segments or semantically similar word segments, the acquisition method of associated word segments is the same as The same reasoning applies to the acquisition of associated word segments based on the two types of error correction data in the above examples, and will not be repeated here.

It should be noted that, for each word segment in the query information, similar word segments of the word segment (such as the three similar word segments in the above-mentioned mode 1 to mode 3) can be used to replace the word segment in the query information to construct Get a candidate information. For example, a word segment in the query information may be replaced, and a candidate information obtained from the replacement includes a similar word segment. For another example, it is also possible to replace two or more word segments in the query information at the same time, then one piece of candidate information obtained by replacement includes two or more similar word segments. In the embodiment of the present application, one word fragment or multiple word fragments can be replaced at a time to construct candidate information. The above is only an example of replacing one word fragment at a time. This application does not limit the specific method of constructing candidate information.

Step 306, the computer device determines second error indication information of the at least one candidate information based on the target model.

The computer device inputs each candidate information into the target model to obtain second error indication information for each candidate information, the second error indication information including the correct probability of the candidate information. Wherein, it is the same as the process of obtaining the correct probability of query information in step 303. Correspondingly, in this step, based on the situation of Example 1 in step 303, the correct probability of the candidate information output by the user group model can be obtained; or, the general The correct probability of the candidate information output by the model. Alternatively, based on the case of Example 2 in step 303, the average value between the third correct probability output by the user group model and the fourth correct probability output by the general model may be obtained as the correct probability of the candidate information. Wherein, the specific implementation manner of obtaining the correct probability of the candidate information is the same as the process in the first example or the second example in the above step 303, and will not be repeated here.

In steps 305-306, by constructing a plurality of candidate information for the query information based on the similarity, the possibility of the candidate matching the correct information is improved. Moreover, candidate information can also be constructed based on multiple methods such as pinyin query index, grapheme query index, or entry query index, which enriches the source of candidate information, and the subsequent correction of user query information is more accurate, thereby reducing user error input query information The time-consuming deletion and re-entry costs improve the actual search accuracy and efficiency of subsequent searches.

Step 307, the computer device determines correction information of the query information from the at least one candidate information based on the second error indication information.

The computer device can filter corrected information satisfying a specified probability condition from at least one candidate information, for example, the specified probability condition can be that the correct probability of the candidate information is greater than 0.8, the correct probability of the candidate information is greater than the correct probability of the query information, and the like.

In a possible implementation manner, the computer device may first correct the correct probability of the candidate information based on the error type probability, and then determine the correction information based on the corrected correct probability. Correspondingly, this step can be realized through the following steps 3071-3072 (not shown in the figure).

Step 3071, the computer device corrects the correct probability of the at least one candidate information based on the wrong type probability of the query information, and obtains the correct score of the at least one candidate information.

In a possible implementation manner, the candidate information includes an associated word segment that replaces the word segment in the query information; the computer device may determine from the error type Among the probabilities of at least one error type included in the probability, the type probability of the matching error type matching the similar type is obtained; the computer device determines the correct score of the candidate information based on the correct probability of the candidate information and the type probability. Exemplarily, the similarity type includes: at least one of similarity in pronunciation, similarity in shape, and similarity in semantics; wherein, the matching error type of similarity in pronunciation includes similarity in pronunciation, and the matching error type in similarity in shape includes similarity in shape. The semantically similar matching error types include at least one of word content similar errors, missing errors, or redundant errors. Wherein, if the number of words in the associated word segment is more than the number of words in the replaced word segment in the original replacement position, it is a missing error, and if the number of words in the associated word segment is less than the number of words in the replaced word segment in the original replacement position, it is a redundant error. Wherein, the similar type between the associated word segment and the replaced word segment in the query information illustrates the source of the associated word segment, for example, if the associated word segment is a pinyin similar word segment obtained based on the pinyin query index, then the corresponding associated word segment and the replaced word segment The type of similarity between the replacement word segments is phonetic similarity; if the related word segment is a similar word segment in shape, the corresponding similar type is similar in shape; if the related word segment is a semantically similar word segment, the object similarity type is semantic similarity.

