[go: up one dir, main page]

WO2022112857A1 - Procédé et appareil permettant de corriger des informations de commande, dispositif, et support de stockage - Google Patents

Procédé et appareil permettant de corriger des informations de commande, dispositif, et support de stockage Download PDF

Info

Publication number
WO2022112857A1
WO2022112857A1 PCT/IB2021/055848 IB2021055848W WO2022112857A1 WO 2022112857 A1 WO2022112857 A1 WO 2022112857A1 IB 2021055848 W IB2021055848 W IB 2021055848W WO 2022112857 A1 WO2022112857 A1 WO 2022112857A1
Authority
WO
WIPO (PCT)
Prior art keywords
information
order
corrected
target
text
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/IB2021/055848
Other languages
English (en)
Chinese (zh)
Inventor
张斌
彭佳玮
陈凯歌
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Sensetime International Pte Ltd
Original Assignee
Sensetime International Pte Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Sensetime International Pte Ltd filed Critical Sensetime International Pte Ltd
Publication of WO2022112857A1 publication Critical patent/WO2022112857A1/fr
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/279—Recognition of textual entities
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/29—Geographical information databases
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/3331—Query processing
    • G06F16/334—Query execution
    • G06F16/3344—Query execution using natural language analysis
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/0601—Electronic shopping [e-shopping]
    • G06Q30/0633—Managing shopping lists, e.g. compiling or processing purchase lists
    • G06Q30/0635—Managing shopping lists, e.g. compiling or processing purchase lists replenishment orders; recurring orders
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/14—Travel agencies

