Intelligent question and answer data processing system based on artificial intelligence

Abstract

The invention discloses an intelligent question and answer data processing system based on artificial intelligence, and belongs to the technical field of intelligent question and answer. The platform end comprises a data statistics module, a knowledge base and a supplementary analysis module; the data statistics module is used for performing real-time statistical analysis on the intelligent question and answer data of each user to generate a question and answer detail graph; the knowledge base is used for storing knowledge data; the supplementary analysis module is used for performing supplementary analysis on the knowledge base according to the question and answer detail graph; the user side comprises an intention analysis module, a personal reserve library and a question and answer module; the personal reserve library is used for storing question-answer reserve data of the user; the intention analysis module is used for performing intention analysis on the user problem data according to the personal reserve library to obtain an intention analysis list; displaying the intention analysis list to the user in real time; determining a target intention according to the intention analysis list, and sending the target intention to a question and answer module; and the question-answering module is used for performing response analysis to obtain response data of the question data of the corresponding user.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent question answering, and specifically is an intelligent question answering data processing system based on artificial intelligence. Background Art

[0002] Intelligent question-answering systems usually provide users with personalized information services in the form of one question and one answer. Intelligent question-answering systems usually classify the accumulated disordered text data in an orderly and scientific manner through natural language processing related technologies, establish various classification models, and combine information retrieval and information extraction technologies to return answers to users.

[0003] Existing intelligent question-answering systems still face many challenges in terms of user question-answering, such as not accurately understanding the user's question-answering intentions, resulting in the answer data not meeting user needs, requiring users to ask questions again, add qualifiers, etc., affecting the user experience.

[0004] Based on this, in order to solve the problem that the existing intelligent question and answer cannot accurately answer user questions, the present invention provides an intelligent question and answer data processing system based on artificial intelligence. Summary of the invention

[0005] In order to solve the problems existing in the above scheme, the present invention provides an intelligent question and answer data processing system based on artificial intelligence.

[0006] The purpose of the present invention can be achieved through the following technical solutions: An intelligent question-answering data processing system based on artificial intelligence, including a platform end and a user end; The platform end includes a data statistics module, a knowledge base and a supplementary analysis module; The data statistics module is used to perform real-time statistical analysis on the intelligent question and answer data of each user, generate a question and answer detail graph, and the question and answer detail graph is used to count the field accuracy of each professional field; and send the question and answer detail graph to the user end.

[0007] Furthermore, the generation of the question-answer detail graph includes: Acquire intelligent question and answer data from each user in real time, wherein the intelligent question and answer data includes user question data, answer data and user behavior records; The intelligent question and answer data are classified according to professional fields to obtain the field question and answer data of the professional fields; the responses of the field question and answer data are checked to obtain the field accuracy of the professional fields; and a question and answer detail graph is generated according to the field accuracy of each professional field.

[0008] Furthermore, the domain question and answer data is checked for responses, including: A behavior judgment model is established. The expression of the behavior judgment model is: ; Where: w i is the input data, which represents the user behavior record of the corresponding field question and answer data, i represents the corresponding field question and answer data in the professional field, i=1, 2, ..., n, n is the number of corresponding field question and answer data in the professional field; the output data is the behavior judgment value XP (w i ), the behavior judgment value is 1 or 0; Identify the user behavior records of the field question and answer data in the professional field, analyze the user behavior records through the behavior judgment model, and obtain the behavior judgment value of the field question and answer data; calculate the field accuracy of the professional field according to the field calculation formula, and the field calculation formula is: ; Where: LQ is the domain accuracy; XP (w i ) is the behavior judgment value of the question and answer data in the corresponding field.

[0009] Furthermore, the user demand share of the corresponding professional field is supplemented in the question and answer details graph.

[0010] The knowledge base is used to store knowledge data.

[0011] The supplementary analysis module is used to perform supplementary analysis on the knowledge base according to the question and answer detail graph, obtain knowledge supplementary data, and store the knowledge supplementary data in the knowledge base.

