Question and answer knowledge base updating method, device, computer equipment and storage medium

Abstract

The present application relates to a method, device, computer equipment and storage medium for updating a question-and-answer knowledge base, the method comprising: obtaining a question; performing feature abstraction processing on the question to obtain an abstract feature of the question; matching the abstract feature with existing data in a preset question-and-answer knowledge base to screen out unanswered questions for which the preset question-and-answer knowledge base has no corresponding answers; obtaining answers corresponding to the unanswered questions, and updating the answers corresponding to the unanswered questions to the preset question-and-answer knowledge base. The above method uses abstract features to screen out questions that cannot be answered by the preset question-and-answer knowledge base, and only needs to obtain corresponding answers to the screened unanswered questions for updating, which reduces the workload of updating the preset question-and-answer knowledge base, shortens the time required for the update process, and reduces labor costs and time costs.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, device, computer equipment and storage medium for updating a question-and-answer knowledge base. Background Art

[0002] An important form of knowledge contained in the knowledge base of the intelligent customer service system is the "question-answer" pair. Professionals in specific fields enter this form of knowledge into the intelligent customer service knowledge base manually; when a user asks the intelligent customer service a question that is similar to the questions in the knowledge base, the intelligent customer service will return the corresponding answer to the user.

[0003] If a user asks a question that is not included in the knowledge base, the intelligent customer service will not be able to answer the user's question; as time goes by, the "question-answer" pairs in the knowledge base also need to be constantly updated so that the intelligent customer service can answer more questions raised by users. Most traditional methods use manual input of "question-answer" into the knowledge base, which is time-consuming and laborious. Summary of the invention

[0004] Based on this, it is necessary to provide a question-and-answer knowledge base updating method, device, computer equipment and storage medium to address the above technical issues, which can reduce labor costs and time costs.

[0005] A method for updating a question-and-answer knowledge base, the method comprising:

[0006] Get the question;

[0007] Performing feature abstraction processing on the problem to obtain abstract features of the problem;

[0008] Based on the abstract features, the existing data in the preset question-and-answer knowledge base is matched to filter out questions to be answered that have no corresponding answers in the preset question-and-answer knowledge base;

[0009] Obtain answers corresponding to the questions to be answered, and update the answers corresponding to the questions to be answered to the preset question and answer knowledge base.

[0010] A question-and-answer knowledge base updating device, the device comprising:

[0011] A question acquisition module is used to acquire questions;

[0012] An abstract processing module is used to perform feature abstract processing on the problem to obtain the abstract features of the problem;

[0013] A matching module, used to match the existing data in a preset question-and-answer knowledge base based on the abstract features, and filter out questions to be answered that have no corresponding answers in the preset question-and-answer knowledge base;

[0014] The updating module is used to obtain the answers corresponding to the questions to be answered, and update the answers corresponding to the questions to be answered to the preset question and answer knowledge base.

[0015] In one embodiment, the device further comprises:

[0016] A clustering module, used for dividing the questions to be answered into different sets of questions to be answered, wherein the questions to be answered contained in each set of questions to be answered belong to the same category;

[0017] A deleting module, used to delete the set of questions to be answered that contains questions with a number less than a preset number threshold, to obtain a final set of questions to be answered;

[0018] The updating module is also used to: obtain answers corresponding to the set of questions to be answered, and update the answers corresponding to the set of questions to be answered to the preset question and answer knowledge base.

[0019] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0020] Get the question;

[0021] Performing feature abstraction processing on the problem to obtain abstract features of the problem;

[0022] Based on the abstract features, the existing data in the preset question-and-answer knowledge base is matched to filter out questions to be answered that have no corresponding answers in the preset question-and-answer knowledge base;

[0023] Obtain answers corresponding to the questions to be answered, and update the answers corresponding to the questions to be answered to the preset question and answer knowledge base.

[0024] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the following steps:

[0025] Get the question;

[0026] Performing feature abstraction processing on the problem to obtain abstract features of the problem;

[0027] Based on the abstract features, the existing data in the preset question-and-answer knowledge base is matched to filter out questions to be answered that have no corresponding answers in the preset question-and-answer knowledge base;

[0028] Obtain answers corresponding to the questions to be answered, and update the answers corresponding to the questions to be answered to the preset question and answer knowledge base.

[0029] The above-mentioned question-and-answer knowledge base updating method, device, computer equipment and storage medium, after obtaining the question raised by the user, abstracts the question to obtain the abstract features corresponding to the question; based on the obtained abstract features of the question, the abstract features of the question are matched with the existing data in the preset question-and-answer knowledge base, thereby screening out the unanswered questions in the questions raised by the user that have no matching answers in the preset question-and-answer knowledge base; obtain the answers to the unanswered questions, update the answers to the preset question-and-answer knowledge base, and complete the update of the preset question-and-answer knowledge base. After extracting the abstract features of the question, the above-mentioned method uses the abstract features of the question to match with the preset question-and-answer knowledge base, screens out the unanswered questions that have no matching answers in the preset question-and-answer knowledge base, and after obtaining the answers to the unanswered questions, updates them to the preset question-and-answer knowledge base. Since the questions that cannot be answered by the preset question-and-answer knowledge base have been screened out by using the abstract features, it is only necessary to obtain the corresponding answers to the screened unanswered questions for updating, which reduces the workload of updating the preset question-and-answer knowledge base, shortens the time required for the update process, and reduces the manpower cost and time cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 An application environment diagram of a question-answer knowledge base updating method in one embodiment;

[0031] Figure 2 A schematic diagram of a process of updating a question-and-answer knowledge base in one embodiment;

[0032] Figure 3 A schematic diagram of a process of performing feature abstraction processing on a problem to obtain abstract features of the problem in an embodiment;

[0033] Figure 4 A schematic diagram of a process for determining an abstract feature of a question based on similarity scores corresponding to each preset similarity algorithm of the question in an embodiment;

[0034] Figure 5 A schematic diagram of a process for matching existing data in a preset question-and-answer knowledge base based on abstract features in one embodiment to filter out unanswered questions for which there are no corresponding answers in the preset question-and-answer knowledge base;

[0035] Figure 6 A schematic diagram of a flow chart of a question-and-answer knowledge base updating method in another embodiment;

[0036] Figure 7 It is a flowchart of a method for updating a question-answer knowledge base in a specific embodiment;

[0037] Figure 8 A schematic diagram of a process of clustering questions to be answered and obtaining clustering results in a specific embodiment;

[0038] Fig. 9 is a structural block diagram of a question-and-answer knowledge base updating device in one embodiment;

[0039] Fig.10 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0041] The question-answer knowledge base updating method provided in this application can be applied to Figure 1 In the application environment shown, the terminal 102 communicates with the server 104 via the network.