Definitions

  • the present disclosure relates to computer vision technology, and in particular, to a method, apparatus, device, and storage medium for correcting order information. Background technique
  • OCR Optical Character Recognition, Optical Character Recognition
  • An embodiment of the present disclosure provides a correction solution for order information.
  • a method for correcting order information comprising: obtaining order information to be corrected according to a text recognition result of the order; determining target search information from the text recognition result ; Acquiring order reference information matching the target search information; and correcting the order information to be corrected by using the order reference information to obtain target order information.
  • an apparatus for correcting order information includes: an acquiring unit, configured to acquire order information to be corrected according to a text recognition result of the order; a determining unit, using for determining target search information from the text recognition result; a matching unit for acquiring order reference information matching the target search information; a correcting unit for correcting the to-be-corrected order information by using the order reference information, to get the target order information.
  • an electronic device the device includes a memory and a processor, where the memory is used for storing computer instructions that can be executed on the processor, and the processor is used for The method for correcting order information described in the first aspect is implemented when the computer instructions are executed.
  • a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the method for correcting order information described in the first aspect is implemented.
  • a computer program including computer-readable codes, and when the computer-readable codes are executed in an electronic device, a processor in the electronic device executes When implementing the correction method for order information described in the first aspect.
  • the order information correction method, device, device, and storage medium obtain order information to be corrected according to the text recognition result of the order, and determine target search information from the text recognition result, Acquiring order reference information matching the target search information, and using the order reference information to correct the order information to be corrected to obtain target order information, can quickly obtain accurate target order information from the text recognition result of the order.
  • FIG. 1 is a flowchart of a method for correcting order information provided by at least one embodiment of the present disclosure
  • FIG. 2 is a schematic structural diagram of a setting database in a method for correcting order information proposed by at least one embodiment of the present disclosure
  • Figures 3A, 3B, and 3C are schematic diagrams of an information extraction method proposed by at least one embodiment of the present disclosure.
  • FIG. 4 is a schematic diagram of a device for correcting order information proposed by at least one embodiment of the present disclosure
  • FIG. 5 is a schematic structural diagram of an electronic device proposed by at least one embodiment of the present disclosure. Detailed ways
  • FIG. 1 shows a flowchart of a method for correcting order information according to some embodiments of the present disclosure. As shown in FIG. 1 , the method includes steps 101 to 104.
  • step 101 the order information to be corrected is obtained according to the text recognition result of the order.
  • the order for text recognition includes at least one of the following: an order image, an order in the form of an electronic document, such as a pdf document. Those skilled in the art should understand that the order may also include other types suitable for text recognition.
  • the text box contained in the order can be obtained by performing text detection on the order; and the text in the text box can be recognized by performing text recognition on the obtained text box, so that Obtain a text recognition result; and directly perform text recognition on the order to be processed, such as OCR, to obtain a text recognition result containing a text box in the order.
  • the embodiments of the present disclosure do not limit the specific method for obtaining the text recognition result.
  • the order information to be corrected is the order information to be corrected obtained from the text recognition result of the order according to the set rule.
  • the order information to be corrected includes address information
  • the address information to be corrected can be obtained from the text recognition result according to the rules of the address information.
  • target search information is determined from the text recognition result.
  • the target search information is information that is related to the order information to be corrected or can reflect the characteristics of the order information to be corrected.
  • the target search information includes a subject name of the order information to be corrected, and at least one of at least one content element.
  • the target search information may include the name of the subject to which the address information belongs (eg, name, place name, etc.) and/or at least one content element included in the address information (eg, administrative regions at all levels, the zip code corresponding to the row area, etc.).
  • step 103 obtain order reference information matching the target search information.
  • order reference information matching the target search information may be acquired from the setting database.
  • the setting database stores a plurality of reference subject names and corresponding reference information.
  • the setting database is a database that stores a plurality of subject names and corresponding address information, according to the subject name corresponding to the order information to be corrected, such as "XX Hotel/ Hotel", and zip code, that is, the matching "XX Hotel/Hotel" can be searched in the setting database, and the corresponding address information can be used as the order reference information.
  • the order reference information matching the target search information can also be obtained through the Internet.
  • a search engine can be used to search the Internet according to the subject name and zip code corresponding to the order information to be corrected, and the retrieved information corresponding to the matching subject name can be used as the order reference information.
  • the order reference information matching the target search information can also be obtained from the setting database and the Internet at the same time.
  • the order reference information is obtained from both the setting database and the Internet, any one of them, or a designated one can be used as the target order reference information; in the case that only the order reference information is obtained from the Internet, the Internet can be used
  • the search results update the settings database.
  • step 104 the order information to be corrected is corrected using the order reference information to obtain target order information.
  • the method, device, device and storage medium for correcting order information proposed by at least one embodiment of the present disclosure obtain order information to be corrected according to the text recognition result of the order, and obtain the order information to be corrected from the text recognition result.
  • Determine the target search information in the middle obtain the order reference information matching the target search information, and use the order reference information to correct the order information to be corrected to obtain the target order information, which can be quickly obtained from the text recognition result of the order.
  • the address database in the related art usually only supports the query from the subject name to the address, and only has certain error tolerance at the beginning and end of the input word. Because the method for correcting order information proposed by the embodiments of the present disclosure is to obtain matching order reference information according to the target search information determined from the text recognition result, and the target search information may be the subject name and/or the subject name in the order information to be corrected. Or at least one content element in the order information to be corrected, so even if there is erroneous information in the order information, such as even a wrong subject name, other information in the order information can be used as the target search information through this correction method, Correcting the order information to be corrected by obtaining the order reference information has high fault tolerance.
  • the method for correcting order information proposed by at least one embodiment of the present disclosure is applicable to orders of different layouts.
