WO2023035623A1 - Answer corpus generation method based on artificial intelligence, and related device - Google Patents

Abstract

The present application relates to the field of artificial intelligence. Disclosed are an answer corpus generation method based on artificial intelligence, and a related device. The method comprises: acquiring an inquiry corpus and an answer corpus to be pushed, and performing word segmentation processing on the basis of a preset linear-chain conditional random field, so as to correspondingly obtain segmented inquiry words and segmented answer words (101); performing professional word semantic matching on the segmented inquiry words and the segmented answer words, so as to correspondingly obtain professional segmented inquiry words and professional segmented answer words (102); performing cross question-answer matching on the professional segmented inquiry words and the professional segmented answer words, and combining the professional segmented inquiry words and the professional segmented answer words according to a cross question-answer matching result, so as to obtain a diagnosis statement (103); and matching, by means of using a preset prior medical knowledge base, therapy product information which corresponds to the diagnosis statement, combining the therapy product information and said answer corpus, so as to obtain a new answer corpus to be pushed, and pushing said new answer corpus (104). By means of the present application, therapy product recommendation during an online inquiry process is realized, thereby improving the degree of intelligence of an online inquiry.

Description

Response corpus generation method and related equipment based on artificial intelligence

This application claims the priority of the Chinese patent application with the application number 202111055021.X and the title of the invention "artificial intelligence-based response corpus generation method and related equipment" submitted to the China Patent Office on September 09, 2021, the entire content of which is passed References are incorporated in the application.

technical field

The present application relates to the field of artificial intelligence, in particular to an artificial intelligence-based answer corpus generation method and related equipment.

Background technique

With the development of computer technology, the existing product recommendation services have evolved from the most original advertising promotion, bidding ranking, etc. to the current advertising recommendation algorithm. At this stage, many video websites are trying to combine product recommendations with video elements. A variety of Internet shopkeeper AI recommendation algorithms, etc. They are all evolving in two technical directions, that is, technically accurate recommendation and formal non-violent display. In order to obtain higher recommendation conversion rate and better user service experience.

The inventor realizes that the existing medical platform has a product recommendation function in the consultation process, generally relying on the doctor to make a verbal recommendation, and when the patient confirms that there is a purchase intention, the product or service is pushed. Although this promotion method relying on manpower has a better conversion effect, it cannot be promoted on a large scale. In the era of vigorous development of big data, accurate intelligent recommendations based on big data analysis have not been fully applied.

Contents of the invention

The main purpose of this application is to solve the technical problem of low accuracy in the existing intelligent product recommendation based on the consultation link.

The first aspect of the present application provides a method for generating response corpus based on artificial intelligence, including: obtaining the query corpus and the response corpus corresponding to the query corpus to be pushed, and based on the preset linear chain conditional random field, respectively The query corpus and the response corpus to be pushed are subjected to word segmentation processing, correspondingly obtaining a plurality of question segmentation words and a plurality of response segmentation words; respectively performing professional word semantic matching on the query diagnosis segmentation words and the response segmentation words, correspondingly obtaining the Inquiry professional participle corresponding to the question participle and the response professional participle corresponding to the response participle; Carry out cross question-and-answer matching to each described interrogation professional participle and each described response professional participle in turn, and according to the result of cross question-answer matching, to The professional participle of the inquiry and the professional participle of the response are combined to obtain a diagnostic statement; using the preset prior medical knowledge base, matching the treatment product information corresponding to the diagnosis statement, and comparing the treatment product information and the response corpus to be pushed Combine them to get a new response corpus to be pushed and push it.

The second aspect of the present application provides an artificial intelligence-based response corpus generation device, including: a word segmentation module, used to obtain the query corpus and the response corpus corresponding to the query corpus to be pushed, and based on the preset linear chain condition. At the airport, word segmentation processing is performed on the inquiry corpus and the response corpus to be pushed, correspondingly obtaining a plurality of question segmentation words and a plurality of response word segmentations; a semantic matching module is used to separately analyze the question segmentation words and the response Segmentation carries out semantic matching of professional words, and correspondingly obtains the corresponding professional word segmentation of the inquiry and the corresponding professional participle of the response participle; Responding to professional word segmentation for cross-question matching, and according to the result of cross-question-answer matching, combining the professional word segmentation of the inquiry and the professional word segmentation of the response to obtain a diagnostic statement; the combination module is used to adopt a preset prior medical knowledge base, Match the treatment product information corresponding to the diagnostic sentence, combine the treatment product information and the response corpus to be pushed, obtain a new response corpus to be pushed, and push it.

The third aspect of the present application provides an artificial intelligence-based answer corpus generation device, including: a memory and at least one processor, instructions are stored in the memory; the at least one processor invokes the instructions in the memory , so that the artificial intelligence-based answer corpus generation device executes the artificial intelligence-based answer corpus generation method as follows: obtain the question corpus and the answer corpus corresponding to the question corpus, and based on the preset linear chain conditional random field, respectively performing word segmentation processing on the inquiry corpus and the response corpus to be pushed, and correspondingly obtaining multiple question segmentation words and multiple response word segmentations; Semantic matching, correspondingly obtain the professional participles corresponding to the question participle and the corresponding participles of the response participles; carry out cross question-and-answer matching to each of the participles of the professional participle of the inquiry and each participle of the professional participle of the response in turn, and according to As a result of cross-question-answer matching, combine the professional participle of the inquiry and the professional participle of the response to obtain a diagnostic statement; use a preset prior medical knowledge base to match the treatment product information corresponding to the diagnosis statement, and compare the treatment product information Combine with the response corpus to be pushed to obtain a new response corpus to be pushed and push it.

The fourth aspect of the present application provides a computer-readable storage medium, and instructions are stored in the computer-readable storage medium, and when it is run on a computer, it causes the computer to perform the following artificial intelligence-based response corpus generation Method: Obtain the question corpus and the response corpus to be pushed corresponding to the question corpus, and perform word segmentation processing on the question corpus and the response corpus to be pushed respectively based on the preset linear chain conditional random field, correspondingly get more An inquiry participle and a plurality of response participles; the professional word semantic matching is carried out to the question participle and the response participle respectively, and correspondingly obtain the question participle corresponding to the question participle and the response specialty corresponding to the response participle Segmentation: Carry out cross question-and-answer matching for each of the professional word segmentations of the inquiry and each of the professional word segmentations of the responses in turn, and according to the results of the cross question-answer matching, combine the professional word segmentation of the inquiry and the professional word segmentation of the response to obtain a diagnostic sentence ; Using the preset prior medical knowledge base, matching the treatment product information corresponding to the diagnosis sentence, combining the treatment product information and the response corpus to be pushed, obtaining a new response corpus to be pushed and pushing it.

In the technical solution provided by this application, by obtaining the inquiry corpus input by the patient and the doctor's reply to the patient's response corpus to be pushed, the inquiry corpus and the corpus to be pushed are converted into professional inquiry corpus and professional word semantic matching Response corpus, and match the professional word segmentation of the inquiry and response to judge the patient's condition and the treatment plan recommended by the doctor, get the diagnosis sentence, and match the treatment product information according to the diagnosis sentence, and together with the response corpus to be pushed sent to patients. Realize the smooth product recommendation function in the consultation process. The recommended products and services are accurate recommendations for this chat scene, providing users with a quick ordering function for specific products, and realizing accurate recommendations of property rights during the consultation process.

Description of drawings

Fig. 1 is the schematic diagram of the first embodiment of the response corpus generation method based on artificial intelligence of the present application;

Fig. 2 is the second embodiment schematic diagram of the application's artificial intelligence-based response corpus generation method;

Fig. 3 is the schematic diagram of the third embodiment of the answer corpus generation method based on artificial intelligence of the present application;

Fig. 4 is a schematic diagram of an embodiment of the artificial intelligence-based response corpus generating device of the present application;

FIG. 5 is a schematic diagram of another embodiment of the artificial intelligence-based response corpus generation device of the present application;

FIG. 6 is a schematic diagram of an embodiment of an artificial intelligence-based response corpus generation device in the present application.