[0042] In some embodiments, after obtaining the question raised by the user, the terminal 102 abstracts the question to obtain the abstract features corresponding to the question; based on the obtained abstract features of the question, the abstract features of the question are sent to the server 104 to match with the existing data in the preset question and answer knowledge base stored in the server 104, thereby filtering out the unanswered questions in the questions raised by the user that have no matching answers in the preset question and answer knowledge base; obtain the answers to the unanswered questions, update the answers to the preset question and answer knowledge base, and complete the update of the preset question and answer knowledge base.

[0043] In other embodiments, only the terminal 102 may obtain questions and send the questions to the server 104. The server 104 abstracts the features of the questions raised by the users, matches answers in a preset question and answer knowledge base based on the abstract features, filters out unanswered questions that have no corresponding answers, obtains the answers to the unanswered questions, and updates them in the preset question and answer knowledge base.

[0044] The terminal 102 may be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, and portable wearable devices, and the server 104 may be implemented as an independent server or a server cluster consisting of multiple servers.

[0045] In one embodiment, Figure 2 As shown, a question-answering knowledge base updating method is provided, which is applied to Figure 1 The terminal in is taken as an example to illustrate, including steps S210 to S240.

[0046] Step S210, obtaining the question.

[0047] In one embodiment, the question is a question raised by the user; the question raised by the user may be a question raised by the user in a historical conversation. Further, in one embodiment, the question raised by the user is a question sent to the server; in different applications and different application scenarios, the questions raised by the user to the server are usually associated and are related to the corresponding application or the field to which the scenario belongs. For example, in medical-related applications, the questions raised by the user are all related to medical treatment; and in applications in the field of history, the questions raised by the user are mostly related to history, and so on.

[0048] In one embodiment, the question raised by the user can be obtained from a historical database, which stores a large number of questions raised by users to the server in history. In other embodiments, obtaining the question can also be achieved by any other method.

[0049] Step S220, performing feature abstraction processing on the question to obtain abstract features of the question.

[0050] Feature abstraction is the process of processing ordered and unordered text classification features in different ways, [AC1] and quantifying their category attributes. The feature corresponding to the question obtained by performing feature abstraction processing on the question is recorded as the abstract feature of the question in this embodiment; the specific process of performing feature abstraction processing on the question will be described in detail in subsequent embodiments and will not be repeated here.

[0051] Step S230, matching the abstract features with the existing data in the preset question and answer knowledge base, and filtering out the unanswered questions that have no corresponding answers in the preset question and answer knowledge base.

[0052] Among them, the preset question and answer knowledge base refers to a database that stores a large number of "question-answer" pairs. At present, many applications are set up with intelligent customer service. Users send questions to intelligent customer service. Based on the preset question and answer knowledge base, the intelligent customer service can search for matching answers from the questions raised by the user. If there are, the intelligent customer service can automatically answer the user without the intervention of manual customer service, and provide answer services to the user. The preset question and answer knowledge base is usually accumulated based on a large amount of historical question and answer dialogue data. As time changes, the questions raised by users may change. Therefore, it is necessary to update the preset question and answer knowledge base in a timely manner so that the preset question and answer knowledge base is continuously updated, and the richness of the "question-answer" pairs in the preset question and answer knowledge base is maintained, and corresponding answers can be automatically given for more questions. The question and answer knowledge base updating method provided in this application is to update the "question-answer" pairs in the preset question and answer knowledge base. In one embodiment, the existing data in the preset question and answer knowledge base includes existing questions and / or existing answers in the preset question and answer knowledge base.

[0053] The existing data in the preset question and answer knowledge base corresponds to the questions one by one, and answers can be given to the corresponding part of the questions. In this embodiment, the abstract features obtained by abstracting the acquired questions are matched in the preset question and answer knowledge base to determine whether there is a matching answer to the acquired question in the preset question and answer knowledge base. If so, it means that the preset question and answer knowledge base can give a corresponding answer to the question. It can be understood that this part of the questions no longer needs to update the answer. For questions that do not have a matching answer in the preset question and answer knowledge base, this part of the questions is recorded as unanswered questions in this embodiment. For unanswered questions, it is necessary to update the matching answers in the preset question and answer knowledge base to ensure that the intelligent customer service can automatically give matching answers based on the preset question and answer knowledge base when users ask these questions.

[0054] In this embodiment, the abstract features obtained by abstracting the question are matched in the preset question and answer knowledge base to find out whether there is an answer matching the question in the preset question and answer knowledge base. The specific process of matching the abstract features with the existing data in the preset question and answer knowledge base to screen out the unanswered questions that have no corresponding answers in the preset question and answer knowledge base will be described in detail in the subsequent embodiments and will not be repeated here.

[0055] Step S240, obtaining the answer corresponding to the question to be answered, and updating the answer corresponding to the question to be answered to the preset question and answer knowledge base.

[0056] After determining that there are no matching answers to the unanswered questions in the preset question and answer knowledge base, it is necessary to obtain the answers corresponding to the unanswered questions and update them to the preset question and answer knowledge base.