  • the subject name corresponding to the order information to be corrected may be obtained as the target search information.
  • the subject name and the order information to be corrected are, for example, key-value pair information, wherein the subject name indicates an attribute, and the order information to be corrected indicates a value of the attribute.
  • the order information to be corrected may be address information, the subject corresponding to the address information is the object to which the address information belongs, and the corresponding subject name is the name of the object to which it belongs.
  • the object to which the address information belongs is an individual
  • the corresponding subject name is a name; when the object to which the address information belongs is a place, the corresponding subject name is the place name.
  • the order information to be corrected may also be identity information, and the subject name corresponding to the identity information is a name. Those skilled in the art should understand that the order information to be corrected may also be other types of information, which is not limited in the present disclosure.
  • the setting database may include reference unit information of a plurality of levels, and the reference unit information of each lowest level of the plurality of levels corresponds to a plurality of reference subject names.
  • the reference unit information is organized and stored according to the hierarchical level from top to bottom, and the lower the level of the reference unit information corresponds to the smaller the scope or the lower the authority.
  • the minimum employment level is the corresponding reference information unit with the smallest scope or the lowest authority.
  • the reference unit information of multiple levels included in the setting database includes reference administrative area information and/or postal code information
  • the reference unit information of the lowest level includes the administrative area with the smallest scope. Name and/or zip code for this borough.
  • the reference unit information in the setting database may be stored in a tree structure, the non-leaf nodes of different levels store the reference unit information of different levels, and the leaf nodes are used to store the information belonging to the upper level node.
  • Reference principal name the reference unit information in the setting database may be stored in a tree structure, the non-leaf nodes of different levels store the reference unit information of different levels, and the leaf nodes are used to store the information belonging to the upper level node. Reference principal name.
  • the setting database further stores first reference information corresponding to each reference subject name.
  • the first reference information is usually complete information corresponding to the reference subject name, including reference unit information at various levels and specific reference information corresponding to the reference subject name. Taking address information as an example, the first reference information may be complete address information, including administrative area information at various levels and specific addresses corresponding to the names of reference entities, such as streets and/or units.
  • the first reference information is obtained in advance, and is the reference information corresponding to the name of the reference subject with higher reliability and accuracy.
  • the reference unit information of multiple levels in the setting database may be administrative regions of multiple levels.
  • the tree structure for storing address information can be a rooted tree structure, and the root node has no actual meaning; the child nodes of the root can be used to store the travel salesman of the order (for example, XX travel agency), and the rest of the non-leaf nodes can be used to store The administrative region composition or postal code of the country; each leaf node can store an object name, and each leaf node can also store the complete address information corresponding to the object name.
  • all non-leaf nodes are unique, and the parent node of the non-leaf node represents its own higher-level administrative district.
  • FIG. 2 is a schematic structural diagram of a setting database in a method for correcting order information proposed by at least one embodiment of the present disclosure.
  • the subtree of the traveling salesman can be constructed according to the top-to-bottom (from shallow to deep) hierarchy: country-province-city-district, in some cases, the next level of district may also include subdistricts , and each administrative region can also be replaced by a postal code, for example, it is constructed as country-province-postal code-district.
  • the zip code can be substituted for any administrative region, which is not limited in the present disclosure.
  • the reference administrative area information of each house level stored in the tree structure can be obtained from the administrative division table of each country and the correspondence table between zip codes and administrative areas published on the Internet;
  • the name of the reference subject stored in the node and the corresponding first reference information can be obtained by manual annotation.
  • the order reference information corresponding to the order information to be corrected may be obtained in the following manner.
  • the unit information of the lowest level in the order information to be corrected may be acquired as target search information according to the level division in the setting database.
  • the order information to be corrected as the address information of a hotel order as an example
  • the hierarchical division of addresses in the set database that is, the tree structure of the database
  • the information contained in the order information to be corrected can be obtained.
  • Unit information at each level For example, according to the tree structure "country-province-city-district" in the database, the order information to be corrected is split, and then the administrative area information of each level included in the address information can be obtained. Among them, the administrative area information of the lowest level can be used as the target search information.
  • the sub-area information may be used as the target search information; if the minimum administrative area included in the address information is a sub-area, the sub-area information may be used as the target search information; information as the target search information; other situations are similar and will not be repeated here.
  • the reference unit information of each lowest level in the database corresponds to multiple reference subject names, so among the multiple reference subject names, the target subject name can be determined according to preset conditions.
  • the subject name corresponding to the order information to be corrected may be matched with a plurality of reference subject names corresponding to the target unit information respectively; and the matching score is the highest and exceeds the first set threshold
  • the reference principal name of identified as the target principal name.
  • the reference information corresponding to each reference subject name is stored in the preset library, according to the first reference information corresponding to the determined target subject name, the information of the pending order information can be obtained.
  • Order reference information >> The reference information stored in the setting database has high reliability and accuracy, and more accurate target order information can be obtained by using the reference information to correct the order information to be corrected.
  • the setting database stores second reference information corresponding to each reference subject name.
  • the second reference information is other reference information other than the reference unit information of each level, and is generally more specific information than the reference unit information of each level.
  • the second reference information may be, for example, the street and/or unit where the hotel is located.
  • the second reference information is obtained in advance, and is the reference information corresponding to the name of the reference subject with high reliability and accuracy.
  • the method of determining the name of the target subject is similar to the above method, the difference is that after the target subject is determined After the name, the order reference information corresponding to the order information to be corrected is obtained according to the reference unit information of each level corresponding to the target subject name and the second reference information corresponding to the target subject name.