Detailed ways

The embodiment of the present application provides an artificial intelligence-based answer corpus generation method and related equipment, which obtains the question corpus and the answer corpus to be pushed, and performs word segmentation processing based on the preset linear chain condition random field, and obtains the question word segmentation and answer correspondingly Word segmentation; carry out semantic matching of professional words for question and answer word segmentation, correspondingly obtain professional word segmentation for inquiry and professional response; carry out cross-question and answer matching for professional word segmentation for inquiry and professional word for response, and according to the results of cross-question-answer matching, Combine the word segmentation for diagnosis and the professional word for response to obtain the diagnosis statement; use the preset prior medical knowledge base to match the treatment product information corresponding to the diagnosis statement, and combine the treatment product information and the response corpus to be pushed to obtain a new response to be pushed Corpus and push. This application realizes the recommendation of treatment products in the process of online consultation, and improves the intelligence of online consultation.

The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above drawings are used to distinguish similar objects, and not necessarily Used to describe a specific sequence or sequence. It is to be understood that the terms so used are interchangeable under appropriate circumstances such that the embodiments described herein can be practiced in sequences other than those illustrated or described herein. Furthermore, the term "comprising" or "having" and any variations thereof, are intended to cover a non-exclusive inclusion, for example, a process, method, system, product or device comprising a sequence of steps or elements is not necessarily limited to those explicitly listed instead, may include other steps or elements not explicitly listed or inherent to the process, method, product or apparatus.

For ease of understanding, the following describes the specific process of the embodiment of the present application. Please refer to FIG. 1. The first embodiment of the artificial intelligence-based answer corpus generation method in the embodiment of the present application includes:

101. Obtain the question corpus and the response corpus to be pushed corresponding to the question corpus, and based on the preset linear chain conditional random field, perform word segmentation processing on the question corpus and the response corpus to be pushed respectively, correspondingly obtain multiple question segmentation words and multiple answer participle;

It can be understood that the execution subject of the present application may be an artificial intelligence-based response corpus generation device, or a terminal or a server, which is not specifically limited here. The embodiment of the present application is described by taking the server as an execution subject as an example.

In this embodiment, the inquiry corpus is input and sent by the patient through the inquiry chat interface, and the response corpus to be pushed is obtained by the docking doctor receiving the patient's inquiry corpus, inputting and sending it through the chat interface, for example, the patient sends "arm Skin allergies, what should I do with the doctor?”, and the doctor replies to the query corpus, “Have you ever taken anti-allergic medicine at home?” The response corpus to be pushed, among which, the corpus to be pushed is sent to the background, and If it is not directly forwarded to the patient, it needs to be implanted with product recommendation information through this application method before sending it.

In addition, the acquisition of product recommendation information is based on the semantic matching of the inquiry corpus and the response corpus to be pushed, so it is necessary to perform semantic recognition on the query corpus and the response corpus to be pushed, and match the corresponding product recommendation information based on the semantic recognition results . Here, by combining the Hidden Markov Model, the pre-set linear chain conditional random field can be used to perform word segmentation processing on the question corpus and the response corpus to be pushed, so as to obtain corresponding multiple question and answer word segmentation. The relationship of "generation-discriminant pair" performs word segmentation processing on the question segmentation and response segmentation.

We generally assume that the question corpus and the response corpus to be pushed have the same structure of multiple response participle X and known part of speech Y, namely: X=(x1, x2,...,xn), Y=(y1, y2,... ..., yn); each Y is connected to X respectively, and each Y is sequentially connected by y1-yn, and the formed structure constitutes a linear chain conditional random field. And learn through the linear random field until the optimal weight is reached, and the question and answer word segmentation can be obtained.

102. Carry out semantic matching of professional words on the question and answer participle respectively, and correspondingly obtain the professional question participle corresponding to the question participle and the answer professional participle corresponding to the answer participle;

In this embodiment, the part of speech is firstly distinguished between the question participle and the response participle, and the parts of speech are marked for the question participle and the response participle, so as to find out the noun question participle and response participle. And the answer participle is directly excluded. It can be directly connected to a custom or existing rule dictionary for exclusion, and the operator can complete the rules, or it can be connected to the AI system for exclusion.

Further, using reserved question and answer word segmentation for semantic matching of professional words can be connected to the existing rule-based synonym thesaurus, and the operation personnel can complete the rules, and can also be connected to the AI synonym function to achieve AI-based System conversion. The ultimate goal is to convert colloquial or non-unified question and answer participle into unified question and answer participle.

103. Carry out cross question-and-answer matching for each professional participle of the inquiry and each participle of the professional response in turn, and according to the result of the cross-question-answer matching, combine the participles of the professional participle of the inquiry and the participle of the professional response to obtain a diagnostic sentence;

In this embodiment, the doctor responds to the question corpus input by the patient. After word segmentation processing and semantic recognition, the doctor obtains the professional question and response word segmentation, which are the key question words and key words in the question corpus. Response words, which may contain multiple symptoms or questions during the consultation process, and the doctor will also respond to different symptoms or questions accordingly. Therefore, it is necessary to perform cross-question and answer matching on the professional participle of the consultation and the professional participle of the response. .

Specifically, it is possible to perform a distributed representation of the word segmentation for each medical inquiry specialty first, and obtain the query word vector sequence and the preset reference word vector sequence bound to the answer professional word segmentation; and then use the deep learning model to Identify the semantic similarity between the query word vector sequence and the reference word vector sequence; then calculate the text similarity between each query professional word and each response professional word segmentation through a preset formula; then use the semantic similarity and text similarity The comprehensive similarity between the two is determined; according to the comprehensive similarity, the corresponding professional word segmentation matching the text professional word segmentation is determined and combined to obtain the diagnostic sentence.

104. Use the preset prior medical knowledge base, match the treatment product information corresponding to the diagnosis sentence, combine the treatment product information and the response corpus to be pushed, obtain and push the new response corpus to be pushed.

In this embodiment, the mapping relationship table between the regular expression of the diagnosis statement and the diagnosis result is configured in the prior medical knowledge base, the corresponding regular expression can be found through the diagnosis statement, and then the mapping relationship can be traversed through the regular expression table to determine the diagnostic result mapped to the diagnostic statement, and then a mapping relationship table between the diagnostic result and the treatment product identification information is also configured in the prior medical knowledge base. By traversing the mapping relationship table through the diagnostic result, you can find To specific treatment product identification information, including recommended links and summary information of quality products, where the summary information may include brand name, instructions for use, sales quantity and favorable rating, etc. Specifically as follows:

(1) Using diagnostic sentences, performing hierarchical traversal in the preset prior medical knowledge base, and according to the results of hierarchical traversal, determine the diagnostic results corresponding to the diagnostic sentences;

(2) Select the therapeutic product identification information matching the diagnosis result from the prior knowledge base, and obtain the therapeutic product information mapped to the therapeutic product identification information, where the therapeutic product information includes the recommended link and summary information of the therapeutic product.

In the embodiment of the present application, by obtaining the inquiry corpus input by the patient and the doctor's reply to the patient's response corpus to be pushed, the inquiry corpus and the corpus to be pushed are converted into professional inquiry corpus and response corpus through semantic matching of professional words , and match the professional word segmentation of the inquiry and response to judge the patient's condition and the treatment plan recommended by the doctor, obtain the diagnosis sentence, and match the treatment product information according to the diagnosis sentence, and push it together with the response corpus to be pushed to patient. Realize the smooth product recommendation function in the consultation process. The recommended products and services are accurate recommendations for this chat scene, providing users with a quick ordering function for specific products, and realizing accurate recommendations of property rights during the consultation process.