[0012] Furthermore, the knowledge base is supplemented with analysis based on the question-answer detail graph, including: The platform sets the unit time period and the benchmark value; the unit value corresponding to each professional field in the corresponding unit time period is counted in real time, and the unit value is the total number of questions and answers of users on the corresponding professional field in the unit time period; According to the question and answer details, the field accuracy rate corresponding to each professional field is identified in real time; the supplementary value of the corresponding professional field is calculated according to the supplementary formula, which is: ; Where: BC is the supplementary value; e is the natural constant; DZ is the unit value; BZ is the benchmark value; LQ is the domain accuracy; Determine whether the professional field needs to be supplemented with knowledge data according to the supplement value; When it is determined that the professional field needs to be supplemented with knowledge data, the knowledge data that needs to be supplemented in the professional field is collected and the knowledge data is supplemented into the knowledge base; When it is determined that the professional field does not need to supplement the knowledge data, no corresponding supplement processing is performed.

[0013] The user terminal includes an intention analysis module, a personal reserve library, and a question-answering module; The personal reserve library is used to store the user's question and answer reserve data, and the question and answer reserve data includes user question data, initial intention sequence, and target intention; the initial intention sequence is the sorted sequence of answer intentions corresponding to the corresponding user question data; the target intention is the answer intention that meets the user's answer requirements.

[0014] Furthermore, the personal repository is also provided with a personal knowledge unit, and the personal knowledge unit is used to store the user's personal knowledge data; the personal knowledge unit is associated with the knowledge repository.

[0015] The intent analysis module is used to perform intent analysis on user question data based on a personal reserve library to obtain an intent analysis list; display the intent analysis list to the user in real time; determine the target intent based on the intent analysis list, and send the target intent to the question and answer module.

[0016] Furthermore, the user's question data is analyzed for intent based on the personal database, including: Identify the question and answer reserve data stored in the personal reserve library, identify the professional fields corresponding to the question and answer reserve data, count the number of question and answer reserve data corresponding to the corresponding professional fields in real time, and mark them as field question and answer values; calculate the field proportion of the corresponding professional fields according to the field question and answer values ​​of each professional field, and mark them as field tendency values; Performing intent analysis on the user question data to determine a number of answer intentions; performing priority analysis on the answer intentions based on the domain tendency value and the question and answer reserve data in the personal reserve library to obtain the priority of the answer intention, and generating an intent analysis list based on the priority of the answer intention.

[0017] The question-answer module is used to perform answer analysis, obtain answer data corresponding to user question data, and display the answer data to the user.

[0018] Further, when the response data does not meet the user's requirements, the user determines a new target intent based on the intent analysis list.

[0019] Compared with the prior art, the present invention has the following beneficial effects: Through the mutual cooperation between various modules, the semantic content of user input can be understood more accurately, thereby accurately identifying the user's question and answer intention. Compared with traditional methods based on keyword matching or simple rules, the recognition ability of the present invention in complex contexts and diversified expressions has been significantly improved, and the misrecognition rate has been effectively reduced; this is of great significance for improving user experience and enhancing system practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 This is a principle block diagram of the present invention. DETAILED DESCRIPTION

[0022] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] like Figure 1 As shown, an intelligent question-answering data processing system based on artificial intelligence includes a platform end and a user end; The platform end is used for operations such as use and maintenance by the platform party, including a data statistics module, a knowledge base and a supplementary analysis module; The data statistics module is used to perform statistical analysis on the intelligent question and answer data of each user, and obtain the intelligent question and answer data of each user in real time. The intelligent question and answer data includes user question data, answer data and user behavior records. User behavior records refer to behavior records related to proving whether the answer data is accurate, such as regeneration, adding qualifiers to ask again, asking the next question, user feedback on inaccurate answers, etc. The user behavior records are collected specifically according to the feedback form of the answer results of the question and answer intelligent body; the intelligent question and answer data are classified according to professional fields to obtain the field question and answer data of the corresponding professional fields; the field question and answer data are checked for answers to obtain the field accuracy of the question and answer intelligent body in the corresponding professional field; a question and answer detail graph is generated according to the field accuracy of each professional field; the question and answer detail graph is sent to the user end to facilitate the user to understand the response situation in each professional field.