  • the complete information of the target subject name can be obtained, Correcting the order information to be corrected based on the complete information can obtain more accurate and complete target order information.
  • order reference information corresponding to the order information to be corrected may also be obtained from the Internet according to the target search information.
  • the Internet may be searched according to, for example, the subject name or at least one content element of the order information to be corrected.
  • One or more candidate order reference information, each candidate order reference information is matched with the to-be-corrected order information, and the candidate order reference information with the highest matching score and exceeding the second set threshold is obtained as the order reference information.
  • the target search information may include the zip code contained in the address information, and/or one of which is administrative area information.
  • a plurality of candidate address information that may be hotel addresses can be obtained from the Internet.
  • the candidate address information with the highest matching score and exceeding the second set threshold can be used as the to-be-corrected address information
  • the order reference information of the order information is corrected to obtain more accurate hotel address information.
  • any one of the candidate address information can be retained and the other candidate address information can be deleted.
  • the organizational storage of administrative regions at all levels and the zip codes corresponding to the administrative regions in the address database can be set according to the regulations of the target country, so the correction method can be easily extended to the itinerary information of any destination country 's correction.
  • retrieval may be performed first in the setting database according to the target search information, and then in the interconnection.
  • the reference information corresponding to the order information to be corrected obtained from the Internet, and the The subject name corresponding to the order information to be corrected is added to the information corresponding to the lowest-level reference unit information in the setting database, that is, the subject name is added to the reference subject name corresponding to the corresponding lowest-level reference unit information .
  • the name of the subject corresponding to the order information to be corrected and the order reference information are stored in the leaf nodes of the tree structure, which become the name of the newly added reference subject and the corresponding reference information.
  • the reference information corresponding to the to-be-corrected order information obtained from the Internet and the The subject name corresponding to the order information updates the information corresponding to the reference unit information of the lowest level in the setting database. That is, the reference information of the target subject name corresponding to the reference unit information of the lowest level in the setting database is replaced with the reference information corresponding to the order to be corrected obtained from the Internet.
  • the reference information corresponding to the order information to be corrected is replaced with the reference information corresponding to the reference subject name originally stored in the leaf nodes of the tree structure, so as to realize the Updates to reference information referring to principal names.
  • the latest update time of the reference information corresponding to the order information to be corrected obtained from the Internet may be obtained, and the determination based on the update time Whether to update the reference information of the reference subject name. For example, if the latest update time is within the set time range, such as within the last year or within the last 6 months, the update can be performed; on the contrary, if the latest update time exceeds the set time range, then the update can be performed. A prompt message is output, and the technician determines whether to update to avoid incorrect update.
  • N-grams are usually used to correct the text recognition results.
  • the training of N-grams relies on thesaurus, the thesaurus of address information, especially the overseas terminology, is usually not available. It is complete, because the correction effect of the N-gram model for the recognition results of the order text of the hotel order class is not good.
  • the hotel address information in the text recognition result of the hotel itinerary can be corrected, such as correcting wrong information in the hotel address, Or complete the incomplete hotel address, which improves the accuracy and reliability of automatic visa information filling, improves user experience, and helps speed up the approval process.
  • the correction method of the present disclosure can use the reference information obtained from the Internet for correction, or update the setting database according to the reference information obtained from the Internet, the problem of incomplete thesaurus can be solved, and better correction can be obtained. Effect.
  • the order information to be corrected includes at least address information and hotel information.
  • the order information to be corrected can be obtained from the text recognition result of the order to be processed by the following method.
  • the key information may include at least one content element of the order information to be corrected and at least one of keywords indicating the order information to be corrected.
  • the key information may include the content element "zip code" in the address information, and if the region to which the address information belongs is known, the digit of the zip code may be determined. number. Taking the order information to be corrected as an example of an address in Thailand, since the postal code of Thailand is 5 digits, it can be determined that the key information is 5 digits. In this step, a text box containing 5 digits is determined as the first text box.
  • the identified content may include more than 5 digits, for example, the text box includes 8 digits, etc.
  • a search may also be performed in the zip code list of the region to which the found zip code belongs, to confirm that the found zip code is indeed the zip code of the region to which the found zip code belongs coding.
  • the number of digits of postal codes around the world can be integrated, and the key information can be determined as a number of 4 to 9 digits. Then in this step, determine the text boxes containing 4 to 9 digits respectively, as the first text box. In a possible implementation manner, in order to reduce additional discrimination operations, a text box containing only 4-9 digits may be determined as the first text box, that is, for a text box containing 10 or more digits Not be considered.
  • the key information may also include content elements in the address information—administrative area information, such as “Thailand” or “Thailand”, then in the multiple text boxes, it can be determined to include “Thailand” or “Thailand” such as The text content of the text box as the first text box.
  • the key information further includes keywords indicating the order information to be corrected.
  • the keywords include "address”, "address”, and expressions in other languages. Keywords for the address. It should be noted that the form of the keyword is not limited in this application, for example, it may include various expressions such as full name and abbreviation.
  • the text box to be merged is determined based on the first text box.
  • the text boxes to be combined may be determined according to the positional relationship with the first text box, and the text boxes to be combined may be combined to obtain a combined text box.
  • the The order information to be corrected is extracted from the merged text box.
  • a first text box containing key information is determined in a plurality of text boxes included in the text recognition result of the order to be processed, and at least part of the text is evaluated according to the first text box.
  • the frames are merged to obtain a merged text frame, and the order information to be corrected is obtained from the merged text frame, which can implement efficient information processing in the pending order according to the key information in the information of the to-be-corrected order.
  • the text boxes may be combined in the following manner to obtain a combined text box.
  • the positional relationship includes the positional relationship between other text boxes (that is, any text box other than the first text box or a specified text box) and the first text box, for example, in the position of the first text box.
  • the distance to the first text box is also included, for example, the distance in pixels from the first text box in the vertical direction and the distance in pixels in the horizontal direction.