Referring to Fig. 2, the second embodiment of the artificial intelligence-based answer corpus generation method in the embodiment of the present application includes:

201. Acquire the question-and-answer corpus and the response corpus to be pushed corresponding to the question-and-answer corpus, and extract the character feature vector and the corresponding pinyin feature vector of the question-and-answer corpus, wherein the question-and-answer corpus includes the question-answer corpus and the response corpus to be pushed;

202. Splicing character feature vectors and corresponding pinyin feature vectors to obtain context information vectors, and performing semantic analysis on context information vectors to obtain semantic features;

203. Use the preset linear chain conditional random field to mark the semantic features, obtain the word segmentation tag sequence, and decode the word segment tag sequence to obtain multiple question and answer word segmentation, wherein the question and answer word segmentation includes question and answer word segmentation;

In this embodiment, each character in the question corpus is divided and sequentially encoded to obtain a character table. Through the character table, the character feature vector of each character is trained using a neural network such as Word2vec, wherein the character feature The vector contains the context information of the question corpus. Each character feature vector represents a character. The dimension of each character feature vector can be adjusted according to the size of the corpus. Generally, the optional dimensions are 50, 100, 200, etc.

Convert each text in the question-and-answer corpus into a letter to construct the alphabet corresponding to the question-and-answer corpus, and construct each letter in the alphabet to be randomly initialized into a vector form to obtain the pinyin vector matrix corresponding to each text. Among them, cL is The vector corresponding to the l-th letter in the pinyin corresponding to the text; L is the maximum length of the pinyin, which is a fixed value by default. Generally speaking, the maximum length of the pinyin corresponding to the text is 6, so L can be set to 6; and if the length L' of the pinyin corresponding to the text is less than L, the elements in the row L'+1~L' in the corresponding pinyin vector matrix will be Set to zero; for example, if the length of the pinyin "shi" corresponding to "市" is 3, all rows 4-6 in the corresponding pinyin vector matrix are set to zero. Each pinyin feature vector matrix is sequentially encoded by a convolutional neural network CNN to obtain a pinyin feature vector of a fixed size.

In this embodiment, the character feature vector and the pinyin feature vector are assembled according to the order of each character in the question corpus for one-to-one correspondence, and the context information vector can be obtained, and then the context information vector is input into the bidirectional LSTM neural network for semantic analysis , where the bidirectional LSTM neural network includes a forward LSTM neural network and a backward LSTM neural network, combined with forgetting and saving mechanisms for backpropagation to learn the semantic features of the context information vector.

Finally, the semantic feature is marked by the linear chain random field CRF, and the word segmentation tag sequence is obtained, where the semantic feature Z={z1, z2,...,zN}, and the tag information Y={y1, y2,...,yN} , then under the condition that the given semantic feature Z takes the value z, the conditional probability of taking the value y on the label sequence Y is p(y|z), and the specific formula is as follows:

Wherein, n=1,2,...,N, tk() and sl() are feature functions, and λk and μl are weights corresponding to tk() and sl() respectively. p(y|z) is the probability of labeling y when the value is z, and S(z) is a normalization factor, in order to normalize the output to a value from 0 to 1. After the sequence of word segmentation is selected and obtained through the above formula, it is decrypted to obtain a plurality of corresponding question and answer word segments.

204. Construct the first phonetic-graphic code of the question-and-answer word segmentation in the preset common word dictionary, and construct the second phonetic-graphic code of each professional word in the preset professional word dictionary, and calculate the first phonetic-graphic code and the second phonetic-graphic code The edit distance between;

205. Combining the question-and-answer participle corresponding to the first phonetic-phonetic code whose editing distance is less than the preset editing distance threshold, to obtain the question-and-answer word group, and selecting the professional word corresponding to the second phonetic-phonetic code whose editing distance is smaller than the editing distance threshold;

206. Using the selected professional words in turn to replace the corresponding question and answer participle in the question and answer participle to obtain multiple professional phrases corresponding to the question and answer participle;

207. Perform semantic analysis on the question and answer phrases to obtain the first semantic analysis result, and perform semantic analysis on each professional phrase to obtain multiple second semantic analysis results;

208. Comparing the first semantic analysis results with the second semantic analysis results respectively, and selecting synonyms of each question and answer participle in the question and answer participle group from multiple professional phrases according to the comparison results;

209. Use the selected synonym as the question-and-answer professional participle corresponding to the question-and-answer participle, wherein the question-and-answer professional participle includes the inquiry-specialized participle and the response-professional participle;

In this embodiment, each character in the question-and-answer word segmentation and each professional word in the professional word dictionary has its special phonetic and font combination. Wherein, the initial consonant, the final vowel, the complement code of the final vowel and the tone of each character are digitally coded to obtain the four-digit code of its pronunciation; the Chinese character structure, five four-corner codes, and the number of strokes of each character are coded to obtain the shape of the character 7-digit code; the combination of the two can form the unique 11-digit phonetic code of each character, including the first phonetic code and the second phonetic code.

Concrete, if represent the coding fields corresponding to 26 initial consonants of order in the table of initials with A1 to A26; Represent the code field corresponding to 39 simple or compound vowels of the order in the table of finals and syllables with B1 to B39; The 39 finals correspond to the coding fields corresponding to the finals complement code; D1 to D4 represent the coding fields corresponding to the first to four tones; then the phonetic code coding information of the word "花" is A11B13C13D1. If the left and right structure, up and down structure, left middle and right structure, upper middle and lower structure, half surround structure, full surround structure, mosaic structure corresponding coding fields of representing commonly used characters with E1 to E7; With F0 to F9, G0 to G9, H0 to H9, J0 to J9, K0 to K9 represent respectively the upper left corner, the upper right corner, the lower left corner, the lower right corner, and the coding field corresponding to the ten classes of strokes corresponding to the attached number; with Li (i is the number of strokes and i is Positive integer) represents the coding field corresponding to the number of strokes; then the font coding information of the word "花" is E2F4G4H2J1K4L7, so the coding information of the commonly used characters of the word "flower" is A11B13C13D1E2F4G4H2J1K4L7.

In this embodiment, the phonetic-graph code includes eleven types of coding fields. If the same type of coding field is different between the first phonetic-graph code and the second phonetic-graph code, the edit distance is increased by 1; otherwise, the original value remains. If the coding fields of all types are consistent between the two, it means that the two commonly used words have the highest similarity, and the edit distance between the two is 0. If the coding fields of all types between the two are inconsistent, it means that the two commonly used words are similar If the degree is the lowest, the edit distance between the two is 11, so the edit distance between the pre-replaced word and the commonly used word is between 0-11.

In this embodiment, since the edit distance is the quantitative value of the similarity between each word in the question and answer segment and each word in the professional dictionary, and the smaller the edit distance, the higher the similarity, so the user can set the preset edit distance Threshold, used to filter professional words for cross combination.

When cross-combining professional words (a1, a2, b1, b2, c1, d1, d2, e1, e2, e3), refer to the question and answer word group (A, B, C, D, E) to select the first group Professional phrase (a1, b1, c1, d1, e1), fixed professional word "B, C, D, E", change "A" to "a1" and "a2" in turn, can get professional phrase (a1, B , C, D, E) and (a2, B, C, D, E), and so on, until all replacement combinations are completed and stop.

In this embodiment, a conventional semantic recognition model is used to perform semantic analysis on the question-and-answer sub-phrases and professional phrases, and obtain the first semantic analysis result and the second semantic analysis result respectively. If there is a small semantic deviation between the two after the comparison, it is determined that the professional word changed in the corresponding professional phrase is a synonym for the corresponding question and answer participle, and it is used as the question and answer professional participle corresponding to the question and answer participle.