[0024] In one embodiment, the domain question and answer data are checked for responses. The domain question and answer data in a professional field may be statistically analyzed using existing data statistics methods to obtain the amount of data corresponding to correct responses and incorrect responses, and then calculate the corresponding domain accuracy rate.

[0025] In one embodiment, the domain question and answer data is checked for answers, including: Establish a behavior judgment model. The behavior judgment model is used to analyze user behavior records and determine whether the answer data corresponding to the user behavior record meets the user's question and answer needs. Use a large amount of historical intelligent question and answer data to set up a training set for training, mark the verification results of the corresponding user behavior records, and then conduct training. You can also use an enumeration method to list various user behavior records that meet the user's question and answer needs and those that do not meet the user's question and answer needs, integrate and establish a behavior library, and then verify according to the behavior library to determine whether it meets the user's question and answer needs. The expression of the behavior judgment model is: ; Where: w i is the input data, which represents the user behavior record of the corresponding field question and answer data, i represents the corresponding field question and answer data in the professional field, i=1, 2, ..., n, n is the number of corresponding field question and answer data in the professional field; the output data is the behavior judgment value XP (w i ), the behavior judgment value is 1 or 0; Identify the user behavior records of the corresponding field question and answer data in the professional field, analyze the user behavior records through the behavior judgment model, and obtain the behavior judgment value of the corresponding field question and answer data; calculate the field accuracy of the corresponding professional field according to the field calculation formula, and the field calculation formula is: ; Where: LQ is the domain accuracy; XP (w i ) is the behavior judgment value of the question and answer data in the corresponding field.

[0026] In one embodiment, in order to facilitate the platform to optimize the question and answer agent based on the question and answer details graph, the user demand share of the corresponding professional field is supplemented in the question and answer details graph, and the user demand share is calculated based on the proportion of the number of questions and answers in the professional field within a preset time period.

[0027] The knowledge base is used to store various knowledge data required for the question-answering agent to answer, and is stored and managed based on a preset data management method.

[0028] The supplementary analysis module is used to perform supplementary analysis on the knowledge base according to the question and answer detail graph, obtain corresponding knowledge supplementary data, and store the knowledge supplementary data in the knowledge base.

[0029] In one embodiment, the knowledge base is supplemented with analysis based on the question-answer detail graph, including: Set a unit period. The unit period is an evaluation cycle, which is used to periodically count the total number of questions and answers in the corresponding professional field within the unit period, which is regarded as the unit value. It is generally a day, a week, etc., and is set according to the user's activity level and the number of users in the question-and-answer agent. It is used to increase credibility and representativeness, and is set by the platform. The platform also sets a benchmark value. The sum of the unit values ​​of each professional field within the unit period can be used as the benchmark value, or the sum of the unit values ​​of each professional field within a certain period of time can be directly specified as the benchmark value, or a certain value can be directly set as the benchmark value. The specific setting is made by the platform according to actual needs.

[0030] Real-time statistics are collected on the unit values ​​corresponding to each professional field in the corresponding unit period, and the field accuracy corresponding to each professional field is identified in real time according to the question and answer details diagram; the supplementary value of the corresponding professional field is calculated according to the supplementary formula, and the supplementary formula is: ; Where: BC is the supplementary value; e is the natural constant; DZ is the unit value; BZ is the benchmark value; LQ is the domain accuracy; According to the supplementary value, it is judged whether the corresponding professional field needs to be supplemented with knowledge data; the supplementary value is compared with a unified preset value. If it is greater than the preset value, it is judged that knowledge data supplementation is needed; if it is not greater than the preset value, it is judged that knowledge data supplementation is not needed; the preset value is set by technicians in this field according to actual conditions or obtained by simulation of a large amount of data; When it is determined that knowledge data supplementation is needed, the knowledge data that needs to be supplemented in the corresponding professional field is collected and the corresponding knowledge data is supplemented into the knowledge base; When it is determined that knowledge data supplementation is not necessary, no corresponding processing is performed.