[0057] In one embodiment, obtaining the answers to the questions to be answered may be obtaining the manually input answers to the questions to be answered [AC2]; for example, in a specific embodiment, after the questions to be answered are screened, the questions to be answered are uniformly sent to relevant personnel, and the answers to the questions to be answered input by the relevant personnel are obtained. In another embodiment, the answers to the questions to be answered may also be obtained from other channels and updated to the preset question and answer knowledge base; for example, the answers to the questions to be answered may be searched and obtained from other databases, etc.

[0058] In one embodiment, updating the answer corresponding to the question to be answered to the preset question and answer knowledge base includes: updating the question to be answered and the corresponding answer as a set of "question-answer" pairs to the preset question and answer knowledge base. Furthermore, in one embodiment, the questions with corresponding answers are classified in the preset question and answer knowledge base, and when the answer corresponding to the question to be answered is updated to the preset question and answer knowledge base, the answer corresponding to the question to be answered is updated to the category to which the question to be answered belongs in the preset question and answer knowledge base. Among them, determining the category to which the question to be answered belongs can be achieved in any way, for example, the category to which the question to be answered belongs can be determined by extracting keywords in the question to be answered.

[0059] The above-mentioned question-and-answer knowledge base updating method, after obtaining the question raised by the user, abstracts the question to obtain the abstract features corresponding to the question; based on the obtained abstract features of the question, the abstract features of the question are matched with the existing data in the preset question-and-answer knowledge base, thereby screening out the unanswered questions in the questions raised by the user that have no matching answers in the preset question-and-answer knowledge base; obtain the answers to the unanswered questions, update the answers to the preset question-and-answer knowledge base, and complete the update of the preset question-and-answer knowledge base. After extracting the abstract features of the question, the above-mentioned method uses the abstract features of the question to match with the preset question-and-answer knowledge base, screens out the unanswered questions that have no matching answers in the preset question-and-answer knowledge base, and after obtaining the answers to the unanswered questions, updates them to the preset question-and-answer knowledge base. Since the questions that cannot be answered by the preset question-and-answer knowledge base have been screened out by using the abstract features, it is only necessary to obtain the corresponding answers to the screened unanswered questions for updating, which reduces the workload of updating the preset question-and-answer knowledge base, shortens the time required for the update process, and reduces the manpower cost and time cost.

[0060] In one embodiment, Figure 3 As shown, performing feature abstraction processing on the problem to obtain the abstract features of the problem includes steps S221 to S223.

[0061] Step S221, reading a preset similarity algorithm.

[0062] The preset similarity algorithm is a preset algorithm for calculating similarity, which can be set according to actual conditions; the preset similarity algorithm can include one algorithm or multiple algorithms. Common similarity algorithms include: cosine similarity and adjusted cosine similarity, Pearson Correlation Coefficient, Jaccard Coefficient, Tanimoto coefficient (generalized Jaccard similarity coefficient), log-likelihood similarity / log-likelihood similarity, mutual information / information gain, relative entropy / KL divergence, term frequency-inverse document frequency (TF-IDF) commonly used in information retrieval, and similarity algorithm based on word vector, similarity algorithm based on sentence vector, etc. In one embodiment, the preset similarity algorithm can be stored in a preset path, and the preset similarity algorithm is read from the preset path before abstracting the acquired question.

[0063] Step S222, calculating similarity scores for the question and existing data in the preset question and answer knowledge base based on various preset similarity algorithms.

[0064] In this step, each preset similarity algorithm is used to calculate the similarity score between the question and the existing data in the preset question and answer knowledge base, and the similarity score between the question and the existing data in the preset question and answer knowledge base under each algorithm is obtained [AC3]. Furthermore, when calculating the similarity score between the question and the existing data in the preset question and answer knowledge base, it is necessary to calculate the similarity score between the question and each existing data in the preset question and answer knowledge base.

[0065] Step S223, determining the abstract features of the question based on the similarity scores corresponding to the preset similarity algorithms of the question.

[0066] In this embodiment, some of the similarity scores between the question and the existing data in the preset question-answer knowledge base are used as the abstract features of the question [AC4]. In one embodiment, each preset similarity algorithm obtains the similarity score between the question and each existing data, that is, each preset similarity algorithm can calculate a group of similarity scores; in this embodiment, some target similarity scores that meet the screening requirements are selected from the similarity score group corresponding to each preset similarity algorithm as the abstract features of the question.

[0067] Among them, the screening requirements can be set according to the actual situation; for example, the similarity scores that meet the requirements can be screened according to the size of the similarity scores, which can be similarity scores with a certain or multiple fixed values, or similarity scores corresponding to fixed order positions from large to small scores [AC5].

[0068] Further, in one embodiment, if Figure 4 As shown, based on the similarity scores corresponding to each preset similarity algorithm of the question, determining the abstract features of the question includes steps S2231 to S2233.

[0069] Step S2231, for any preset similarity algorithm, sort the similarity scores between the question and the existing data in the preset question and answer knowledge base according to the score values ​​to obtain a similarity sorting result.

[0070] It can be understood that the sorting based on the similarity score can be from large to small or from small to large, and so on.

[0071] Step S2232, reading the highest similarity score in the similarity sorting result to each target similarity score in the preset sequence position in order from the largest similarity score to the smallest similarity score.

[0072] Step S2233, calculate the average score of each target similarity score and determine it as the abstract feature of the problem.