210. Carry out cross question-and-answer matching for each professional participle of inquiry and each participle of professional response in turn, and according to the result of cross-question-answer matching, combine the participle of professional questioning and participle of professional response to obtain a diagnostic sentence;

211. Using the preset prior medical knowledge base, matching the treatment product information corresponding to the diagnosis statement, combining the treatment product information and the response corpus to be pushed, obtaining and pushing the new response corpus to be pushed.

In the embodiment of the present application, the phonetic-phonetic codes of question and answer word segmentation and professional word segmentation are constructed through the preset common word dictionary and professional word dictionary, and the number of each question and answer word segmentation is determined through the matching of phonetic-phonetic codes. Synonyms, and replace them to obtain the corresponding professional participle for consultation and professional participle for answering, which will be more accurate in subsequent product matching.

Referring to Fig. 3, the third embodiment of the artificial intelligence-based answer corpus generation method in the embodiment of the present application includes:

301. Obtain the medical questioning corpus and the response corpus to be pushed corresponding to the medical questioning corpus, and use a preset fully homomorphic encryption algorithm to convert the medical questioning corpus and the response corpus to be pushed into corresponding plaintext;

302. Perform encryption operation on the base plaintext to obtain the encrypted corpus, and calculate the original ciphertext code, inverse ciphertext code, and ciphertext complement of the encrypted corpus according to the preset modulus;

303. Using the original code of the ciphertext, the reverse code of the ciphertext, and the complement of the ciphertext, perform a modular operation on the encrypted corpus to obtain the modular encrypted corpus, wherein the modular encrypted corpus includes the first modular encrypted corpus corresponding to the medical inquiry corpus and the response to be pushed The second encrypted corpus corresponding to the corpus;

304. Use the first encrypted corpus as a new question corpus and the second encrypted corpus as a new response corpus to be pushed, and perform word segmentation on the question corpus and the response corpus to be pushed based on the preset linear chain conditional random field Processing, and correspondingly obtain multiple question and response participle words;

In this embodiment, the type (Type) of the plaintext m of the inquiry corpus and the response corpus to be pushed is T, and the set of T is {integer, real number, character, date, Boolean, etc.}, and the known plaintext ms, where s Indicates the data system (System), that is, binary, decimal, hexadecimal, 521, etc., denoted as (T, ms); for example: s=2 means binary, binary is usually represented by B, and plain text m is represented as Binary bit mB is written as (T, mB); s=16 represents hexadecimal, and hexadecimal is usually represented by H, and plaintext m is expressed as hexadecimal bit mH, and is written as (T, mH); s =512 means 512 base, and plaintext m is expressed as 512 base m512, which is recorded as (T, m512) and so on. For example, a medical inquiry corpus and a response corpus to be pushed are respectively m1=5 and m2=3, which need to be encrypted with binary digits respectively, and the obtained binary digit plaintexts are 101 and 011 respectively.

In this embodiment, when performing encryption operations on the base plaintext, the following encryption operation formula can be used: c=(m+s*r+p*r) mod x0, wherein, c represents the ciphertext, and m represents the consultation The base digits in the corpus and the plaintext of the response corpus to be pushed, s represents the base used in encryption, r represents a random number, p is an encryption key, x0 is an intermediate variable, which is equal to the encryption key p and another encryption key The product of the key q.

For example, suppose the working key p=111, q=11, x0=p*q=1221, let r=1, then for the numbers 0 and 1 in the binary bit plaintext 101 of plaintext 5 and the binary bit plaintext 011 of plaintext 3 In other words, after calculation using the above formula in this step, it can be obtained: the ciphertext encrypted by number 0 is equal to 113; the ciphertext encrypted by number 1 is equal to 114.

In addition, through the calculation method of original code, inverse code and complement code, the corresponding original ciphertext code, inverse ciphertext code and complement code of ciphertext can be calculated through the encrypted corpus, and the original ciphertext code, When the ciphertext inverse code and the ciphertext complement code are used for encryption calculation, for the addition operation of the encrypted corpus, the ciphertext combination in it is directly summed in place without using the original ciphertext code, ciphertext inverse code and ciphertext Text complement.

Among them, when performing subtraction on the encrypted corpus, first obtain the inverse code of the encrypted corpus of the subtrahend, then obtain the corresponding complement according to the inverse code, and finally compare the complement with the original code of the encrypted corpus of the minuend bit sum operation.

Among them, when performing multiplication for the encrypted corpus, first create an n*(2n-1) matrix according to the number n of text elements in the encrypted corpus, and then sum each column of the constructed matrix to obtain a The new row vector, take this row vector as the multiplication result of the encrypted corpus.

Among them, when performing a division operation on the encrypted corpus, create an empty storage format for the division result. The total length of the storage format is 32 bits, 64 bits or 80 bits, and includes sign bits, integer bits and decimal places, and according to This storage format expands the binary bit plaintext; performs encryption operations on the expanded binary bit plaintext, and combines the results of the encrypted operations to obtain the corresponding ciphertext as the dividend and divisor respectively; set the initial value of the decimal counter count equal to the storage Format length-L, where L is the length of integer bits in the storage format; judge whether the ciphertext of the dividend is greater than the ciphertext of the divisor, if greater, then add the ciphertext of the dividend to the complement of the encrypted corpus, and obtain the remainder as The new dividend, and add the ciphertext of 1 in the integer position, that is, the ciphertext quotient is obtained; otherwise, judge whether the ciphertext of the remainder is all zero or the decimal place counter count is greater than the total length of the storage format, if yes, then Store directly according to the previously preset storage format, otherwise add 0 ciphertext to the far right of the remainder ciphertext to obtain a new remainder ciphertext, and determine whether the new remainder ciphertext in the step is greater than the ciphertext of the divisor, if it is greater than then Add the new remainder ciphertext and the ciphertext complement of the divisor to obtain the new remainder ciphertext again, and set the value of the countth decimal place to the ciphertext value corresponding to 1; set the countth decimal place Set the value to the ciphertext value corresponding to 0, add 1 to the decimal counter count, obtain the integer part and decimal part of the quotient according to the obtained ciphertext value, and store it in the previously preset storage format.

305. Construct the first phonetic-graphic code of the question-and-answer word segmentation in the preset common word dictionary, and construct the second phonetic-graphic code of each professional word in the preset professional word dictionary, and calculate the first phonetic-graphic code and the second phonetic-graphic code The edit distance between;

306. Combining the question-and-answer participle corresponding to the first phonetic-phonetic code whose edit distance is less than the preset edit-distance threshold, to obtain the question-and-answer word group, and selecting the professional word corresponding to the second phonetic-phonetic code whose edit distance is less than the edit-distance threshold;

307. Using the selected professional words in turn to replace the corresponding question and answer participle in the question and answer participle to obtain multiple professional phrases corresponding to the question and answer participle;

308. Perform semantic analysis on the question and answer phrases to obtain a first semantic analysis result, and perform semantic analysis on each professional phrase to obtain multiple second semantic analysis results;

309. Comparing the first semantic analysis result and each second semantic analysis result respectively;

310. According to the comparison results, respectively calculate the degree of difference between the first semantic analysis result and each second semantic analysis result, and determine the similarity between the question and answer phrase group and each professional phrase group according to the degree of difference;

311. Classify the professional phrases corresponding to the professional words in the question-and-answer word segmentation group for each question-and-answer word segmentation, and obtain multiple professional phrases of the question-and-answer word segmentation categories;

312. Select the professional phrase with the highest similarity from the professional phrases of each question and answer word segmentation category, and use the professional words corresponding to the question and answer word segmentation category in the selected professional phrases as synonyms for the question and answer word segmentation;

In the present embodiment, when carrying out cross-combination to professional phrase (a1, a2, b1, b2, c1, d1, d2, e1, e2, e3), refer to question and answer word segmentation (A, B, C, D, E) , change "a1" and "a2" for the professional words corresponding to "A", and divide the corresponding professional phrases (a1, B, C, D, E) and (a2, B, C, D, E) into one category , and so on, get the professional phrases of multiple question and answer word segmentation categories, and the professional phrases with the greatest similarity with the question and answer word segmentation, and the changed professional words are also the most similar to the question and answer word segmentation in the question and answer word segmentation, then it can be used as the question and answer word segmentation synonyms of .