  • the distance between the text boxes is determined according to the distance between the center points of the two text boxes.
  • the text box whose positional relationship between each of the text boxes and the first text box belongs to the set range For example, the text box above the first text box may be determined as the second text box, or the text box whose pixel distance from the first text box in the vertical direction is within a set threshold may be determined as the second text box text boxes, etc.
  • the merging of the to-be-merged text boxes may be performed on a line basis. That is, according to the row to which each text box in the to-be-combined text box belongs, the to-be-combined text boxes are combined to obtain the combined text box.
  • Figure 3A shows an exemplary merge result. As shown in FIG. 3A , it includes multiple lines of merged text boxes, including merged text boxes 301 to 303, wherein the merged text boxes in each row are obtained by merging one or more text boxes included in the row.
  • the first threshold may be specifically determined according to the format feature of the order information to be corrected.
  • the order information to be corrected may be acquired from the merged text box according to the format feature of the order to be processed.
  • the format feature of the order to be processed includes the distance feature between each line of text, the font feature of each line of text, the positional relationship feature between texts, and so on.
  • the target direction for obtaining the order information to be corrected can be determined, and the order information to be corrected is obtained according to the target direction.
  • the order information to be corrected is address information and the key information is a zip code
  • the zip code is usually located at the end of the address information
  • the target direction for extracting the order information to be corrected can be determined, and the extraction is performed according to the target direction.
  • the order information to be corrected is address information
  • the key information is a keyword indicating the address information
  • the key word “address” since the key word “address” is usually located at the forefront of the address information, it can be determined that the order information to be corrected is located below the first text box, so that it can be determined that the order to be corrected is extracted
  • the target direction of the information is extracted according to the target direction.
  • the target direction includes a first target direction and a second target direction
  • the first target direction is used to indicate that the merge is traversed in the process of locating the area where the order information to be corrected is located
  • the direction of the text box, the second target direction is used to indicate the direction of reading the order information to be corrected from the area where the order information to be corrected is located.
  • the key information may include a keyword indicating the order information to be corrected, at least one content element of the order information to be corrected, a subject name of the order information to be corrected, and the like. Taking the order information to be corrected as address information as an example, the keywords indicating the address information include "address",
  • the key information is "10110" (zip code), starting from the first text box containing "10110", that is, from the merged text box 301 Initially, the combined text box is traversed upward until the combined text box 302 where the key information "Address” is located is found. Then take the key information "Address” as the starting position, traverse the merged text box downward until the merged text box 301 where the key information "10110" is located is found, and obtain the content traversed downward as the waiting Correct order information.
  • the uppercase and lowercase forms of some or all letters in a word are not limited, and can be adjusted according to the actual situation. That is to say, in the actual identification and other processing processes, the same processing method can be adopted for ADDRESS, Address, address, etc., that is, they are all identified as "addresses”.
  • the method further includes: obtaining a distance between adjacent merged text boxes.
  • the adjacent merged text boxes include two merged text boxes that are adjacent in the vertical direction.
  • the multiple merged text boxes obtained from the text recognition result include multiple pairs of adjacent merged text boxes.
  • the merged text boxes 311-314 include adjacent merged text boxes 311-312, adjacent merged text boxes 312-313, and adjacent merged text boxes
  • the traversing includes acquiring the text content in the combined text box, and also includes acquiring the distance between the combined text box and its adjacent combined text box, wherein the adjacent combined text box is traversing the combined text box. Iterates between merged text boxes. Next, take the first traversed merged text box as the starting position in the adjacent merged text boxes whose distance satisfies the first set condition, and traverse the merged text frame according to the second target direction until the merged text frame is found.
  • that the distance of adjacent merged text boxes satisfies the first set condition includes: the distance of the adjacent merged text boxes is greater than the first inter-frame distance threshold.
  • the key information is "10400" (zip code)
  • the first text box containing the zip code is taken as the starting position, that is, the first text box containing "10400”
  • the text box is the starting position, That is, starting from the merged text box 311, the merged text box is traversed upward.
  • traversing to the merged text box 312 includes acquiring the content in the merged text box 312 and obtaining the distance between the merged text box 312 and the merged text box 311 .
  • the distance between the two text boxes may be the pixel distance between the center points of the two text boxes in the vertical direction, or the pixel distance between the corresponding positions of the two text boxes may be used as the distance between the two text boxes , for example, in the case of left-aligned two text boxes, the corner points of the two text boxes at the upper left corner or the lower left corner can be used as the two vertices for determining the distance, and the pixels between the two vertices can be used to determine the distance. distance as the distance between two text boxes.
  • other methods similar to the above-mentioned contents can also be used to determine the distance between the two text boxes.
  • the specific implementation process is not limited in this application, and may include but not be limited to the above exemplified situations.
  • the distance between the merged text box 312 and the merged text box 311 does not satisfy the first set condition, that is, the distance between the merged text box 312 and the merged text box 311 is less than or equal to the first inter-frame distance threshold.
  • the distance between the combined text box 314 and the combined text box 313 satisfies the first set condition, that is, the distance between the combined text box 314 and the combined text box 313 is greater than the first inter-box distance threshold, then Stop traversing upwards.
  • the relationship between the first target direction and the direction to which the second target direction points respectively is not limited, that is, the first target direction and the second target direction may be at a certain angle, for example, the The first target direction and the second target direction may be opposite (ie, 180°), or may be the same (ie, 0°).
  • the first target direction may indicate a downward traversal of the merged text box, by traversing the merged text box downward until searching for Go to the last key information, or find adjacent merged text boxes whose distance satisfies the first set condition.
  • the first target direction and the second target direction are the same, and the traversal is performed again in the above-mentioned traversed area according to the second target direction, and the traversed area is obtained.
  • the content is used as the order information to be corrected.
  • the adjacent merged text box is used as the target adjacent merged text box, then the first inter-frame distance threshold corresponding to the target adjacent merged text box is determined according to at least one of the following : the height of the merged text box first traversed in the adjacent merged text boxes of the target; the distance between the merged text boxes contained in the traversed adjacent merged text boxes and the height of the merged text box first traversed.
  • the target adjacent merged text boxes are two adjacent merged text boxes for which the first inter-frame distance threshold is to be determined.