[0073] The preset sequence position can be set in advance according to actual conditions; the preset sequence position can include one or more; in a specific embodiment, the preset sequence position includes the 1st (i.e., the highest score), the 3rd, the 10th, and the 100th from large to small. In this embodiment, reading the highest similarity score in the similarity sorting result to each target similarity score in the preset sequence position in sequence includes: reading the similarity scores of the 1st (highest score), 1-3 (top 3 high scores), 1-10 (top 10 high scores), and 1-100 (top 100 high scores) from large to small. It can be understood that in other embodiments, the similarity scores corresponding to other sequence positions can also be selected as the target similarity scores to calculate the average score as the abstract feature of the problem. Further, the average score is calculated for the similarity scores from the highest score to the third place from large to small, that is, the first 3 similarity scores from large to small; the average score is calculated for the similarity scores from the highest score to the tenth place from large to small, that is, the first 10 similarity scores from large to small; ...; and so on. In this embodiment, 4 scores are calculated: the highest score, the average score of the top 3 from large to small, the average score of the top 10 from large to small, and the average score of the top 100 from large to small, as the abstract features of the problem.

[0074] Furthermore, if the preset similarity algorithm includes multiple algorithms, when determining the abstract feature, the target similarity score is read from the similarity ranking results corresponding to each preset similarity algorithm, and the average score is calculated, and the average scores corresponding to each preset similarity algorithm are used as the abstract feature of the problem. [AC6]

[0075] Taking the preset similarity scores including BM25 and TFIDF as examples, firstly, the similarity scores are calculated based on the BM25 calculation question and the existing data in the preset question-answering knowledge base, and the similarity scores of the highest score, the top 3 average scores, the top 10 average scores and the top 100 average scores are taken from the highest to the lowest, and are represented by BM25_top1, BM25_average_top3, BM25_average_top10 and BM25_average_top100 respectively; similarly, TFIDF_top1, TFIDF_top3, TFIDF_top10 and TFIDF_top100 are calculated based on TFIDF. TFIDF_average_top3, TFIDF_average_top10 and TFIDF_average_top100; in this embodiment, the abstract features of the problem include: BM25_top1, BM25_average_top3, BM25_average_top10, BM25_average_top100, TFIDF_top1, TFIDF_average_top3, TFIDF_average_top10 and TFIDF_average_top100.

[0076] Furthermore, in one embodiment, if Figure 5 As shown, based on the abstract features, matching is performed with existing data in a preset question-and-answer knowledge base to filter out unanswered questions that have no corresponding answers in the preset question-and-answer knowledge base, including steps S231 to S233.

[0077] Step S231, performing binary classification based on the abstract features to obtain a binary classification value of the problem.

[0078] Among them, binary classification is one of the classification algorithms, and the classification algorithm is one of the tasks of machine learning; the classification algorithm is a supervised algorithm that requires labeled data to train. In this embodiment, the abstract features of the question are used for binary classification to determine whether there is a matching answer to the acquired question in the preset question and answer knowledge base based on the obtained binary classification value. Among them, binary classification based on abstract features to obtain binary classification values ​​can be achieved in any way, such as sigmoid function (an activation function) or softmax function (also known as normalized exponential function).

[0079] In a specific embodiment, the binary classification algorithm uses a sigmoid function. Taking the preset similarity scores including BM25 and TFIDF as examples, the binary classification value obtained by binary classification based on the abstract features can be expressed as:

[0080] y=sigmoid([BM25_top1,BM25_average_top3,BM25_average_top10,

[0081] BM25_average_top100,TFIDF_top1,TFIDF_average_top3,

[0082] TFIDF_average_top1,TFIDF_average_top100]).

[0083] Step S232, determining whether there is a corresponding answer to the question in the preset question-and-answer knowledge base based on the binary classification value.

[0084] Based on the abstract features of the question, binary classification is performed, and the obtained binary classification value can determine whether there is a matching answer to the question in the preset question and answer knowledge base.

[0085] Step S233, determining questions that do not have corresponding answers in the preset question and answer knowledge base as questions to be answered.

[0086] It is determined based on the binary classification value whether each question has a matching answer in the preset question and answer knowledge base, and then questions that do not have a matching answer in the preset question and answer knowledge base can be screened out, that is, the questions to be answered in this embodiment.

[0087] In the above embodiment, the similarity scores of some of the similarity scores that meet the screening conditions are calculated by calculating the similarity scores between the question and the existing data in the preset question and answer knowledge base as the abstract features of the question, rather than the features of the question itself (such as vocabulary, syntax, etc.), and then the abstract features are used to perform binary classification operations, so that the labeled data used in the binary classification algorithm can be applied to the update of question and answer knowledge bases in different fields, achieving one-time labeling and multiple uses; since the characteristics of the question itself are related to the question, the problems involved in different fields are usually quite different, and the data contained in the question and answer knowledge bases corresponding to different fields are also different. Therefore, if the characteristics of the question itself are used for binary classification, the labeled data used for binary classification cannot be used in question and answer knowledge bases in different fields, resulting in high labeling costs. In this embodiment, the abstract features of the question are used for binary classification to determine whether there is a matching answer to the question in the preset question and answer knowledge base. The binary classification labeled data used in this method can be reused in different fields, reducing the labeling cost.

[0088] In one embodiment, Figure 6 As shown, obtaining the answer corresponding to the question to be answered, and updating the answer corresponding to the question to be answered to the preset question and answer knowledge base, also includes:

[0089] Step S610: divide the questions to be answered into different sets of questions to be answered, and the questions to be answered contained in each set of questions to be answered belong to the same category.

[0090] The division of the questions to be answered into different categories can be achieved in any way. In one embodiment, by calculating the similarity between the questions to be answered, the questions to be answered with greater similarity are divided into the same set.

[0091] Furthermore, in one embodiment, the questions to be answered are divided into different sets of questions to be answered, and the questions to be answered contained in each set of questions to be answered belong to the same category, including: initializing a question to be answered as an initial set of questions to be answered; calculating the set similarity between each pair of initial sets of questions to be answered, and merging the initial sets of questions to be answered whose set similarities meet the merging condition to obtain a set of questions to be answered.