[0031] In one embodiment, the knowledge data that needs to be supplemented in the corresponding professional field is collected, including: According to the field question and answer data of the professional field, the answer accuracy and share ratio of the corresponding user questions are counted, and the priority of the corresponding user questions is calculated based on the answer accuracy rate and share ratio. The calculation can be based on the existing priority calculation method, such as multiplying the answer accuracy rate and share ratio to calculate the priority value, or adding a proportional coefficient of the answer accuracy rate and share ratio and then multiplying them. The priority value can also be calculated based on a variety of methods such as the exponential base, and priority sorting is performed according to the priority value.

[0032] Conduct problem-solving simulations for each user's question to determine what knowledge data needs to be collected in order to improve the accuracy of answering the user's question, and prioritize the role of the knowledge data in improving the accuracy of answering the user's question. For example, simulate and supplement the knowledge data for answer analysis to understand the improvement in the accuracy of answering.

[0033] When there is sufficient storage space, the above knowledge data is filtered out after deduplication and the like as the knowledge data that needs to be supplemented; When there is insufficient storage space, the storage capacity that is allowed to be supplemented is identified, and then the storage capacity is divided according to the priority of the user's question, and then the knowledge data that needs to be supplemented is determined based on the divided storage capacity and the priority of the knowledge data.

[0034] The user terminal includes an intention analysis module, a personal reserve library, and a question-and-answer module; The personal reserve library is used to store the user's question and answer reserve data, which includes user question data, initial intention sequence, target intention and other data. The initial intention sequence is to analyze the user question data according to the answer analysis to determine the answer intentions and the ranking of each answer intention. For example, there are three answer methods for a certain question, and each answer method corresponds to an answer intention. The initial intention sequence is determined according to the actual answer analysis results. For example, the first display is the answer data corresponding to the first-ranked answer intention. If the user is not satisfied, the answer data corresponding to the second-ranked answer intention will be regenerated; the target intention refers to the answer intention corresponding to the answer data recognized or adopted by the user; the question and answer reserve data are all statistically obtained based on the user's historical question and answer data.

[0035] In one embodiment, a personal knowledge unit is also provided in the personal reserve library. The personal knowledge unit is used to store personal knowledge data that meets the needs of users. The personal knowledge unit is generally added by users themselves and is mainly used to make the generated response data as close to the relevant personal knowledge data as possible, that is, for the data source of the response, when the personal knowledge data meets the requirements, it is used as a benchmark for response analysis or its weight is increased; the personal knowledge unit is associated with the knowledge base, that is, when performing response analysis later, it is first determined whether the personal knowledge unit contains relevant personal knowledge data. If not, no corresponding analysis is performed. If so, it is supplemented as the basis for the response, and the response analysis is performed according to the preset personal weight, such as adjusting the training of the corresponding large model so that it can realize response analysis based on personal knowledge data.

[0036] The intent analysis module is used to perform intent analysis on user question data according to a personal reserve library to obtain an intent analysis list. The intent analysis list is used to sort the response intents of the user question data, i.e., the initial intent sequence; the intent analysis list is displayed to the user in real time, and the user can manually adjust the sorting of the corresponding response intents according to the intent analysis list. Generally, the user will directly specify the corresponding response intent as the target intent when adjusting, otherwise the response intent ranked first will be the target intent; the target intent is determined according to the intent analysis list, and the target intent is sent to the question and answer module.

[0037] In one embodiment, the user question data is analyzed for intent based on the personal reserve library, including: Identify the question and answer reserve data stored in the personal reserve library, identify the professional fields corresponding to the question and answer reserve data, and count the number of question and answer reserve data corresponding to the corresponding professional fields in real time, which refers to the number of questions and answers, and mark them as field question and answer values; calculate the field proportion of the corresponding professional field based on the field question and answer values ​​of each professional field, and mark them as field tendency values; Perform intent analysis on user question data to determine the various answer intentions, that is, perform answer analysis according to the current agent answer mode to determine the various answer intentions, or determine the answer intentions of user question data through other existing answer analysis technologies; perform priority analysis on each answer intention according to the domain tendency value and the question and answer reserve data in the personal reserve library to determine the priority of each answer intention, and generate an intent analysis list according to the priority of each answer intention.