  • the first inter-frame distance thresholds corresponding to each pair of adjacent merged text boxes may be different.
  • the first inter-frame distance threshold is determined according to the height of the merged text box first traversed in the target adjacent merged text frame.
  • each merged text frame is from bottom to top
  • the adjacent merged text boxes 311 and 312 are first traversed adjacent merged text boxes in this example, and the first inter-frame distance threshold corresponding to the two can be determined according to the height of the merged text box 311 .
  • the first inter-box distance threshold is set to 0.65*mean_heightl (the height of the merged text box 311).
  • the first inter-frame distance threshold may be based on the traversed adjacent merged text boxes included The distance between the included merged text boxes and the height of the first traversed merged text box are determined.
  • the first traversed merged text box is the first traversed merged text box in the process of locating the region where the order information to be corrected is located.
  • the first inter-frame distance threshold corresponding to the target adjacent merged text boxes may be determined by: obtaining the updated inter-frame distance of the target adjacent merged text boxes, and the updated inter-frame distances The distance is obtained by weighted summation of the distances between the merged text boxes included in the reference adjacent merged text boxes and the updated inter-frame distances between the merged text boxes included in the reference adjacent merged text boxes, wherein , the reference adjacent text frame is the adjacent merged text frame closest to the target merged text frame; the update disturbance value of the target adjacent merged text frame is obtained, and the updated disturbance value is obtained by comparing all the first traversed The absolute value of the disturbance value of the adjacent merged text box and the distance difference value are obtained by weighted summation, wherein the distance difference value is the updated inter-frame distance of the target adjacent merged text frame and the reference adjacent merge.
  • the difference between the distances between the merged text boxes included in the text box, the disturbance value is determined according to the height of the merged text box that is first traversed; according to the distance between the update boxes and the update disturbance value, it is determined that the target is adjacent The first inter-box distance threshold for merging text boxes.
  • the text box closest to the region where the extracted target region is located is the text box of the subject name corresponding to the order information to be corrected.
  • the text box above the extracted address information is the name of the hotel, the subject of the address information. The same is true for documents such as business cards and shopping orders.
  • the text box closest to the area where the address information, identity information, etc. are located is the text box where the name of the subject of the information is located.
  • the subject name corresponding to the order information to be corrected may be determined by the following method.
  • the content contained in the merged text boxes 321-322 is the order information to be corrected extracted according to the method for correcting order information described in any embodiment of the present disclosure.
  • the area where the merged text boxes 321-322 are located is determined as the area where the order information to be corrected is located.
  • the region where the order information to be corrected is located is in the first target direction (the direction of search traversal, in In this example, it is up) and the closest merged text box is 323 (there is a non-target language text between the merged text box 322 and the merged text box 323, as shown in gray, which is ignored).
  • the merged text box 323 is traversed upward. Since the distance between the adjacent merged text boxes above the merged text box 323 and the merged text box 323 exceeds the second inter-frame threshold, that is, the second set condition is satisfied (there is no other merged text above the merged text box 323 .
  • the second setting condition is met
  • the text box 323 is merged, so that the content "XXXXX Hotel" in the merged text box can be determined as the name of the subject of the order information to be corrected, that is, "XXXXX Hotel" is determined as the name of the subject of the order information to be corrected.
  • the distance threshold between the second boxes can be set to 0.4*mean_height (the average of adjacent merged text boxes high).
  • the second inter-box distance threshold can be set to 0.6*mean_height (adjacent merged the average height of the text box).
  • the information extraction method proposed by any embodiment of the present disclosure can be applied to images or electronic documents of various formats, and various formats include at least one of the following: a hotel order, an airplane itinerary, a passport, an ID card, etc.
  • the electronic document may be a pdf document.
  • FIG. 4 is an apparatus for correcting order information provided by at least one embodiment of the present disclosure.
  • the apparatus includes: an obtaining unit 401 for obtaining order information to be corrected according to a text recognition result of the order; a determining unit 402 for obtaining order information from The target search information is determined in the text recognition result; the matching unit 403 is used to obtain order reference information matching the target search information; the correction unit 404 is used to correct the order information to be corrected by using the order reference information, to get the target order information.
  • the target search information includes at least one of the following: the target search information includes a subject name of the order information to be corrected and at least one content element of the order information to be corrected. at least one.
  • the matching unit is specifically used for at least one of the following: obtaining order reference information matching the target search information from the setting database; obtaining the target search information through the Internet Matching order reference information.
  • the setting database includes reference unit information of a plurality of levels, and the reference unit information of the lowest level in the plurality of levels corresponds to a plurality of reference subject names.
  • the setting database stores the first reference information corresponding to the name of the reference subject; the determining unit is specifically configured to: obtain the to-be-corrected according to the hierarchical division in the setting database the unit information of the lowest level in the order information; the matching unit is specifically configured to: determine the target unit that matches the unit information of the lowest level in the order information to be corrected in the reference unit information of the lowest level of the setting database information; determine a target subject name that meets a preset condition among the multiple reference subject names corresponding to the target unit information; obtain a target subject name matching the target search information according to the first reference information corresponding to the target subject name Order reference information.
  • the setting database stores second reference information corresponding to the name of the reference subject; the determining unit is specifically configured to: obtain the to-be-corrected data according to the hierarchical division in the setting database the unit information of the lowest level in the order information; the matching unit is specifically configured to: determine the target unit that matches the unit information of the lowest level in the order information to be corrected in the reference unit information of the lowest level of the setting database information; determine the target subject name that meets the preset condition among the multiple reference subject names corresponding to the target unit information; according to the reference unit information of each level corresponding to the target subject name, and the target subject name Corresponding second reference information, obtain order reference information matching the target search information.
  • the matching unit when the matching unit determines a target subject name that meets a preset condition among the multiple reference subject names corresponding to the target unit information, the matching unit is specifically configured to: match the order to be corrected The subject names corresponding to the information are respectively matched with multiple reference subject names corresponding to the target unit information; the reference subject name with the highest matching score and exceeding the first set threshold is determined as the target subject name.