In a possible example, the computer device may determine the product value of the type probability and the correct probability as the correct score of the candidate information. For example, based on the target model, the correct probability of the candidate information is P3, the wrong type probability of the query information is P2, and the type probability of the matching wrong type is extracted from P2, then the correct score P4 can be expressed as: P4=P3×P2[ The matching error type corresponding to the candidate information], wherein, P2[the matching error type corresponding to the candidate information] represents the type probability of the matching error type in P2.

There can be one or more similar types between the associated word segment that the candidate information can include and the replaced word segment; At least two of the fragments overlap. For example, the pinyin-like word segment and the font-like word segment of the replaced word segment are related word segments M; for example, the pinyin-like word segment and the font-like word segment of the query information "coupon" both include "coupon".

In a possible example, the candidate information includes a related word segment, and the replaced word segments do not overlap with each other, and the candidate information corresponds to a matching error type, then directly obtain the Type probability of matching wrong type for query information. The associated word segment is one of the pinyin-like word segment, the shape-like word segment or the semantically similar word segment of the replaced word segment. For example, for the query information "ABXXX" and its corresponding candidate information "CDXXX", the source of "CD" only includes pinyin similar word segments obtained based on the pinyin query index, that is, "CD" is the first word segment "AB ", then obtain the error probability of similar phonetic errors in the error type probability of the query information "ABXXX".

In another possible example, a piece of candidate information includes an associated word segment, and there are at least two similarity types between the associated word segment and the replaced word segment, that is, the pinyin-like word segment of the replaced word segment, the shape-like word At least two of the fragments or semantically similar word fragments overlap. Then the computer device determines at least two similarities corresponding to the at least two similarity types based on at least two similarity types between the associated segment and the replaced segment, and based on the target similarity of the at least two similarity levels that meet the threshold condition type, to get the type probability of the matching error type matching with this target similar type. Wherein, those that meet the threshold condition may include but are not limited to: the similarity is higher than a certain threshold, the similarity is the largest, the similarity is higher than the minimum threshold and does not exceed the maximum threshold, and the like. For example, the similarity is above a certain threshold of 60%.

For example, for the query information "ABXXX" and its corresponding candidate information "JKXXX", "JK" is the pinyin-like word segment and the font-like segment of the first word segment "AB", wherein the "JK" and "AB" If the pinyin similarity is 91% and the font similarity is 80%, then based on the pinyin similarity of 91%, the type probability of the phonetic similarity error matching with the phonetic similarity is obtained.

The above two examples are only described by taking the candidate information including a related word segment as an example. Of course, one piece of candidate information may also include at least two associated word segments, and there are at least one or two similar types between the at least two associated word segments and the replaced word segment. Then the type probability can be determined in combination with at least one or two similar types between the at least two associated word segments and the replaced word segment respectively. In a possible example, the type probability P0 can be calculated based on Formula 1:

Formula 1: P0＝∑x _i p _i ;

Wherein, P0 is the type probability; i is the ith associated word segment in this at least two associated word segments of candidate information; x _i is the corresponding similarity (such as pinyin) of the similar type between the ith associated word segment and the replaced word segment similarity, font similarity or word content similarity); p _i is the type probability of the matching error type matching with the i-th associated word segment corresponding to the similar type.

In another possible implementation, the correction factor can also be used to calculate the correct score, and the error habits of each user are different; for example, a certain user often has certain fixed types of errors; therefore, based on each user's For error habits, a personalized target relationship is configured for each user, and the target relationship includes the correspondence between error types and correction factors. Then this step may also include: the computer device may obtain a correction factor corresponding to the matching error type from the target association relationship matched with the user based on the matching error type, and calculate the relationship between the type probability, the correction factor, and the correct probability The product value of is determined as the correct score of the candidate information.

Step 3072: Based on the correct score of the at least one candidate information and the first error indication information, the computer device screens out corrected information whose correct score meets the target condition from the at least one candidate information.