[0038] In one embodiment, considering the difference in the impact of the number of questions and answers at different times on the current field tendency value, the number of questions and answers at the corresponding time can be reduced in combination with the existing time decay technology, that is, the field question and answer value is adjusted. For example, if the number of questions and answers at the current time is 1, the number of times of a certain historical event corresponds to 0.5; the cumulative total value of the professional field is calculated, and then the field tendency value of the professional field is calculated.

[0039] In one embodiment, each answer intention is analyzed for priority based on the domain tendency value and the question and answer reserve data in the personal reserve library, that is, firstly match the user question data in the personal reserve library to determine each question and answer reserve data with a similarity that meets the requirements, such as a similarity of not less than 80%, 70%, etc., and sort each answer intention according to the initial intention sequence and target intention corresponding to the question and answer reserve data, that is, sort according to the sorting method of the initial intention sequence and the target intention or the determination change method of the target intention. If the target intention is ranked first in the initial intention sequence, then sort each answer intention according to the original analysis sorting method. If the target intention is ranked first in the initial intention sequence, then sort each answer intention according to the original analysis sorting method. If the intention is not ranked first, the ranking is adjusted according to the ranking adjustment method of the target intention to achieve the ranking of each response intention, and the corresponding ranking is marked as a reference ranking; the weight coefficient of the corresponding reference ranking is calculated according to the similarity corresponding to each reference ranking, that is, the proportion of each similarity in the whole is used as the weight coefficient; a comprehensive ranking is determined according to each reference ranking and the weight coefficient, the professional field corresponding to the corresponding response intention is identified, and the corresponding field tendency value is matched; the response intention in the comprehensive ranking is ranked according to the field tendency value, mainly to determine the response intention ranked first. If the analysis does not affect the response intention ranked first, no adjustment is required.

[0040] In one embodiment, priority analysis is performed on each answer intention based on the domain tendency value and the question and answer reserve data in the personal reserve library. The priority analysis can be performed based on existing methods, such as establishing a priority analysis model based on an existing priority evaluation algorithm or a neural network such as a CNN network or a DNN network, training the priority analysis model manually, and analyzing the successfully trained priority analysis model to determine the priority of the corresponding answer data.

[0041] The question-and-answer module is used to perform answer analysis based on user question data and target intent, obtain corresponding answer data, and generate high-quality answers using the existing large model; the specific question-and-answer analysis process is determined based on the question-and-answer intelligent agent in actual application; the answer data is displayed to the user, and when the answer data does not meet the user's needs, the user can determine the accurate target intent based on the intent analysis list.

[0042] The above formulas are all calculated by removing dimensions and taking numerical values. The formula is a formula that is closest to the actual situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained by simulating a large amount of data.

[0043] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An intelligent question-answering data processing system based on artificial intelligence, characterized in that: Including platform side and user side; The platform end includes a data statistics module, a knowledge base and a supplementary analysis module; The data statistics module is used to perform real-time statistical analysis on the intelligent question and answer data of each user and generate a question and answer detail graph, which is used to count the field accuracy of each professional field; Sending the question and answer details graph to the user terminal; The knowledge base is used to store knowledge data; The supplementary analysis module is used to perform supplementary analysis on the knowledge base according to the question-answer detail graph, obtain knowledge supplementary data, and store the knowledge supplementary data in the knowledge base; The user terminal includes an intention analysis module, a personal reserve library, and a question-and-answer module; The personal reserve library is used to store the user's question and answer reserve data, and the question and answer reserve data includes user question data, initial intention sequence, and target intention; The initial intention sequence is a sorted sequence of answer intentions corresponding to the corresponding user question data; the target intention is an answer intention that satisfies the user's answer requirements; The intention analysis module is used to perform intention analysis on user question data according to the personal reserve library to obtain an intention analysis list; Displaying the intention analysis list to the user in real time; Determine the target intent based on the intent analysis list and send the target intent to the question-answering module; The question-answer module is used to perform answer analysis, obtain answer data corresponding to user question data, and display the answer data to the user.