313. Use the selected synonym as the question-and-answer professional participle corresponding to the question-and-answer participle, wherein the question-and-answer professional participle includes the inquiry-specialized participle and the response-professional participle;

314. Carry out cross question-and-answer matching on each question segment and each response segment in turn, and according to the cross-question matching result, combine the query segment and the response segment to obtain a diagnostic statement;

315. Use the preset prior medical knowledge base to match the treatment product information corresponding to the diagnosis statement, combine the treatment product information and the response corpus to be pushed, and obtain and push the new response corpus to be pushed.

In the embodiment of this application, by further encrypting the inquiry corpus and the response corpus to be pushed, and by calculating the ciphertext, product recommendation and other data processing processes, the patient's personal privacy information can be guaranteed, and the patient's consultation can be improved. sense of experience.

The above describes the artificial intelligence-based response corpus generation method in the embodiment of the present application, and the following describes the artificial intelligence-based response corpus generation device in the embodiment of the present application. Please refer to FIG. 4, the artificial intelligence-based response in the embodiment of the present application One embodiment of the corpus generation device includes:

The word segmentation module 401 is used to obtain the query corpus and the response corpus to be pushed corresponding to the query corpus, and perform word segmentation processing on the query corpus and the response corpus to be pushed based on the preset linear chain conditional random field , correspondingly obtain multiple question and response participle words;

The semantic matching module 402 is used to carry out semantic matching of professional words to the question participle and the response participle respectively, and correspondingly obtain the professional question participle corresponding to the question participle and the response professional participle corresponding to the response participle;

The question-and-answer matching module 403 is used to sequentially perform cross question-and-answer matching on each of the professional word segmentations of the inquiry and each of the professional word segmentations of the responses, and perform cross-question matching on the professional word segmentation of the inquiry and the professional word segmentation of the responses according to the results of the cross-question matching. Combination to get the diagnostic statement;

The combination module 404 is used to match the treatment product information corresponding to the diagnostic sentence using the preset prior medical knowledge base, combine the treatment product information and the response corpus to be pushed, obtain and push new response corpus to be pushed.

In the embodiment of the present application, by obtaining the inquiry corpus input by the patient and the doctor's reply to the patient's response corpus to be pushed, the inquiry corpus and the corpus to be pushed are converted into professional inquiry corpus and response corpus through semantic matching of professional words , and match the professional word segmentation of the inquiry and response to judge the patient's condition and the treatment plan recommended by the doctor, obtain the diagnosis sentence, and match the treatment product information according to the diagnosis sentence, and push it together with the response corpus to be pushed to patient. Realize the smooth product recommendation function in the consultation process. The recommended products and services are accurate recommendations for this chat scene, providing users with a quick ordering function for specific products, and realizing accurate recommendations of property rights during the consultation process.

Please refer to Fig. 5, another embodiment of the artificial intelligence-based answer corpus generation device in the embodiment of the present application includes:

The word segmentation module 401 is used to obtain the query corpus and the response corpus to be pushed corresponding to the query corpus, and perform word segmentation processing on the query corpus and the response corpus to be pushed based on the preset linear chain conditional random field , correspondingly obtain multiple question and response participle words;

The semantic matching module 402 is used to carry out semantic matching of professional words to the question participle and the response participle respectively, and correspondingly obtain the professional question participle corresponding to the question participle and the response professional participle corresponding to the response participle;

The question-and-answer matching module 403 is used to sequentially perform cross question-and-answer matching on each of the professional word segmentations of the inquiry and each of the professional word segmentations of the responses, and perform cross-question matching on the professional word segmentation of the inquiry and the professional word segmentation of the responses according to the results of the cross-question matching. Combination to get the diagnostic statement;

The combination module 404 is used to match the treatment product information corresponding to the diagnostic sentence using the preset prior medical knowledge base, combine the treatment product information and the response corpus to be pushed, obtain and push new response corpus to be pushed.

Specifically, the word segmentation module 401 includes:

An extraction unit 4011, configured to extract character feature vectors and corresponding pinyin feature vectors of the question-and-answer corpus, wherein the question-and-answer corpus includes question-and-answer corpus and response corpus to be pushed;

A splicing unit 4012, configured to splice the character feature vectors and corresponding pinyin feature vectors to obtain context information vectors, and perform semantic analysis on the context information vectors to obtain semantic features;

The decoding unit 4013 is configured to use a preset linear chain conditional random field to mark the semantic features to obtain a word segmentation tag sequence, and decode the word segment tag sequence to obtain a plurality of question and answer word segmentation, wherein the question and answer word segmentation includes Questions and answers.

Specifically, the semantic matching module 402 includes:

Construction unit 4021, for constructing the first phonetic-phonetic code of the question-and-answer participle in the preset common word dictionary, and constructing the second phonetic-phonetic code of each professional word in the preset professional word dictionary, and calculating the first phonetic-phonetic code and the edit distance between the second phonetic-graph code;

The combination unit 4022 is used to combine the question and answer participle corresponding to the first phonetic-phonetic code whose editing distance is less than the preset editing distance threshold to obtain the question-answering word group, and select the second phonetic-phonetic code corresponding to the second phonetic-phonetic code whose editing distance is less than the editing distance threshold professional term;

The replacement unit 4023 is used to replace the corresponding question and answer participle in the question and answer participle with the selected professional words in turn, so as to obtain a plurality of professional phrases corresponding to the question and answer participle;

Semantic analysis unit 4024, configured to perform semantic analysis on the question and answer phrases to obtain a first semantic analysis result, and perform semantic analysis on each of the professional phrases to obtain multiple second semantic analysis results;

The comparison unit 4025 is configured to compare the first semantic analysis result with each of the second semantic analysis results, and select each question and answer in the question and answer sub-phrase group from a plurality of professional phrases according to the comparison result A synonym of the participle; use the selected synonym as the question and answer professional participle corresponding to the question and answer participle, wherein the question and answer professional participle includes an inquiry professional participle and a response professional participle.

Specifically, the comparison unit 4025 is also used for:

According to the result of the comparison, respectively calculate the degree of difference between the first semantic analysis result and each of the second semantic analysis results, and according to the degree of difference, determine the difference between the question and answer phrase group and each of the professional phrases the similarity;

Classify the professional phrases where each question and answer participle corresponds to the professional word in the question and answer participle, and obtain the professional phrases of a plurality of question and answer participle categories;

Select the professional phrase with the highest similarity from the professional phrases of each question and answer word segmentation category, and use the professional words corresponding to the question and answer word segmentation category in the selected professional phrases as synonyms for the question and answer word segmentation.

Specifically, the combining module 404 includes:

The traversal unit 4041 is configured to use the diagnostic statement to perform hierarchical traversal in the preset priori medical knowledge base, and determine the diagnosis result corresponding to the diagnostic statement according to the result of the hierarchical traversal;

The screening unit 4042 is configured to select the therapeutic product identification information matching the diagnosis result from the prior knowledge base, and obtain the therapeutic product information mapped to the therapeutic product identification information, wherein the therapeutic product Information includes referral links and summary information for therapeutic products.