The target condition may include but not limited to: the correct score is greater than the correct probability of the query information, the correct score exceeds the target score threshold, the correct score is the largest, and the correct score is greater than the correct probability corresponding to the error type probability of the query information (that is, (1-P2 ))wait. The computer device can use the candidate information whose final correct score satisfies the target condition as the corrected information, and subsequently search based on the corrected information.

In order to more clearly introduce the execution process of the embodiment of the present application, the following uses Figure 4 as an example to illustrate. Figure 4 is a flow chart of query error correction provided by the embodiment of the present application. For example, when a user searches on a video platform, the user inputs a user Query, and the computer device performs personalized error judgment on the user Query input by the user. For example, the user group model is used to determine the correct probability and error type probability of the user Query. When there is an error in the user Query, perform personalized error correction on the user Query and return the corrected user Query for searching. Among them, the personalized data can be used to construct the user group model, for example, the user group model can be obtained through the training of the user's historical query information. When performing personalized error correction, for immature users and mature users, at least one of the personalized error correction data of the user group and the user's personalized error correction data can be used for correction. Of course, it can also be combined with general error correction data. In addition, for users whose user maturity is new, general error correction data can be used for correction.

It should be noted that this application builds personalized query error correction models, error correction data, and error correction strategies for users with different interests to improve the error correction functions of target applications such as video platforms, search engines, social applications, and live broadcast applications. The ability to satisfy query error correction, more intelligently and accurately correct the query information entered by the user, further reduce the re-input cost when the user incorrectly enters the query information to retrieve information such as videos, articles, and commodities, and improves the search function of the target application availability. Especially for video platforms, the number of users is relatively large, and the interest distribution of different users or different user groups is different; there are great differences among different users in terms of the stars they follow, the types of videos they like, and the topics discussed. The search behavior of the video platform is a means for users to subjectively discover video-related information, which is bound to be affected by the user's subjective interests. However, in related technologies, some error correction schemes for users' video search queries fail to make full use of the personalization of video platform users. interest, which leads to the inability to personalize the error correction problem of users with different interests; and the user personalized video search query error correction implementation method proposed in this application, based on the user interests of the video platform, builds a personalized error judgment model and error correction for users with different interests The data enables the query error correction function of the video platform to meet the individual needs of thousands of people, more accurately judge and correct errors for the video search query input by the user, and greatly avoid deletion, Re-entry improves the user's personalized search efficiency and improves the actual accuracy of subsequent searches.

Through the above step 307, personalized error determination and personalized correction of query information of users with different personalization levels are realized, and user experience of video search is improved. And based on the error type probability, the similarity type matching is performed on the associated word and the replaced word, so that based on the type probability of the matching error type, the correct probability of the candidate information can be directional and personalized, and the accuracy of the corrected information obtained is improved. Accuracy, to further improve the accuracy of the query.

By determining the first error indication information of the query information based on the target model matched with the user, the target model is used to determine whether the information input into the model is wrong based on the user's interest data, so that the user can be personalized and wrongly judged based on the user's interest data , so that the judgment result more matches the user's actual interest needs; and when the query information is wrong, based on the target error correction data that matches the user, determine at least one candidate information associated with the query information, and subsequently determine the candidate information based on the target model The second error indication information is used to further filter out the correction information; and the target error correction data includes at least the user's interest data; thus, the query information can be personalized error correction based on the user's interest angle, ensuring that the correction information obtained after correction is more accurate Matching the user's real query intent can meet the error correction needs of thousands of people, improve the actual accuracy of the error judgment process and correction process, and then improve the actual accuracy of query error correction.

FIG. 5 is a schematic structural diagram of a query error correction device provided by an embodiment of the present application. As shown in Figure 5, the device includes:

The first determination module 501 is configured to, in response to receiving the query information entered into the search box by the user, determine the first error indication information of the query information based on the target model matching the user, and the target model is used to The interest data of the application determines whether the information input into the target model is wrong;

A candidate determination module 502, configured to respond to the first error indication information indicating that the query information is wrong, and determine at least one candidate information associated with the query information based on the target error correction data matching the user and the query information, the target The error correction data includes at least the interest data of the user in the target application;

A second determining module 503, configured to determine second error indication information of the at least one candidate information based on the target model;

A correction module 504, configured to determine correction information of the query information from the at least one candidate information based on the second error indication information.