[0092] In this embodiment, when dividing the questions to be answered, each question to be answered is first regarded as a category, that is, the initial set of questions to be answered, and then the set similarities between the initial sets of questions to be answered are calculated in turn, and the sets are merged based on the set similarities, and the sets whose set similarities meet the merging conditions are merged. In one embodiment, calculating the set similarity between the sets of questions to be answered includes: calculating the similarity between the questions to be answered in the set of questions to be answered; further, in a specific embodiment, when judging whether the merging condition is met, the minimum value of the similarity and the average value of the similarity are taken to represent the similarity between the two sets. Among them, the merging condition can be set according to the actual situation; in a specific embodiment, the merging condition includes that the minimum value in the set similarity is within the minimum value threshold range, and the average value in the set similarity is within the average value threshold range. Among them, the minimum value threshold range and the average value threshold range can be set according to the actual situation.

[0093] It can be understood that the unanswered questions contained in the question sets with a large set similarity are likely to be the same or similar. After merging these unanswered questions into the same unanswered question set, it is only necessary to obtain the answers corresponding to the unanswered question set, which reduces the amount of calculation and workload, and can reduce the calculation cost and time cost.

[0094] Step S620, deleting the set of questions to be answered that includes the number of questions to be answered that is less than a preset number threshold, to obtain a final set of questions to be answered.

[0095] After the process of dividing the unanswered questions belonging to the same category into the same divided set of unanswered questions is completed, the set of unanswered questions containing a small number of unanswered questions is deleted; further, in one embodiment, before dividing the set, each unanswered question is initialized as an initial set of unanswered questions, and the set similarity of the initial set of unanswered questions is calculated in turn, and after the sets that meet the merging conditions based on the set similarity are merged, the set of unanswered questions obtained after the merging is deleted, that is, the set with the number of unanswered questions in the set less than the preset number threshold is deleted to obtain the final set of unanswered questions. In this embodiment, the set of unanswered questions with the number of questions less than the preset number threshold is deleted, and the set retained after the deletion is recorded as the final set of unanswered questions. Among them, the preset number threshold can be set according to the actual situation, for example, it is set to be less than 3 questions or 5 questions in the set, that is, it is determined that the condition is met and the set of unanswered questions needs to be deleted; further, the preset number threshold can be set in combination with the number of questions obtained.

[0096] After the questions to be answered are divided into sets based on categories, the obtained sets may contain a small number of questions, which may be questions that are not often asked by users or may be irrelevant questions. For example, a user may ask an intelligent customer service in a biology-related field, such as "What's the weather like today?". Such questions are not strongly relevant to the intelligent customer service and do not need to be included in the preset question and answer knowledge base. Therefore, in this embodiment, the sets containing such questions are deleted to reduce unnecessary workload.

[0097] Furthermore, in one embodiment, before deleting the set of unanswered questions containing less than a preset number threshold, the set of unanswered questions containing less than a preset number threshold can also be screened out, recorded as the target set of unanswered questions, and the target set of unanswered questions is sent to the relevant operation and maintenance personnel to obtain the feedback information of the operation and maintenance personnel on the target set of unanswered questions; if it is determined according to the feedback information that the target set of unanswered questions needs to be deleted, the step of deleting the set of unanswered questions containing less than a preset number threshold is entered. In this embodiment, whether the set of unanswered questions containing a small number of questions needs to be deleted is manually confirmed twice to reduce misjudgment, so that the update process of the preset question and answer knowledge base can update as many relevant questions that users may ask as possible.

[0098] Further, in this embodiment, obtaining answers corresponding to the questions to be answered, and updating the answers corresponding to the questions to be answered to a preset question and answer knowledge base, includes step S630: obtaining answers corresponding to a set of questions to be answered, and updating the answers corresponding to the set of questions to be answered to a preset question and answer knowledge base.

[0099] In the above embodiment, before obtaining the answers corresponding to the questions to be answered, the questions to be answered are first divided into sets based on categories, that is, clustered, and the sets obtained by the division that contain a smaller number of questions are deleted. Then, the answers to the retained set of questions to be answered are obtained. Only the answers corresponding to the remaining set of questions to be answered need to be obtained. Compared with the method of obtaining the answer to a single question to be answered, the amount of calculation and workload are reduced, and the calculation cost and time cost can be reduced.

[0100] In a specific embodiment, the above-mentioned question-answer knowledge base updating method is applied to the question-answer knowledge base of the intelligent customer service, such as Figure 7 The flowchart of this embodiment is shown, which includes the following steps:

[0101] 1. Obtain the chat log between the user and the intelligent customer service, that is, the questions raised by the user.

[0102] 2. Based on the binary classification algorithm, the unanswered questions that the intelligent customer service cannot answer are cleaned out, that is, questions that do not have matching answers in the preset question and answer database.

[0103] Among them, the binary classification algorithm is used to separate the questions that the intelligent customer service cannot answer from the chat logs between the user and the intelligent customer service. The binary classification algorithm is a mathematical expression as follows: y = f(x);

[0104] Among them, x: represents the user's question, and y is 1 or 0, indicating whether it can be answered.

[0105] Classification algorithms are supervised algorithms that require labeled data for training. There are two problems that need to be solved:

[0106] (1) The labeling cost is high. Different intelligent customer service systems contain different knowledge bases. For the same user question, some intelligent customer service systems can answer it while others cannot. How can different intelligent customer service systems reuse the same labeling data? (2) For the same intelligent customer service system, as the knowledge base becomes richer, questions that were previously unanswerable can become answerable questions.

[0107] In order to solve the above problems, abstract features are used to replace the features of the user questions themselves in the feature selection during binary classification, that is, y=f(user question) is changed to y=f(abstract features of user questions).

[0108] Furthermore, the abstract problem of the question is determined, and a preset similarity algorithm is used to search the question-answer knowledge base based on the user question. For example, BM25, TFIDF, a word vector-based similarity algorithm, a sentence vector-based similarity algorithm, etc. can be selected.