Specifically, the artificial intelligence-based answer corpus generation device also includes an encryption module 405, which is used for:

Using a preset fully homomorphic encryption algorithm, converting the medical inquiry corpus and the response corpus to be pushed into corresponding base plaintext;

Encrypting the base plaintext to obtain encrypted corpus, and calculating the original ciphertext, inverse ciphertext and complement of ciphertext of the encrypted corpus according to the preset modulus;

Using the original code of the ciphertext, the inverse code of the ciphertext, and the complementary code of the ciphertext, perform a modular operation on the encrypted corpus to obtain a modular encrypted corpus, wherein the modular encrypted corpus includes the corresponding The first modular encrypted corpus and the second encrypted corpus corresponding to the response corpus to be pushed;

The first encrypted corpus is used as a new interrogation corpus, and the second encrypted corpus is used as a new response corpus to be pushed.

In the embodiment of the present application, the phonetic-phonetic codes of question and answer word segmentation and professional word segmentation are constructed through the preset common word dictionary and professional word dictionary, and the number of each question and answer word segmentation is determined through the matching of phonetic-phonetic codes. Synonyms, and replace them, to get the corresponding professional word segmentation and response professional word segmentation, which is more accurate in subsequent product matching; by further encrypting the query corpus and the response corpus to be pushed, and by calculating the ciphertext , product recommendation and other data processing processes can better guarantee the patient's personal privacy information and improve the patient's consultation experience.

Figure 4 and Figure 5 above describe in detail the artificial intelligence-based response corpus generation device in the embodiment of the present application from the perspective of modular functional entities, and the following describes the artificial intelligence-based response corpus generation device in the embodiment of the present application from the perspective of hardware processing Describe in detail.

FIG. 6 is a schematic structural diagram of an artificial intelligence-based response corpus generation device provided by an embodiment of the present application. The artificial intelligence-based response corpus generation device 600 may have relatively large differences due to different configurations or performances, and may include one or More than one processor (central processing units, CPU) 610 (for example, one or more processors) and memory 620, one or more storage media 630 for storing application programs 633 or data 632 (for example, one or more mass storage devices ). Wherein, the memory 620 and the storage medium 630 may be temporary storage or persistent storage. The program stored in the storage medium 630 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations for the artificial intelligence-based answer corpus generation device 600 . Furthermore, the processor 610 may be configured to communicate with the storage medium 630 , and execute a series of instruction operations in the storage medium 630 on the artificial intelligence-based response corpus generating device 600 .

The artificial intelligence-based answer corpus generating device 600 may also include one or more power sources 640, one or more wired or wireless network interfaces 650, one or more input and output interfaces 660, and/or, one or more operating systems 631 , such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art can understand that the structure of the artificial intelligence-based answer corpus generation device shown in FIG. 6 does not constitute a limitation to the artificial intelligence-based answer corpus generation device, and may include more or less components than those shown in the illustration, or Combining certain parts, or different arrangements of parts.

The present application also provides an artificial intelligence-based response corpus generation device, the computer device includes a memory and a processor, and computer-readable instructions are stored in the memory, and when the computer-readable instructions are executed by the processor, the processor executes the above-mentioned tasks. The steps of the artificial intelligence-based response corpus generation method in the embodiments.

The present application also provides a computer-readable storage medium. The computer-readable storage medium may be a non-volatile computer-readable storage medium. The computer-readable storage medium may also be a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium, and when the instructions are run on the computer, the computer is made to execute the steps of the method for generating answer corpus based on artificial intelligence as follows: obtaining the question corpus and the information corresponding to the question corpus The response corpus to be pushed, and based on the preset linear chain conditional random field, respectively perform word segmentation processing on the inquiry corpus and the response corpus to be pushed, correspondingly obtain a plurality of question segmentation words and a plurality of response word segmentation; The questioning participle and the response participle carry out professional word semantic matching, correspondingly obtain the questioning specialty participle corresponding to the described questioning participle and the answering specialty participle corresponding to the response participle; According to the results of the cross-question-answer matching, the professional word segmentation of the inquiry and the professional word segmentation of the response are combined to obtain the diagnostic sentence; the pre-set prior medical knowledge base is used to match the corresponding diagnostic sentence. Combining the treatment product information with the response corpus to be pushed to obtain and push the new response corpus to be pushed.

Those skilled in the art can clearly understand that for the convenience and brevity of the description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the foregoing method embodiment, which will not be repeated here.

If the integrated unit is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or part of the contribution to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , including several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (read-only memory, ROM), random access memory (random access memory, RAM), magnetic disk or optical disc and other media that can store program codes. .

As mentioned above, the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: it can still understand the foregoing The technical solutions described in each embodiment are modified, or some of the technical features are equivalently replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the application.

Claims (20)