In a possible implementation manner, the candidate determination module 502 includes:

An associated word acquisition unit, configured to acquire at least one associated word segment corresponding to the at least one word segment from the target error correction data based on at least one word segment included in the query string;

The candidate acquisition unit is configured to replace at least one corresponding word segment in the query information based on the at least one associated word segment, to obtain the at least one candidate information.

In a possible implementation manner, the target error correction data further includes a query index of the interest data, and the query index is used to index similar segments in the interest data;

The associated word acquisition unit is used for at least one of the following:

For each word segment, based on the pinyin and the pinyin query index of the word segment, query at least one pinyin similar word segment of the word segment from the target error correction data, the query index includes a pinyin query index, and the pinyin query index is used for Index the word fragments with similar pinyin in the interest data;

Based on the stroke sequence and stroke query index of the word segment, query at least one similar word segment of the word segment from the target error correction data, the query index includes a stroke query index, and the stroke query index is used to index the interest data Word fragments with similar strokes;

Based on the word segment and the semantic query index, at least one semantically similar word segment of the word segment is queried from the target error correction data, the query index includes a semantic query index, and the semantic query index is used to index the keyword semantics in the interest data Similar word fragments.

In a possible implementation manner, the first error indication information includes a correct probability and an error type probability of the query information, and the error type probability is used to indicate the possibility of at least one type of error in the query information; 2. The error indicates the correct probability that the information includes the candidate information;

The correction module 504 includes:

A correction unit, configured to correct the correct probability of the at least one candidate information based on the wrong type probability of the query information, to obtain the correct score of the at least one candidate information;

A screening unit, configured to, based on the correct score of the at least one candidate information and the first error indication information, filter out corrected information whose correct score meets the target condition from the at least one candidate information.

In a possible implementation manner, the candidate information includes an associated word segment replacing a word segment in the query information;

The correcting unit is configured to obtain a match matching the similar type from the probability of at least one error type included in the error type probability based on the similar type between the associated word segment and the replaced word segment in the query information A type probability of the wrong type; based on the correct probability of the candidate information and the type probability, a correct score of the candidate information is determined.

In a possible implementation manner, the similarity type includes: at least one of similarity in pronunciation, similarity in shape, and similarity in semantics;

Wherein, the matching error types with similar phonetics include similar phonetic errors, the matching error types with similar fonts include similar font errors, and the matching error types with similar semantics include at least one of similar word content errors, missing errors, or redundant errors. one item.

In a possible implementation manner, the correction unit is used for any of the following:

determining the product value between the type probability and the correct probability as the correct score of the candidate information;

Based on the matching error type, the correction factor corresponding to the matching error type is obtained from the target association relationship matched with the user, and the product value among the type probability, correction factor and correct probability is determined as the correctness of the candidate information. score, the target correlation includes the correspondence between error types and correction factors.

In a possible implementation manner, the first error indication information includes a correct probability and an error type probability of the query information; the first determination module is configured to extract at least one description information of the query information through the target model, The at least one descriptive information includes at least one item of pinyin, strokes, position codes, keywords included in the query information, or phrases included in the query information; based on at least one description information of the query information, the Error judgment is performed on the query information to obtain the correct probability and error type probability of the query information.

In a possible implementation manner, the target model includes a user group model, and the user group model is used to determine whether the information input to the user group model is wrong based on the interest data set of the user group, and the user group includes at least two of the user. users with similar interests;

The device also includes a model training module, the model training module is used to obtain the historical query information of at least two users in the user group, and generate a negative sample including noise data by performing a noise generation operation on part of the information in the historical query information data; the negative sample data is predicted by the initial model to obtain the predicted sample data; based on the similarity between the predicted sample data and the positive sample data of the historical query information, the model parameters of the initial model are adjusted until Stop adjusting when the initial model meets the model training conditions, and obtain the user group model;

Wherein, the noise generation operation includes at least one of replacing a word segment operation, deleting a word segment operation or adding a word segment operation, and the replacing word segment operation includes replacing with a phonetic similar word segment, replacing with a font similar word segment Or replace it with at least one item in the semantically similar word segment.