[0109] Read the preset similarity algorithm. For each similarity algorithm, select the highest score, top 3 average scores, top 10 average scores, top 100 average scores and other numerical features obtained when comparing the user question with the question-answer knowledge base from large to small as the abstract features of the user question. The binary classification algorithm uses sigmoid (other binary classification algorithms can also be used in other embodiments). The binary classification algorithm model is expressed as follows:

[0110] y=sigmoid( [

[0112] BM25_top1,

[0113] BM25_average_top3,

[0114] BM25_average_top10,

[0115] BM25_average_top100,

[0116] TFIDF_top1,

[0117] TFIDF_average_top3 ...... ] )

[0120] Since the question itself is not used as the classification feature, but the similarity score between the user question and the question-answering knowledge base is used as the abstract feature, and this similarity score feature shows strong consistency in different question-answering knowledge bases, it can be labeled once and used multiple times.

[0121] 3. Cluster the questions to be answered, obtain the clustering results, and delete the clusters with fewer questions to be answered. The specific steps are as follows: Figure 8 As shown, each question to be answered is initialized as an initial set of questions to be answered, and the set similarities between each pair of initial sets of questions to be answered are calculated respectively. The sets whose set similarities meet the merging conditions are merged together to obtain a merged set of questions to be answered, and the sets whose number of questions to be answered is less than a preset threshold are deleted to obtain the final clustering result, that is, the final set of questions to be answered.

[0122] Figure 8As shown in , clusters.length indicates the number of question sets to be answered; "Calculate the similarity between cluster i and cluster j" uses the minimum and average similarities between the two question sets to be answered; "Similarity is within the threshold range" means that the minimum and average similarities between the two sentences calculated in the previous step must be within a certain threshold, which can be determined according to the specific situation; "Delete clusters with less than a certain value of elements in clusters": After clustering, some categories contain very few user questions. The questions contained in this category are considered to be questions that are not frequently asked and need to be removed from the clustering results.

[0123] 4. Obtain answers corresponding to the final set of questions to be answered, and update the answers to the preset question and answer knowledge base. In this embodiment, the final set of questions to be answered is sent to the operation and maintenance personnel, who make corresponding answers to the set of questions to be answered and update the answers to the preset question and answer knowledge base.

[0124] The clustering algorithm aims to cluster unanswerable questions into multiple categories, so that questions with the same meaning are grouped together to facilitate secondary confirmation by operation and maintenance personnel. The above clustering algorithm can not only group together questions with the same meaning that users often ask, but also eliminate sparse questions that users rarely ask, because there is no need to enter seldom asked questions into the knowledge base.

[0125] Since the "binary classification algorithm" and "clustering algorithm" cannot guarantee 100% accuracy of the results, secondary confirmation can ensure the quality of the intelligent customer service knowledge base with a small amount of human participation; the result of clustering is a collection of user questions, and operation and maintenance personnel need to combine their professional knowledge to add answers to these frequently asked questions and enter them into the knowledge base.

[0126] The above-mentioned question-answer knowledge base update method provides a systematic solution to semi-automatically enrich the intelligent customer service "question-answer" pairs; it solves the problem of intelligent customer service knowledge base maintenance, and its effectiveness has been verified in specific scenarios. It provides a binary classification algorithm that can be migrated between different knowledge bases. By using similarity abstract features, it realizes one-time labeling and multiple uses; it solves the problem of high labeling costs; and as the number of "question-answer" pairs in the knowledge base increases, some questions change from unanswerable to answerable. Finally, it provides a question clustering method that can eliminate sparse questions, which can not only group together frequently asked questions with the same meaning, but also eliminate relatively sparse and infrequently asked questions.

[0127] It should be understood that, although each step in each flow chart involved in the above-described embodiment is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have strict order restrictions, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each flow chart involved in the above-described embodiment may include multiple steps or multiple stages, and these steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of the steps or stages in other steps or other steps.

[0128] In one embodiment, Fig. 9 As shown, a question-answer knowledge base updating device is provided, comprising: a question acquisition module 910, an abstraction processing module 920, a matching module 930 and an updating module 940, wherein:

[0129] A question acquisition module 910 is used to acquire questions;

[0130] Abstraction processing module 920, used to perform feature abstraction processing on the problem to obtain the abstract features of the problem;

[0131] A matching module 930 is used to match the existing data in the preset question-and-answer knowledge base based on the abstract features, and filter out questions to be answered that have no corresponding answers in the preset question-and-answer knowledge base;

[0132] The updating module 940 is used to obtain the answers corresponding to the questions to be answered, and update the answers corresponding to the questions to be answered to the preset question and answer knowledge base.

[0133] After obtaining the question raised by the user, the above-mentioned question and answer knowledge base updating device performs abstract processing on the question to obtain the abstract features corresponding to the question; based on the obtained abstract features of the question, the abstract features of the question are matched with the existing data in the preset question and answer knowledge base, thereby screening out the unanswered questions in the questions raised by the user that have no matching answers in the preset question and answer knowledge base; obtain the answers to the unanswered questions, update the answers to the preset question and answer knowledge base, and complete the update of the preset question and answer knowledge base. After extracting the abstract features of the question, the above-mentioned device uses the abstract features of the question to match with the preset question and answer knowledge base, screens out the unanswered questions that have no matching answers in the preset question and answer knowledge base, and after obtaining the answers to the unanswered questions, updates them to the preset question and answer knowledge base. Since the questions that cannot be answered by the preset question and answer knowledge base have been screened out by using the abstract features, it is only necessary to obtain the corresponding answers to the screened unanswered questions for updating, which reduces the workload of updating the preset question and answer knowledge base, shortens the time required for the update process, and reduces the manpower cost and time cost.

[0134] In one embodiment, the abstract processing module 920 of the above-mentioned device includes: an algorithm reading unit, used to read a preset similarity algorithm; a similarity calculation unit, used to calculate the similarity score for the question and the existing data in the preset question and answer knowledge base based on each preset similarity algorithm; the abstract feature determination unit is also used to determine the abstract feature of the question based on the similarity score corresponding to each preset similarity algorithm of the question.