A method for generating answer corpus based on artificial intelligence, wherein the method for generating answer corpus based on artificial intelligence includes: Obtaining the question corpus and the response corpus to be pushed corresponding to the question corpus, and performing word segmentation processing on the question corpus and the response corpus to be pushed respectively based on the preset linear chain conditional random field, correspondingly obtaining a plurality of question Diagnosis participle and multiple response participle; Carry out professional word semantic matching to described question participle and described response participle respectively, correspondingly obtain the professional question participle corresponding to described question participle and the response professional participle corresponding to described response participle; Carry out cross question-and-answer matching on each of the professional participle of the inquiry and each of the professional participle of the response in turn, and according to the result of the cross question-answer matching, combine the professional participle of the inquiry and the professional participle of the response to obtain a diagnostic sentence; The preset prior medical knowledge base is used to match the treatment product information corresponding to the diagnosis statement, and the treatment product information and the response corpus to be pushed are combined to obtain and push a new response corpus to be pushed. The method for generating response corpus based on artificial intelligence according to claim 1, wherein, the word segmentation processing is performed on the question corpus and the response corpus to be pushed respectively, and a plurality of question segmentation words and a plurality of response segmentation words are correspondingly obtained include: Extracting character feature vectors and corresponding pinyin feature vectors of the question-and-answer corpus, wherein the question-and-answer corpus includes question-and-answer corpus and response corpus to be pushed; Splicing the character feature vector and the corresponding pinyin feature vector to obtain a context information vector, and performing semantic analysis on the context information vector to obtain semantic features; Using the preset linear chain conditional random field to mark the semantic features to obtain a word segmentation tagging sequence, and decoding the word segmentation tagging sequence to obtain a plurality of question and answer word segmentation, wherein the question and answer word segmentation includes question word segmentation and response word segmentation . The method for generating response corpus based on artificial intelligence according to claim 2, wherein said professional word semantic matching is performed on said question participle and said response participle respectively, and correspondingly obtains the inquiry specialty corresponding to said question participle Participle and the response professional participle corresponding to the response participle include: Construct the first phonetic-graphic code of the question-and-answer participle in the preset common word dictionary, and construct the second phonetic-graphic code of each professional word in the preset professional word dictionary, and calculate the first phonetic-shaped code and the second phonetic-shaped code The edit distance between; Combining the question-and-answer participle corresponding to the first phonetic-phonetic code whose edit distance is less than the preset edit-distance threshold, obtains the question-and-answer word group, and selecting the professional word corresponding to the second phonetic-phonetic code whose edit distance is less than the edit-distance threshold; Adopting selected professional words to replace the corresponding question and answer participle in the question and answer participle in turn, obtain a plurality of professional phrases corresponding to the question and answer participle; Performing semantic analysis on the question and answer phrases to obtain a first semantic analysis result, and performing semantic analysis to each of the professional phrases to obtain a plurality of second semantic analysis results; Comparing the first semantic analysis result with each of the second semantic analysis results, and selecting synonyms for each question and answer participle in the question and answer participle group from a plurality of the professional phrases according to the results of the comparison; The selected synonyms are used as the question and answer professional participle corresponding to the question and answer participle, wherein the question and answer professional participle includes the question and answer professional participle. The artificial intelligence-based answer corpus generation method according to claim 3, wherein, according to the result of the comparison, selecting the synonyms of each question and answer participle in the question and answer participle group from a plurality of said professional phrases includes: According to the result of the comparison, respectively calculate the degree of difference between the first semantic analysis result and each of the second semantic analysis results, and according to the degree of difference, determine the difference between the question and answer phrase group and each of the professional phrases the similarity; Classify the professional phrases where each question and answer participle corresponds to the professional word in the question and answer participle, and obtain the professional phrases of a plurality of question and answer participle categories; Select the professional phrase with the highest similarity from the professional phrases of each question and answer word segmentation category, and use the professional words corresponding to the question and answer word segmentation category in the selected professional phrases as synonyms of the question and answer word segmentation. According to the artificial intelligence-based response corpus generation method according to any one of claims 1-4, wherein said use of a preset prior medical knowledge base, matching the treatment product information corresponding to the diagnosis statement includes: Using the diagnostic statement, perform hierarchical traversal in the preset priori medical knowledge base, and determine the diagnosis result corresponding to the diagnostic statement according to the result of the hierarchical traversal; Selecting the therapeutic product identification information matching the diagnosis result from the prior knowledge base, and obtaining the therapeutic product information mapped to the therapeutic product identification information, wherein the therapeutic product information includes the recommendation of the therapeutic product Links and Summary Information. The artificial intelligence-based answer corpus generation method according to claim 1, wherein, after said obtaining the question corpus and the answer corpus to be pushed corresponding to the question corpus, further comprising: Using a preset fully homomorphic encryption algorithm, converting the medical inquiry corpus and the response corpus to be pushed into corresponding base plaintext; Encrypting the base plaintext to obtain encrypted corpus, and calculating the original ciphertext, inverse ciphertext and complement of ciphertext of the encrypted corpus according to the preset modulus; Using the original code of the ciphertext, the inverse code of the ciphertext, and the complementary code of the ciphertext, perform a modular operation on the encrypted corpus to obtain a modular encrypted corpus, wherein the modular encrypted corpus includes the corresponding The first modular encrypted corpus and the second encrypted corpus corresponding to the response corpus to be pushed; The first encrypted corpus is used as a new interrogation corpus, and the second encrypted corpus is used as a new response corpus to be pushed. An artificial intelligence-based response corpus generation device, wherein the artificial intelligence-based response corpus generation device includes: a memory and at least one processor, instructions are stored in the memory; The at least one processor invokes the instruction in the memory, so that the artificial intelligence-based answer corpus generation device executes the following steps of the artificial intelligence-based answer corpus generation method: Obtaining the question corpus and the response corpus to be pushed corresponding to the question corpus, and performing word segmentation processing on the question corpus and the response corpus to be pushed respectively based on the preset linear chain conditional random field, correspondingly obtaining a plurality of question Diagnosis participle and multiple response participle; Carry out professional word semantic matching to described question participle and described response participle respectively, correspondingly obtain the professional question participle corresponding to described question participle and the response professional participle corresponding to described response participle; Carry out cross question-and-answer matching on each of the professional participle of the inquiry and each of the professional participle of the response in turn, and according to the result of the cross question-answer matching, combine the professional participle of the inquiry and the professional participle of the response to obtain a diagnostic sentence; The preset prior medical knowledge base is used to match the treatment product information corresponding to the diagnosis statement, and the treatment product information and the response corpus to be pushed are combined to obtain and push a new response corpus to be pushed. The artificial intelligence-based response corpus generating device according to claim 7, wherein said word segmentation processing is performed on said question corpus and said response corpus to be pushed, correspondingly obtaining a plurality of question segmentation words and a plurality of response segmentation words include: Extracting character feature vectors and corresponding pinyin feature vectors of the question-and-answer corpus, wherein the question-and-answer corpus includes question-and-answer corpus and response corpus to be pushed; Splicing the character feature vector and the corresponding pinyin feature vector to obtain a context information vector, and performing semantic analysis on the context information vector to obtain semantic features; Using the preset linear chain conditional random field to mark the semantic features to obtain a word segmentation tagging sequence, and decoding the word segmentation tagging sequence to obtain a plurality of question and answer word segmentation, wherein the question and answer word segmentation includes question word segmentation and response word segmentation . According to the artificial intelligence-based response corpus generation device according to claim 8, wherein said professional word semantic matching is performed on said question participle and said response participle respectively, and correspondingly obtains the inquiry specialty corresponding to said question participle Participle and the response professional participle corresponding to the response participle include: Construct the first phonetic-graphic code of the question-and-answer participle in the preset common word dictionary, and construct the second phonetic-graphic code of each professional word in the preset professional word dictionary, and calculate the first phonetic-shaped code and the second phonetic-shaped code The edit distance between; Combining the question-and-answer participle corresponding to the first phonetic-phonetic code whose edit distance is less than the preset edit-distance threshold, obtains the question-and-answer word group, and selecting the professional word corresponding to the second phonetic-phonetic code whose edit distance is less than the edit-distance threshold; Adopting selected professional words to replace the corresponding question and answer participle in the question and answer participle in turn, obtain a plurality of professional phrases corresponding to the question and answer participle; Performing semantic analysis on the question and answer phrases to obtain a first semantic analysis result, and performing semantic analysis to each of the professional phrases to obtain a plurality of second semantic analysis results; Comparing the first semantic analysis result with each of the second semantic analysis results, and selecting synonyms for each question and answer participle in the question and answer participle group from a plurality of the professional phrases according to the results of the comparison; The selected synonyms are used as the question and answer professional participle corresponding to the question and answer participle, wherein the question and answer professional participle includes the question and answer professional participle. According to the artificial intelligence-based response corpus generation device according to claim 9, wherein, according to the result of the comparison, selecting synonyms of each question and answer participle in the question and answer participle group from a plurality of said professional phrases includes: According to the result of the comparison, respectively calculate the degree of difference between the first