In a possible implementation manner, the device also includes:

The user maturity determination module is configured to determine the user maturity of the user in response to receiving the query information input by the user into the search box, and the user maturity is used to indicate the activity degree of the user in the target application;

The target model determining module is configured to determine at least one target model matching the user based on a user group having interests similar to the user in response to the user's maturity exceeding a maturity threshold.

In a possible implementation manner, the target model determination module is used for at least one of the following:

Based on a user group with similar interests to the user, determine a user group model corresponding to the user group as a target model matching the user;

Based on a user group with similar interests to the user, determine the general model and the user group model as target models matching the user;

Wherein, the general model is used to determine whether the information input into the general model is wrong based on the global user data set of the target application.

In a possible implementation, the user maturity determination module is configured to determine the user's active time in the target application; based on the user's interaction operation object associated label, count the user's interest in the active time The number of tags; based on the active time and the number of tags of interest, determine the maturity of the user.

In a possible implementation manner, the device further includes at least one of the following:

The first query module is configured to receive a query string input by the user in the search box of the target application, and use the query string as the query information;

The second query module is used to receive the query image input by the user through the search box, and extract the character string included in the query image, and use the character string extracted from the query image as the query information;

The third query module is used to receive the emoticon input by the user through the search box, and obtain the text description information of the emoticon, and use the text description information of the emoticon as the query information;

The fourth query module is used to receive the query string and the emoticon input by the user through the search box, and obtain the text description information of the emoticon, and use the query string and the text description information of the emoticon as the query information.

In a possible implementation manner, the device further includes an error correction data determination module, configured for any of the following:

Based on the user group of the user, the error correction data of the user group is determined as the target error correction data matching the user, and the error correction data of the user group includes the interest data set of the user group;

Based on the user group of the user, determine the error correction data of the user and the error correction data of the user group as the target error correction data, and the error correction data of the user includes the interest data of the user;

Based on the user group of the user, determine the error correction data and the general error correction data of the user group as the target error correction data, and the general error correction data includes the interest data set of the global users of the target application;

Based on the user group of the user, the error correction data of the user, the error correction data of the user group and the common error correction data are determined as the target error correction data.

By determining the first error indication information of the query information based on the target model matched with the user, the target model is used to determine whether the information input into the model is wrong based on the user's interest data, so that the user can be personalized and wrongly judged based on the user's interest data , so that the judgment result more matches the user's actual interest needs; and when the query information is wrong, based on the target error correction data that matches the user, determine at least one candidate information associated with the query information, and subsequently determine the candidate information based on the target model The second error indication information is used to further filter out the correction information; and the target error correction data includes at least the user's interest data; thus, the query information can be personalized error correction based on the user's interest angle, ensuring that the correction information obtained after correction is more accurate Matching the user's real query intent can meet the error correction needs of thousands of people, improve the actual accuracy of the error judgment process and correction process, and then improve the actual accuracy of query error correction.

The query error correction device in this embodiment can execute the query error correction method shown in the above embodiments of the present application, and its realization principle is similar, so it will not be repeated here.

FIG. 6 is a schematic structural diagram of a computer device provided in an embodiment of the present application. As shown in Figure 6, the computer device includes: a memory and a processor; at least one program, stored in the memory, for being executed by the processor, compared with related technologies, can realize:

By determining the first error indication information of the query information based on the target model matched with the user, the target model is used to determine whether the information input into the model is wrong based on the user's interest data, so that the user can be personalized and wrongly judged based on the user's interest data , so that the judgment result more matches the user's actual interest needs; and when the query information is wrong, based on the target error correction data that matches the user, determine at least one candidate information associated with the query information, and subsequently determine the candidate information based on the target model The second error indication information is used to further filter out the correction information; and the target error correction data includes at least the user's interest data; thus, the query information can be personalized error correction based on the user's interest angle, ensuring that the correction information obtained after correction is more accurate Matching the user's real query intent can meet the error correction needs of thousands of people, improve the actual accuracy of the error judgment process and correction process, and then improve the actual accuracy of query error correction.