  • the matching unit is specifically configured to: perform a search on the Internet according to the target search information, and obtain one or more candidate order reference information matching the target search information; The order reference information is matched with the order information to be corrected; and the candidate order reference information with the highest matching score and exceeding the second set threshold is obtained as the order reference information.
  • the device further includes an adding unit for adding the order reference information obtained from the Internet and the subject name corresponding to the order information to be corrected, and adding force P to the device In the information corresponding to the reference unit information of the lowest level in the database.
  • the device further includes an update unit, configured to update the information in the setting database according to the order reference information obtained from the Internet and the subject name corresponding to the order information to be corrected.
  • the information corresponding to the reference unit information of the lowest level is updated.
  • the order information to be corrected includes at least address information, and at least one content element included in the address information includes at least one of the following: administrative area, postal code; Reference unit information at multiple levels includes reference borough information or zip code information.
  • the obtaining unit is specifically configured to: obtain a text recognition result of the object to be processed, where the text recognition result includes multiple text boxes;
  • the first text box of the information, the key information includes at least one content element of the order information to be corrected and at least one of the keywords indicating the order information to be corrected; according to the first text box, to At least a part of the multiple text boxes is combined to obtain a combined text box; and the order information to be corrected is acquired from the combined text box.
  • An embodiment of the present disclosure further provides an electronic device, the device includes a memory and a processor, where the memory is used to store computer instructions that can be executed on the processor, and the processor is used to execute the computer instructions.
  • the method for correcting order information described in any embodiment of the present disclosure is implemented.
  • the order information correction method, device, device and storage medium obtain order information to be corrected according to the text recognition result of the order, and determine target search information from the text recognition result, Acquiring order reference information matching the target search information, and using the order reference information to correct the order information to be corrected to obtain target order information, can quickly obtain accurate target order information from the text recognition result of the order.
  • FIG. 5 provides an electronic device according to at least one embodiment of the present disclosure, the device includes a memory and a processor, where the memory is used to store computer instructions that can be executed on the processor, and the processor is used to execute all The method for correcting order information described in any embodiment of the present disclosure is implemented when the computer instruction is used.
  • At least one embodiment of the present disclosure also provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the method for correcting order information described in any embodiment of the present disclosure is implemented .
  • At least one embodiment of the present disclosure also provides a computer program, comprising computer-readable code, When the computer-readable code is executed in an electronic device, the processor in the electronic device implements the method for correcting order information described in the first aspect when executed.
  • the computer program product of order information provided by the embodiments of the present disclosure includes a computer-readable storage medium storing computer-readable codes, and the instructions included in the computer-readable codes can be used to execute the methods described in the foregoing method embodiments.
  • the correction method of the order information includes a computer-readable storage medium storing computer-readable codes, and the instructions included in the computer-readable codes can be used to execute the methods described in the foregoing method embodiments.
  • the correction method of the order information includes a computer-readable storage medium storing computer-readable codes, and the instructions included in the computer-readable codes can be used to execute the methods described in the foregoing method embodiments.
  • the correction method of the order information includes a computer-readable storage medium storing computer-readable codes, and the instructions included in the computer-readable codes can be used to execute the methods described in the foregoing method embodiments.
  • the correction method of the order information includes a computer-readable storage medium storing computer-readable codes, and the instructions included in the computer-readable codes can be used to execute the methods described in
  • one or more embodiments of this specification may be provided as a method, system or computer program product. Accordingly, one or more embodiments of this specification may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, one or more embodiments of this specification may employ a computer program implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein form of the product.
  • computer-usable storage media including, but not limited to, disk storage, CD-ROM, optical storage, etc.
  • Embodiments of the subject matter and functional operations described in this specification can be implemented in: digital electronic circuits, tangible embodied computer software or firmware, computer hardware including the structures disclosed in this specification and their structural equivalents, or a combination of one or more of them.
  • Embodiments of the subject matter described in this specification may be implemented as one or more computer programs, ie, one or more of computer program instructions encoded on a tangible non-transitory program carrier for execution by or to control the operation of data processing apparatus or multiple modules.
  • the program instructions may be encoded on an artificially generated propagated signal, such as a machine-generated electrical, optical or electromagnetic signal, which is generated to encode and transmit information to a suitable receiver device for use by the data
  • the processing device executes.
  • the computer storage medium may be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or a combination of one or more of them.
  • Computers suitable for the execution of a computer program include, for example, general and/or special purpose microprocessors, or any other type of central processing unit.
  • the central processing unit will receive instructions and data from read only memory and/or random access memory.
  • the basic components of a computer include a central processing unit for implementing or executing instructions and one or more memory devices for storing instructions and data.
  • a computer will also include one or more mass storage devices for storing data, such as magnetic disks, magneto-optical disks, or optical disks, or the computer will be operably interfaced with such mass storage devices to receive data from or to It transmits data, or both.
  • the computer does not have to have such a device.
  • the computer can be embedded in another device, such as a mobile phone, a personal Digital Assistant (PDA) mobile audio or video players, game consoles, Global Positioning System (GPS) receivers, or portable storage devices such as Universal Serial Bus (USB) flash drives, to name a few.
  • PDA personal Digital Assistant
  • GPS Global Positioning System
  • USB Universal Serial Bus
  • Computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media, and memory devices, including, for example, semiconductor memory devices (eg, EPROM, EEPROM, and flash memory devices), magnetic disks (eg, internal hard disk or removable disk), magneto-optical disks, and CD-ROM and DVD-ROM disks.
  • semiconductor memory devices eg, EPROM, EEPROM, and flash memory devices
  • magnetic disks eg, internal hard disk or removable disk
  • magneto-optical disks e.g, CD-ROM and DVD-ROM disks.
  • the processor and memory may be supplemented by or incorporated in special purpose logic circuitry.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Remote Sensing (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • General Health & Medical Sciences (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