[0135] Furthermore, in one embodiment, the above-mentioned abstract feature determination unit includes: a sorting subunit, which is used to sort the similarity scores of the question and the existing data in the preset question and answer knowledge base according to the score values ​​for any preset similarity algorithm to obtain a similarity sorting result; a score reading subunit, which is used to read the highest similarity score in the similarity sorting result to each target similarity score in the preset sequence position in order from large to small; an average value calculation subunit, which is used to calculate the average score of each target similarity score to determine it as the abstract feature of the question.

[0136] In one embodiment, the matching module 930 of the above-mentioned device includes: a binary classification unit, which is used to perform binary classification based on abstract features to obtain a binary classification value of the question; a matching unit, which is used to determine whether the question has a corresponding answer in a preset question and answer knowledge base based on the binary classification value; and a to-be-answered question determination unit, which is used to determine a question that does not have a corresponding answer in the preset question and answer knowledge base as a to-be-answered question.

[0137] In one embodiment, the above-mentioned device also includes: a classification module, which is used to divide the questions to be answered into different sets of questions to be answered, and the questions to be answered contained in each set of questions to be answered belong to the same category; a deletion module, which is used to delete the set of questions to be answered in which the number of questions to be answered is less than a preset number threshold, so as to obtain the final set of questions to be answered; in this embodiment, the update module 940 is used to obtain the answers corresponding to the set of questions to be answered, and update the answers corresponding to the set of questions to be answered to the preset question and answer knowledge base.

[0138] In one embodiment, the classification module of the above-mentioned device includes: an initialization unit, used to initialize a question to be answered into an initial set of questions to be answered; a similarity calculation module, used to respectively calculate the set similarity between each pair of initial sets of questions to be answered, and merge the initial sets of questions to be answered whose set similarities meet the merging conditions to obtain a set of questions to be answered.

[0139] For the specific limitations of the question-and-answer knowledge base updating device, please refer to the limitations of the question-and-answer knowledge base updating method above, which will not be repeated here. Each module in the above-mentioned question-and-answer knowledge base updating device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0140] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig.10 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for updating a question and answer knowledge base is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0141] Those skilled in the art will understand that Fig.10The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0142] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0143] Obtain questions; perform feature abstraction processing on the questions to obtain abstract features of the questions; match the abstract features with existing data in a preset question-and-answer knowledge base to filter out unanswered questions that have no corresponding answers in the preset question-and-answer knowledge base; obtain answers corresponding to the unanswered questions, and update the answers corresponding to the unanswered questions to the preset question-and-answer knowledge base.

[0144] In one embodiment, when the processor executes the computer program, the following steps are also implemented: reading a preset similarity algorithm; calculating similarity scores for the question and existing data in a preset question and answer knowledge base based on each preset similarity algorithm; and determining the abstract features of the question based on the similarity scores corresponding to each preset similarity algorithm of the question.

[0145] In one embodiment, when the processor executes the computer program, the following steps are also implemented: for any preset similarity algorithm, the similarity scores between the question and the existing data in the preset question and answer knowledge base are sorted according to the score values ​​to obtain a similarity sorting result; in order from large to small, the highest similarity score in the similarity sorting result is read in sequence to each target similarity score in the preset sequence position; the average score of each target similarity score is calculated to determine it as the abstract feature of the question.

[0146] In one embodiment, when the processor executes the computer program, the following steps are also implemented: binary classification is performed based on the abstract features to obtain the binary classification value of the question; whether the question has a corresponding answer in the preset question and answer knowledge base based on the binary classification value; and questions that do not have corresponding answers in the preset question and answer knowledge base are determined as questions to be answered.

[0147] In one embodiment, when the processor executes the computer program, the following steps are also implemented: dividing the questions to be answered into different sets of questions to be answered, and the questions to be answered contained in each set of questions to be answered belong to the same category; deleting the sets of questions to be answered that contain a number of questions to be answered that is less than a preset number threshold, and obtaining a final set of questions to be answered; obtaining answers corresponding to the set of questions to be answered, and updating the answers corresponding to the set of questions to be answered to a preset question and answer knowledge base.

[0148] In one embodiment, when the processor executes the computer program, the following steps are also implemented: initializing a question to be answered as an initial set of questions to be answered; calculating the set similarities between each pair of initial sets of questions to be answered, and merging the initial sets of questions to be answered whose set similarities meet the merging conditions to obtain a set of questions to be answered.

[0149] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0150] Obtain questions; perform feature abstraction processing on the questions to obtain abstract features of the questions; match the abstract features with existing data in a preset question-and-answer knowledge base to filter out unanswered questions that have no corresponding answers in the preset question-and-answer knowledge base; obtain answers corresponding to the unanswered questions, and update the answers corresponding to the unanswered questions to the preset question-and-answer knowledge base.

[0151] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: reading a preset similarity algorithm; calculating similarity scores for the question and existing data in a preset question and answer knowledge base based on each preset similarity algorithm; and determining the abstract features of the question based on the similarity scores corresponding to each preset similarity algorithm of the question.

[0152] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: for any preset similarity algorithm, the similarity scores between the question and the existing data in the preset question and answer knowledge base are sorted according to the score values ​​to obtain a similarity sorting result; in order from large to small, the highest similarity score in the similarity sorting result is read in sequence to each target similarity score in the preset sequence position; the average score of each target similarity score is calculated to determine it as the abstract feature of the question.

[0153] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: binary classification is performed based on the abstract features to obtain the binary classification value of the question; whether the question has a corresponding answer in the preset question and answer knowledge base based on the binary classification value; and questions that do not have corresponding answers in the preset question and answer knowledge base are determined as questions to be answered.