A computer device is provided in an optional embodiment. As shown in FIG. 6 , the computer device 600 shown in FIG. 6 includes: a processor 601 and a memory 603 . Wherein, the processor 601 is connected to the memory 603 , such as through a bus 602 . Optionally, the computer device 600 may further include a transceiver 604, which may be used for data interaction between the computer device and other computer devices, such as sending and/or receiving data. It should be noted that, in practical applications, the transceiver 604 is not limited to one, and the structure of the computer device 600 does not limit the embodiment of the present application.

Processor 601 can be CPU (Central Processing Unit, central processing unit), general purpose processor, DSP (Digital Signal Processor, data signal processor), ASIC (Application Specific Integrated Circuit, application specific integrated circuit), FPGA (Field Programmable Gate Array, field programmable gate array) or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various illustrative logical blocks, modules and circuits described in connection with the present disclosure. The processor 601 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, and the like.

Bus 602 may include a path for communicating information between the components described above. The bus 602 may be a PCI (Peripheral Component Interconnect, Peripheral Component Interconnect Standard) bus or an EISA (Extended Industry Standard Architecture, Extended Industry Standard Architecture) bus, etc. The bus 602 can be divided into address bus, data bus, control bus and so on. For ease of representation, only one thick line is used in FIG. 6 , but it does not mean that there is only one bus or one type of bus.

Memory 603 can be ROM (Read Only Memory, read-only memory) or other types of static storage devices that can store static information and instructions, RAM (Random Access Memory, random access memory) or other types of memory that can store information and instructions Dynamic storage devices can also be EEPROM (Electrically Erasable Programmable Read Only Memory, Electrically Erasable Programmable Read-Only Memory), CD-ROM (Compact DiscRead Only Memory, CD-ROM) or other CD-ROM storage, CD-ROM storage (including compact CD-ROM, laser discs, compact discs, digital versatile discs, blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer media, but not limited thereto.

The memory 603 is used to store the application program code (computer program) for executing the solution of the present application, and the execution is controlled by the processor 601 . The processor 601 is configured to execute the application program code stored in the memory 603, so as to implement the content shown in the foregoing method embodiments.

Wherein, the computer equipment includes, but is not limited to: a server, a terminal, or any electronic equipment that supports query and error correction functions.

An embodiment of the present application provides a computer-readable storage medium, and a computer program is stored on the computer-readable storage medium. When it is run on a computer device, the computer device can execute the query error correction method in the foregoing method embodiments. Corresponding content.

An embodiment of the present application provides a computer program product or computer program, where the computer program product or computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the above query error correction method.

The terms "first", "second", "third", "fourth", "1", "2", etc. (if any) in the description and claims of this application and the above drawings are used for Distinguishes between similar objects and does not necessarily describe a particular order or sequence. It is to be understood that the data so used are interchangeable under appropriate circumstances such that the embodiments of the application described herein can be practiced in sequences other than those illustrated or described in writing.

It should be understood that although the various steps in the flow chart of the accompanying drawings are displayed sequentially according to the arrows, these steps are not necessarily executed sequentially in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in the flowcharts of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages may not necessarily be executed at the same time, but may be executed at different times, and the order of execution is also It is not necessarily performed sequentially, but may be performed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

The above descriptions are only part of the embodiments of the present invention. It should be pointed out that those skilled in the art can make some improvements and modifications without departing from the principles of the present invention. It should be regarded as the protection scope of the present invention.