L'invention concerne un procédé permettant de corriger des informations de commande. Le procédé consiste : à acquérir, selon un résultat de reconnaissance de texte d'une commande, des informations de commande à corriger ; à déterminer des informations de recherche cibles à partir du résultat de reconnaissance de texte ; à acquérir des informations de référence de commande correspondant aux informations de recherche cibles ; et à corriger les informations de commande à corriger en utilisant les informations de référence de commande de sorte à obtenir des informations de commande cibles.
PCT/IB2021/055848 2020-11-25 2021-06-30 Procédé et appareil permettant de corriger des informations de commande, dispositif, et support de stockage Ceased WO2022112857A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011339777.2 2020-11-25
CN202011339777.2A CN112395874B (zh) 2020-11-25 2020-11-25 订单信息的校正方法、装置、设备及存储介质

Publications (1)

Publication Number Publication Date
WO2022112857A1 true WO2022112857A1 (fr) 2022-06-02

Family

ID=74603919

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/IB2021/055848 Ceased WO2022112857A1 (fr) 2020-11-25 2021-06-30 Procédé et appareil permettant de corriger des informations de commande, dispositif, et support de stockage

Country Status (2)

Country Link
CN (1) CN112395874B (fr)
WO (1) WO2022112857A1 (fr)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114092684B (zh) * 2021-11-17 2024-11-19 中国银联股份有限公司 一种文本校准方法、装置、设备及存储介质
CN114120322B (zh) * 2022-01-26 2022-05-10 深圳爱莫科技有限公司 订单商品数量识别结果校正方法及处理设备

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050137991A1 (en) * 2003-12-18 2005-06-23 Bruce Ben F. Method and system for name and address validation and correction
WO2009005492A1 (fr) * 2007-06-29 2009-01-08 United States Postal Service Systèmes et procédés pour valider une adresse
CN107239453A (zh) * 2016-03-28 2017-10-10 平安科技（深圳）有限公司 信息写入方法和装置
WO2020134991A1 (fr) * 2018-12-29 2020-07-02 益萃网络科技（中国）有限公司 Procédé de saisie automatique pour formulaire papier, appareil et dispositif informatique et support d'informations

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5043735B2 (ja) * 2008-03-28 2012-10-10 インターナショナル・ビジネス・マシーンズ・コーポレーション 情報分類システム、情報処理装置、情報分類方法およびプログラム
CN110442702B (zh) * 2019-08-15 2022-09-02 北京上格云技术有限公司 搜索方法、装置、可读存储介质和电子设备
CN110674396B (zh) * 2019-08-28 2021-04-27 北京三快在线科技有限公司 文本信息处理方法、装置、电子设备及可读存储介质

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050137991A1 (en) * 2003-12-18 2005-06-23 Bruce Ben F. Method and system for name and address validation and correction
WO2009005492A1 (fr) * 2007-06-29 2009-01-08 United States Postal Service Systèmes et procédés pour valider une adresse
CN107239453A (zh) * 2016-03-28 2017-10-10 平安科技（深圳）有限公司 信息写入方法和装置
WO2020134991A1 (fr) * 2018-12-29 2020-07-02 益萃网络科技（中国）有限公司 Procédé de saisie automatique pour formulaire papier, appareil et dispositif informatique et support d'informations

Also Published As

Publication number Publication date
CN112395874A (zh) 2021-02-23
CN112395874B (zh) 2025-04-22

Similar Documents

Publication Publication Date Title
AU2020279921B2 (en) Representative document hierarchy generation
US20120278339A1 (en) Query parsing for map search
CN111652176B (zh) 信息提取方法、装置、设备及存储介质
JP7149721B2 (ja) 情報処理装置、文字認識エンジン最適化方法及びプログラム
CN110674396B (zh) 文本信息处理方法、装置、电子设备及可读存储介质
CN109344387B (zh) 形近字字典的生成方法、装置和形近字纠错方法、装置
CN103544186A (zh) 挖掘图片中的主题关键词的方法和设备
US20250363302A1 (en) Mapping entities in unstructured text documents via entity correction and entity resolution
WO2022112857A1 (fr) Procédé et appareil permettant de corriger des informations de commande, dispositif, et support de stockage
CN113704427A (zh) 一种文本出处的确定方法、装置、设备及存储介质
JP2019169025A (ja) 情報処理装置、文字認識エンジン選択方法及びプログラム
CN115470307A (zh) 一种地址匹配方法及装置
US8996501B2 (en) Optimally ranked nearest neighbor fuzzy full text search
US10331717B2 (en) Method and apparatus for determining similar document set to target document from a plurality of documents
CN114201480A (zh) 一种基于nlp技术的多源poi融合方法、装置及可读存储介质
CN111460325A (zh) Poi搜索方法、装置与设备
JP6146209B2 (ja) 情報処理装置、文字認識方法、及びプログラム
CN113626536B (zh) 一种基于深度学习的新闻地理编码方法
JP2005107931A (ja) 画像検索装置
US12405132B2 (en) Apparatus and method for matching POI entities
CN112579713A (zh) 地址识别方法、装置、计算设备及计算机存储介质
HK40039018A (en) Correction method, device, equipment and storage medium of order information
WO2022164387A1 (fr) Procédé et système de déduplication de bases de données de points d'intérêt
KR101743289B1 (ko) 자동 주소 번역 시스템 및 방법
JP2655087B2 (ja) 文字認識後処理方式

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21897263

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21897263

Country of ref document: EP

Kind code of ref document: A1