2. The intelligent question-answering data processing system based on artificial intelligence according to claim 1, characterized in that: The generation of question and answer detail graph includes: Acquire intelligent question and answer data from each user in real time, wherein the intelligent question and answer data includes user question data, answer data and user behavior records; The intelligent question and answer data are classified according to professional fields to obtain the field question and answer data of the professional fields; the responses of the field question and answer data are checked to obtain the field accuracy of the professional fields; and a question and answer detail graph is generated according to the field accuracy of each professional field.

3. The intelligent question-answering data processing system based on artificial intelligence according to claim 2, characterized in that: Verify responses to domain question-answering data, including: A behavior judgment model is established. The expression of the behavior judgment model is: ; Where: w i is the input data, which represents the user behavior record of the corresponding field question and answer data, i represents the corresponding field question and answer data in the professional field, i=1, 2, ..., n, n is the number of corresponding field question and answer data in the professional field; the output data is the behavior judgment value XP (w i ), the behavior judgment value is 1 or 0; Identify the user behavior records of the field question and answer data in the professional field, analyze the user behavior records through the behavior judgment model, and obtain the behavior judgment value of the field question and answer data; calculate the field accuracy of the professional field according to the field calculation formula, and the field calculation formula is: ; Where: LQ is the domain accuracy; XP (w i ) is the behavior judgment value of the question and answer data in the corresponding field.

4. The intelligent question-answering data processing system based on artificial intelligence according to claim 2, characterized in that: The user demand share in the corresponding professional field is supplemented in the question and answer detail diagram.

5. The intelligent question-answering data processing system based on artificial intelligence according to claim 1, characterized in that: The knowledge base is supplemented with analysis based on the question-answer detail graph, including: The platform sets the unit time period and the benchmark value; the unit value corresponding to each professional field in the corresponding unit time period is counted in real time, and the unit value is the total number of questions and answers of users on the corresponding professional field in the unit time period; According to the question and answer details, the field accuracy rate corresponding to each professional field is identified in real time; the supplementary value of the corresponding professional field is calculated according to the supplementary formula, which is: ; Where: BC is the supplementary value; e is the natural constant; DZ is the unit value; BZ is the benchmark value; LQ is the domain accuracy; Determine whether the professional field needs to be supplemented with knowledge data according to the supplement value; When it is determined that the professional field needs to be supplemented with knowledge data, the knowledge data that needs to be supplemented in the professional field is collected and the knowledge data is supplemented into the knowledge base; When it is determined that the professional field does not need to supplement the knowledge data, no corresponding supplement processing is performed.

6. The intelligent question-answering data processing system based on artificial intelligence according to claim 1, characterized in that: The personal repository is also provided with a personal knowledge unit, and the personal knowledge unit is used to store the user's personal knowledge data; the personal knowledge unit is associated with the knowledge repository.

7. The intelligent question-answering data processing system based on artificial intelligence according to claim 1, characterized in that: Perform intent analysis on user question data based on personal database, including: Identify the question and answer reserve data stored in the personal reserve library, identify the professional fields corresponding to the question and answer reserve data, count the number of question and answer reserve data corresponding to the corresponding professional fields in real time, and mark them as field question and answer values; calculate the field proportion of the corresponding professional fields according to the field question and answer values ​​of each professional field, and mark them as field tendency values; Performing intent analysis on the user question data to determine a number of answer intentions; performing priority analysis on the answer intentions based on the domain tendency value and the question and answer reserve data in the personal reserve library to obtain the priority of the answer intention, and generating an intent analysis list based on the priority of the answer intention.

8. The intelligent question-answering data processing system based on artificial intelligence according to claim 1, characterized in that: When the response data does not meet the user's requirements, the user determines a new target intent based on the intent analysis list.

Images