Claims (18)

1. A query error correction method, characterized in that the method comprises: In response to receiving query information entered into the search box by the user, determining first error indication information of the query information based on a target model matched with the user, the target model being used to base the user's interest data in the target application determining whether the information input into the target model is wrong; In response to the first error indication information indicating that the query information is wrong, determining at least one candidate information associated with the query information based on the target error correction data matched with the user and the query information, the target correction The error data at least includes the interest data of the user in the target application; determining second error indication information for the at least one candidate information based on the target model; Correction information for the query information is determined from the at least one candidate information based on the second error indication information. 2. The query error correction method according to claim 1, wherein the query information includes a query character string, and the determination is based on the target error correction data matched with the user and the query information, At least one candidate information associated with the query information, including: Based on at least one word segment included in the query string, at least one associated word segment corresponding to the at least one word segment is obtained from the target error correction data; Based on the at least one associated word segment, replace at least one corresponding word segment in the query information to obtain the at least one candidate information. 3. The query error correction method according to claim 2, wherein the target error correction data also includes a query index of the data of interest, and the query index is used to index similar segments in the data of interest; The obtaining at least one associated word segment corresponding to the at least one word segment from the target error correction data based on at least one word segment included in the query string includes at least one of the following: For each word segment, based on the pinyin of the word segment and a pinyin query index, query at least one pinyin-like word segment of the word segment from the target error correction data, the query index includes a pinyin query index, the The pinyin query index is used to index word fragments with similar pinyin in the interest data; Based on the stroke sequence and stroke query index of the word segment, query at least one font-similar word segment of the word segment from the target error correction data, the query index includes a stroke query index, and the stroke query index is used for Indexing word segments with similar strokes in the interest data; Based on the word segment and the semantic query index, at least one semantically similar word segment of the word segment is queried from the target error correction data, the query index includes a semantic query index, and the semantic query index is used to index the Word fragments with semantically similar keywords in interest data. 4. The query error correction method according to claim 1, wherein the first error indication information includes the correct probability and error type probability of the query information, and the error type probability is used to indicate the query information there is a probability of an error of at least one error type; the second error indication information includes a probability of correctness of the candidate information; The determining correction information of the query information from the at least one candidate information based on the second error indication information includes: Correcting the correct probability of the at least one candidate information based on the wrong type probability of the query information to obtain the correct score of the at least one candidate information; Based on the correct score of the at least one candidate information and the first error indication information, the corrected information whose correct score meets the target condition is screened out from the at least one candidate information. 5. The query error correction method according to claim 4, wherein the candidate information comprises an associated word segment replacing the word segment in the query information; The correcting the correct probability of the at least one candidate information based on the error type probability of the query information to obtain the correct score of the at least one candidate information includes: Based on the similarity type between the associated word segment and the replaced word segment in the query information, from the probability of at least one error type included in the error type probability, a matching error type matching the similar type is obtained type probability; A correct score for the candidate information is determined based on the correct probability of the candidate information and the type probability. 6. The query error correction method according to claim 5, wherein the similarity types include: at least one of similarity in pronunciation, similarity in font shape, and similarity in semantics; Wherein, the matching error types with similar phonetics include similar phonetic errors, the matching error types with similar fonts include similar font errors, and the matching error types with similar semantics include similar word content errors, missing errors, or redundant errors. At least one of the . 7. The query error correction method according to claim 5, wherein said determining the correct score of said candidate information based on the correct probability of said candidate information and said type probability comprises any of the following: determining the product value between the type probability and the correct probability as the correct score of the candidate information; Based on the type of matching error, the correction factor corresponding to the type of matching error is obtained from the target association relationship matched with the user, and the product value of the probability of the type, the correction factor, and the correct probability is determined as the The correct score of the candidate information, the target correlation includes the correspondence between error types and correction factors. 8. The query error correction method according to claim 1, wherein the first error indication information includes the correct probability and the error type probability of the query information; the target model based on the user matching, Determining the first error indication information of the query information includes: Through the target model, at least one kind of descriptive information of the query information is extracted, the at least one kind of descriptive information includes pinyin, strokes, position codes of the query information, keywords included in the query information or the query information at least one of the phrases included in the message; Based on at least one description information of the query information, an error judgment is performed on the query information to obtain a correct probability and a wrong type probability of the query information. 