semantic analysis result and each of the second semantic analysis results, and according to the degree of difference, determine the difference between the question and answer phrase group and each of the professional phrases the similarity; Classify the professional phrases where each question and answer participle corresponds to the professional word in the question and answer participle, and obtain the professional phrases of a plurality of question and answer participle categories; Select the professional phrase with the highest similarity from the professional phrases of each question and answer word segmentation category, and use the professional words corresponding to the question and answer word segmentation category in the selected professional phrases as synonyms for the question and answer word segmentation. According to the artificial intelligence-based response corpus generation device according to any one of claims 7-10, wherein said use of a preset prior medical knowledge base, matching the treatment product information corresponding to the diagnosis statement includes: Using the diagnostic statement, perform hierarchical traversal in the preset priori medical knowledge base, and determine the diagnosis result corresponding to the diagnostic statement according to the result of the hierarchical traversal; Selecting the therapeutic product identification information matching the diagnosis result from the prior knowledge base, and obtaining the therapeutic product information mapped to the therapeutic product identification information, wherein the therapeutic product information includes the recommendation of the therapeutic product Links and Summary Information. The artificial intelligence-based answer corpus generating device according to claim 7, wherein, after said acquiring the question corpus and the answer corpus to be pushed corresponding to the question corpus, further comprising: Using a preset fully homomorphic encryption algorithm, converting the medical inquiry corpus and the response corpus to be pushed into corresponding base plaintext; Encrypting the base plaintext to obtain encrypted corpus, and calculating the original ciphertext, inverse ciphertext and complement of ciphertext of the encrypted corpus according to the preset modulus; Using the original code of the ciphertext, the inverse code of the ciphertext, and the complementary code of the ciphertext, perform a modular operation on the encrypted corpus to obtain a modular encrypted corpus, wherein the modular encrypted corpus includes the corresponding The first modular encrypted corpus and the second encrypted corpus corresponding to the response corpus to be pushed; The first encrypted corpus is used as a new interrogation corpus, and the second encrypted corpus is used as a new response corpus to be pushed. A computer-readable storage medium, the computer-readable storage medium is stored with instructions, wherein, when the instructions are executed by a processor, the steps of the method for generating answer corpus based on artificial intelligence are realized as follows: Obtaining the question corpus and the response corpus to be pushed corresponding to the question corpus, and performing word segmentation processing on the question corpus and the response corpus to be pushed respectively based on the preset linear chain conditional random field, correspondingly obtaining a plurality of question Diagnosis participle and multiple response participle; Carry out professional word semantic matching to described question participle and described response participle respectively, correspondingly obtain the professional question participle corresponding to described question participle and the response professional participle corresponding to described response participle; Carry out cross question-and-answer matching on each of the professional participle of the inquiry and each of the professional participle of the response in turn, and according to the result of the cross question-answer matching, combine the professional participle of the inquiry and the professional participle of the response to obtain a diagnostic sentence; The preset prior medical knowledge base is used to match the treatment product information corresponding to the diagnosis statement, and the treatment product information and the response corpus to be pushed are combined to obtain and push a new response corpus to be pushed. The computer-readable storage medium according to claim 13, wherein said performing word segmentation processing on said question corpus and said response corpus to be pushed respectively, correspondingly obtaining a plurality of question segmentation words and multiple response word segmentations comprises: Extracting character feature vectors and corresponding pinyin feature vectors of the question-and-answer corpus, wherein the question-and-answer corpus includes question-and-answer corpus and response corpus to be pushed; Splicing the character feature vector and the corresponding pinyin feature vector to obtain a context information vector, and performing semantic analysis on the context information vector to obtain semantic features; Using the preset linear chain conditional random field to mark the semantic features to obtain a word segmentation tagging sequence, and decoding the word segmentation tagging sequence to obtain a plurality of question and answer word segmentation, wherein the question and answer word segmentation includes question word segmentation and response word segmentation . The computer-readable storage medium according to claim 14, wherein the professional word semantic matching is performed on the question participle and the response participle respectively, and correspondingly obtain the professional participle of the questioning and the corresponding participle of the question participle. The professional participle corresponding to the above participles include: Construct the first phonetic-graphic code of the question-and-answer participle in the preset common word dictionary, and construct the second phonetic-graphic code of each professional word in the preset professional word dictionary, and calculate the first phonetic-shaped code and the second phonetic-shaped code The edit distance between; Combining the question-and-answer participle corresponding to the first phonetic-phonetic code whose edit distance is less than the preset edit-distance threshold, obtains the question-and-answer word group, and selecting the professional word corresponding to the second phonetic-phonetic code whose edit distance is less than the edit-distance threshold; Adopting selected professional words to replace the corresponding question and answer participle in the question and answer participle in turn, obtain a plurality of professional phrases corresponding to the question and answer participle; Performing semantic analysis on the question and answer phrases to obtain a first semantic analysis result, and performing semantic analysis to each of the professional phrases to obtain a plurality of second semantic analysis results; Comparing the first semantic analysis result with each of the second semantic analysis results, and selecting synonyms for each question and answer participle in the question and answer participle group from a plurality of the professional phrases according to the results of the comparison; The selected synonyms are used as the question and answer professional participle corresponding to the question and answer participle, wherein the question and answer professional participle includes the question and answer professional participle. The computer-readable storage medium according to claim 15, wherein, according to the result of the comparison, selecting the synonyms of each question and answer participle in the question and answer participle from a plurality of said professional phrases includes: According to the result of the comparison, respectively calculate the degree of difference between the first semantic analysis result and each of the second semantic analysis results, and according to the degree of difference, determine the difference between the question and answer phrase group and each of the professional phrases the similarity; Classify the professional phrases where each question and answer participle corresponds to the professional word in the question and answer participle, and obtain the professional phrases of a plurality of question and answer participle categories; Select the professional phrase with the highest similarity from the professional phrases of each question and answer word segmentation category, and use the professional words corresponding to the question and answer word segmentation category in the selected professional phrases as synonyms for the question and answer word segmentation. The computer-readable storage medium according to any one of claims 13-16, wherein, using the preset prior medical knowledge base, matching the treatment product information corresponding to the diagnostic statement includes: Using the diagnostic statement, perform hierarchical traversal in the preset priori medical knowledge base, and determine the diagnosis result corresponding to the diagnostic statement according to the result of the hierarchical traversal; Selecting the therapeutic product identification information matching the diagnosis result from the prior knowledge base, and obtaining the therapeutic product information mapped to the therapeutic product identification information, wherein the therapeutic product information includes the recommendation of the therapeutic product Links and Summary Information. The computer-readable storage medium according to claim 13, wherein, after the acquisition of the medical inquiry corpus and the response corpus to be pushed corresponding to the medical inquiry corpus, further comprising: Using a preset fully homomorphic encryption algorithm, converting the medical inquiry corpus and the response corpus to be pushed into corresponding base plaintext; Encrypting the base plaintext to obtain encrypted corpus, and calculating the original ciphertext, inverse ciphertext and complement of ciphertext of the encrypted corpus according to the preset modulus; Using the original code of the ciphertext, the inverse code of the ciphertext, and the complementary code of the ciphertext, perform a modular operation on the encrypted corpus to obtain a modular encrypted corpus, wherein the modular encrypted corpus includes the corresponding The first modular encrypted corpus and the second encrypted corpus corresponding to the response corpus to be pushed; The first encrypted corpus is used as a new interrogation corpus, and the second encrypted corpus is used as a new response corpus to be pushed. An artificial intelligence-based response corpus generation device, wherein the artificial intelligence-based response corpus generation device includes: The word segmentation module is used to obtain the query corpus and the response corpus to be pushed corresponding to the query corpus, and perform word segmentation processing on the query corpus and the response corpus to be pushed based on the preset linear chain conditional random field, respectively, Correspondingly obtain multiple question and answer participle words; The semantic matching module is used to carry out semantic matching of professional words on the question participle and the response participle respectively, and correspondingly obtain the professional question participle corresponding to the question participle and the response professional participle corresponding to the response participle; A question-and-answer matching module, which is used to sequentially perform cross question-and-answer matching on each of the professional participle of the inquiry and each of the professional participle of the response, and combine the participle of the professional participle of the inquiry with the professional participle of the response according to the result of the cross-question matching , get the diagnostic statement; The combination module is used to match the treatment product information corresponding to the diagnostic statement using the preset prior medical knowledge base, and combine the treatment product information and the response corpus to be pushed to obtain and push new response corpus to be pushed. The artificial intelligence-based answer corpus generation device according to claim 19, wherein, the artificial intelligence-based answer corpus generation device also includes an encryption module for: Using a preset fully homomorphic encryption algorithm, converting the medical inquiry corpus and the response corpus to be pushed into corresponding base plaintext; Encrypting the base plaintext to obtain encrypted corpus, and calculating the original ciphertext, inverse ciphertext and complement of ciphertext of the encrypted corpus according to the preset modulus; Using the original code of the ciphertext, the inverse code of the ciphertext, and the complementary code of the ciphertext, perform a modular operation on the encrypted corpus to obtain a modular encrypted corpus, wherein the modular encrypted corpus includes the corresponding The first modular encrypted corpus and the second encrypted corpus corresponding to the response corpus to be pushed; The first encrypted corpus is used as a new interrogation corpus, and the second encrypted corpus is used as a new response corpus to be pushed.

Images