[0154] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: dividing the questions to be answered into different sets of questions to be answered, and the questions to be answered contained in each set of questions to be answered belong to the same category; deleting the sets of questions to be answered that contain a number of questions to be answered that is less than a preset number threshold, and obtaining a final set of questions to be answered; obtaining answers corresponding to the set of questions to be answered, and updating the answers corresponding to the set of questions to be answered to a preset question and answer knowledge base.

[0155] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: initializing a question to be answered as an initial set of questions to be answered; calculating the set similarities between each pair of initial sets of questions to be answered, and merging the initial sets of questions to be answered whose set similarities meet the merging conditions to obtain a set of questions to be answered.

[0156] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0157] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0158] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A question-answer knowledge base updating method, characterized in that: The method comprises: Get the question; Performing feature abstraction processing on the problem to obtain abstract features of the problem; Based on the abstract features, the existing data in the preset question-and-answer knowledge base is matched to filter out questions to be answered that have no corresponding answers in the preset question-and-answer knowledge base; Obtaining answers corresponding to the questions to be answered, and updating the answers corresponding to the questions to be answered to the preset question and answer knowledge base; The performing feature abstraction processing on the question to obtain the abstract features of the question includes: calculating similarity scores between the question and existing data in the preset question-answer knowledge base based on each preset similarity algorithm; determining the abstract features of the question based on the similarity scores corresponding to each preset similarity algorithm of the question; The method matches the abstract features with the existing data in the preset question and answer knowledge base to screen out unanswered questions that have no corresponding answers in the preset question and answer knowledge base, including: performing binary classification based on the abstract features to obtain the binary classification value of the question; determining whether the question has a corresponding answer in the preset question and answer knowledge base according to the binary classification value; and determining the questions that do not have corresponding answers in the preset question and answer knowledge base as the unanswered questions.

2. The method according to claim 1, characterized in that Determining the abstract features of the problem based on the similarity scores corresponding to the preset similarity algorithms of the problem includes: For any preset similarity algorithm, the similarity scores of the question and the existing data in the preset question-and-answer knowledge base are sorted according to the score values ​​to obtain a similarity sorting result; Reading the target similarity scores from the highest similarity score in the similarity sorting result to the preset sequence position in order from the largest to the smallest score; The average score of each target similarity score is calculated and determined as the abstract feature of the problem.

3. The method according to claim 2, characterized in that If the preset similarity algorithm includes multiple algorithms, the target similarity score is read from the similarity ranking result corresponding to each preset similarity algorithm, the average score is calculated, and the average scores corresponding to each preset similarity algorithm are used as the abstract features of the problem.

4. The method according to any one of claims 1 to 3, characterized in that: The step of obtaining the answer corresponding to the question to be answered and updating the answer corresponding to the question to be answered to the preset question and answer knowledge base further includes: Dividing the questions to be answered into different sets of questions to be answered, wherein the questions to be answered contained in each set of questions to be answered belong to the same category; Deleting the set of questions to be answered that includes questions with a number less than a preset number threshold, to obtain a final set of questions to be answered; The obtaining of answers corresponding to the questions to be answered and updating the answers corresponding to the questions to be answered to the preset question and answer knowledge base includes: obtaining answers corresponding to the set of questions to be answered and updating the answers corresponding to the set of questions to be answered to the preset question and answer knowledge base.

5. The method according to claim 4, characterized in that The dividing the questions to be answered into different sets of questions to be answered, wherein the questions to be answered contained in each set of questions to be answered belong to the same category, includes: Initialize a question to be answered as an initial set of questions to be answered; The set similarities between the two initial sets of questions to be answered are calculated respectively, and the initial sets of questions to be answered whose set similarities satisfy a merging condition are merged to obtain the set of questions to be answered.

6. The method according to claim 4, characterized in that The method further comprises: Before deleting the set of unanswered questions including the number of unanswered questions less than the preset number threshold, screen out the set of unanswered questions including the number of unanswered questions less than the preset number threshold, and record it as the target set of unanswered questions; Send the target set of questions to be answered to relevant operation and maintenance personnel, and obtain feedback information from the operation and maintenance personnel on the target set of questions to be answered; If it is determined according to the feedback information that the target set of questions to be answered needs to be deleted, the step of deleting the set of questions to be answered including the number of questions to be answered that is less than a preset number threshold is entered.

7. A question-answer knowledge base updating device, characterized in that: The device comprises: A question acquisition module is used to acquire questions; An abstract processing module is used to perform feature abstract processing on the problem to obtain the abstract features of the problem; A matching module, used to match the existing data in a preset question-and-answer knowledge base based on the abstract features, and filter out questions to be answered that have no corresponding answers in the preset question-and-answer knowledge base; An updating module, used for obtaining answers corresponding to the questions to be answered, and updating the answers corresponding to the questions to be answered to the preset question and answer knowledge base; The abstract processing module is used to calculate similarity scores between the question and the existing data in the preset question-answer knowledge base based on each preset similarity algorithm; based on the similarity scores corresponding to each preset similarity algorithm of the question, determine the abstract features of the question; The matching module includes: a binary classification unit, used to perform binary classification based on abstract features to obtain a binary classification value of the question; a matching unit, used to determine whether the question has a corresponding answer in a preset question and answer knowledge base based on the binary classification value; and a to-be-answered question determination unit, used to determine a question that does not have a corresponding answer in the preset question and answer knowledge base as a to-be-answered question.

8. The device according to claim 7, characterized in that The device also includes: A clustering module, used for dividing the questions to be answered into different sets of questions to be answered, wherein the questions to be answered contained in each set of questions to be answered belong to the same category; A deleting module, used to delete the set of questions to be answered that contains questions with a number less than a preset number threshold, to obtain a final set of questions to be answered; The updating module is also used to: obtain answers corresponding to the set of questions to be answered, and update the answers corresponding to the set of questions to be answered to the preset question and answer knowledge base.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images