9. The query error correction method according to claim 8, wherein the target model includes a user group model, and the user group model is used to determine the information input into the user group model based on the interest data set of the user group Whether it is wrong, the user group includes at least two users with similar interests of the user; The training process of the user group model includes: Obtain historical query information of at least two users in the user group, and generate negative sample data including noise data by performing a noise generation operation on part of the information in the historical query information; Wherein, the noise generating operation includes at least one of replacing a word segment operation, deleting a word segment operation, or adding a word segment operation, and the replacing word segment operation includes replacing with a phonetic similar word segment, replacing with a font similar word segment or replaced by at least one of the semantically similar word fragments; Predicting the negative sample data through the initial model to obtain predicted sample data; Based on the similarity between the predicted sample data and the positive sample data of the historical query information, adjust the model parameters of the initial model until the initial model meets the model training conditions and stop adjusting, and obtain the user group model. 10. The query error correction method according to claim 1, wherein, before determining the first error indication information of the query information based on the target model matched with the user, the method further comprises: In response to receiving query information input by the user into the search box, determine the user maturity of the user, where the user maturity is used to indicate the activity of the user in the target application; In response to when the user maturity exceeds a maturity threshold, at least one target model matching the user is determined based on a user group having interests similar to the user. 11. The query error correction method according to claim 10, wherein said determining at least one target model matching said user based on a user group similar in interest to said user comprises at least one of the following: Determining a user group model corresponding to the user group as a target model matching the user based on a user group with similar interests to the user; determining a general model and the user group model as a target model matching the user based on a user group having interests similar to the user; Wherein, the general model is used to determine whether the information input into the general model is wrong based on the global user data set of the target application. 12. The query error correction method according to claim 10, wherein said determining the user maturity of said user comprises: determining the active time of the user in the target application; Based on the operation object associated tags of the user's interactive operation, counting the number of interest tags of the user within the active time; Based on the active time and the number of interest tags, the user maturity is determined. 13. The query error correction method according to claim 1, characterized in that, in response to receiving the query information entered into the search box by the user, based on the target model matched with the user, the first element of the query information is determined. Before the error indication information, the method also includes at least one of the following: receiving a query string input by the user in the search box of the target application, and using the query string as the query information; receiving a query image input by the user through the search box, extracting a character string included in the query image, and using the character string extracted from the query image as the query information; receiving the emoticon input by the user through the search box, and obtaining the text description information of the emoticon, and using the text description information of the emoticon as the query information; receiving the query string and the emoticon input by the user through the search box, and obtaining the text description information of the emoticon, and using the query string and the text description information of the emoticon as the query information . 14. The query error correction method according to claim 1, characterized in that, before determining at least one candidate information of the query information based on the query information and the target error correction data matched with the user, the Said method also includes any of the following: Based on the user group of the user, determining the error correction data of the user group as the target error correction data matching the user, the error correction data of the user group includes the interest data set of the user group; Based on the user group of the user, determining the error correction data of the user and the error correction data of the user group as the target error correction data, and the error correction data of the user includes the interest data of the user; Based on the user group of the user, determining the error correction data and general error correction data of the user group as the target error correction data, the general error correction data including the interest data set of global users of the target application; Based on the user group of the user, the error correction data of the user, the error correction data of the user group and the common error correction data are determined as the target error correction data. 15. A query error correction device, characterized in that the device comprises: The first determination module is configured to determine first error indication information of the query information based on a target model matched with the user in response to receiving query information input by the user into the search box, and the target model is used to determine the first error indication information of the query information based on the target model The user's interest data in the target application determines whether the information input into the target model is wrong; A candidate determination module, configured to determine at least one candidate associated with the query information based on the target error correction data matching the user and the query information in response to the first error indication information indicating that the query information is wrong Information, the target error correction data at least includes the interest data of the user in the target application; A second determining module, configured to determine second error indication information of the at least one candidate information based on the target model; A correction module, configured to determine correction information of the query information from the at least one candidate information based on the second error indication information. 16. A computer device comprising a memory, a processor and a computer program stored on the memory, wherein the processor executes the computer program to realize the query error correction described in any one of claims 1 to 14 method. 17. A computer-readable storage medium, on which a computer program is stored, wherein, when the computer program is executed by a processor, the query error correction method according to any one of claims 1 to 14 is implemented. 18. A computer program product, comprising a computer program, characterized in that, when the computer program is executed by a processor, the query error correction method according to any one of claims 1 to 14 is implemented.

Images