Insurance service question and answer method and related device

Abstract

The invention discloses an insurance business question and answer method and a related device. The method comprises the steps of obtaining a to-be-answered question of a user; through a knowledge base association model, based on the to-be-answered question, obtaining a target knowledge base associated with the to-be-answered question; similarity calculation is carried out on the multiple pieces of knowledge in the target knowledge base and the to-be-answered question, and K pieces of candidate knowledge with the highest similarity with the to-be-answered question are determined; k > 0; and based on the K pieces of candidate knowledge and the to-be-answered question, obtaining an answer to the to-be-answered question. Therefore, the pre-constructed knowledge base is fully utilized, the semantics of the to-be-answered question of the user is fully mined in combination with the vector similarity, and the K pieces of candidate knowledge with strong semantic association can be determined for the to-be-answered question, so that a more accurate answer to the to-be-answered question can be obtained based on the K pieces of candidate knowledge, and the user experience is further improved.

Description

Technical Field

[0001] The present application relates to the technical field of natural language processing, and in particular to a question-answering method and related devices for insurance business. Background Art

[0002] With the rapid development of technology, question-answering is increasingly used in various fields. For example, in the insurance field, it can provide instant answers to questions related to insurance business raised by users, help users solve questions related to insurance business, and enable users to better understand insurance business.

[0003] In the related art, users can rely on predefined rules or templates to answer various questions about insurance business. However, predefined rules or templates are difficult to cover all questions that users may ask, lack flexibility, and cannot cope with the diverse needs of users, which easily affects the accuracy of the final generated answers.

[0004] Therefore, how to improve the accuracy of the answers generated for questions raised by users has become a problem that needs to be solved urgently. Summary of the invention

[0005] In view of this, an embodiment of the present application provides a question-and-answer method and related devices for insurance business, the purpose of which is to improve the accuracy of the answers generated for questions raised by users.

[0006] In a first aspect, an embodiment of the present application provides a question-and-answer method for insurance business, the method comprising:

[0007] Obtaining the user's unanswered questions; the unanswered questions are related to insurance business;

[0008] Based on the question to be answered, a target knowledge base associated with the question to be answered is obtained through a knowledge base association model; the knowledge base association model is obtained by training multiple knowledge bases, each of which stores multiple pieces of knowledge related to the insurance business;

[0009] Calculate the similarity between the plurality of pieces of knowledge in the target knowledge base and the question to be answered respectively, and determine K pieces of candidate knowledge with the highest similarity to the question to be answered; wherein K>0;

[0010] Based on the K pieces of candidate knowledge and the question to be answered, an answer to the question to be answered is obtained.

[0011] In a possible implementation, obtaining an answer to the question to be answered based on the K pieces of candidate knowledge and the question to be answered includes:

[0012] By using a semantic relevance determination model, based on the K pieces of candidate knowledge and the question to be answered, N pieces of semantically relevant knowledge semantically relevant to the question to be answered and reasons why the KN pieces of candidate knowledge are not semantically relevant to the question to be answered are obtained; the K pieces of candidate knowledge include the N pieces of semantically relevant knowledge and the KN pieces of candidate knowledge; 0≤N≤K;

[0013] If the number of the N pieces of semantically related knowledge is greater than or equal to a preset number, an answer to the question to be answered is obtained through a question-answering model based on the N pieces of semantically related knowledge and the question to be answered.

[0014] In a possible implementation, the insurance business question-and-answer method further includes:

[0015] If the number of the N semantically relevant knowledge pieces is less than the preset number, a new target knowledge base associated with the question to be answered is obtained based on the question to be answered and the semantically irrelevant reason through the knowledge base association model;

[0016] Calculate the similarity between the plurality of pieces of knowledge in the new target knowledge base and the question to be answered respectively, and determine K new candidate pieces of knowledge with the highest similarity to the question to be answered;

[0017] Through a semantic relevance determination model, based on the K new candidate knowledge and the question to be answered, M new semantically relevant knowledge semantically relevant to the question to be answered and reasons why the KM new candidate knowledge are not relevant to the new semantics of the question to be answered are obtained; the K new candidate knowledge includes the M new semantically relevant knowledge and the KM new candidate knowledge; 0≤M≤K;

[0018] If the sum of the number of the N semantically related knowledge and the M new semantically related knowledge is greater than or equal to the preset number, then, through the question-answering model, based on the N semantically related knowledge, the M new semantically related knowledge and the question to be answered, an answer to the question to be answered is obtained;

[0019] If the sum of the N semantically-related knowledge and the M new semantically-related knowledge is less than the preset number, the latest semantically-related knowledge is determined based on the question to be answered and the reason why the new semantics are irrelevant until the number of semantically-related knowledge meets the preset number, thereby obtaining the answer to the question to be answered.

[0020] In a possible implementation, obtaining an answer to the question to be answered based on the K pieces of candidate knowledge and the question to be answered includes:

[0021] Through the question-answering model, based on the K pieces of candidate knowledge and the question to be answered, an answer to the question to be answered is obtained.

[0022] In a possible implementation, the knowledge base association model is trained in the following manner:

[0023] Obtaining a first sample of questions to be answered and a knowledge base associated with the first sample of questions to be answered;

[0024] Obtaining a prediction knowledge base associated with the first to-be-answered question sample based on the first to-be-answered question sample by using the first to-be-trained model;

[0025] The first model to be trained is trained according to the prediction knowledge base, the knowledge base associated with the first question sample to be answered, and the first loss function to obtain the knowledge base associated model.

[0026] In one possible implementation, the question-answering model is trained in the following manner:

[0027] Acquire a second question sample to be answered, a semantically relevant knowledge sample semantically relevant to the second question sample to be answered, and an answer sample to the second question sample to be answered;

[0028] Generate a predicted answer to the second question sample to be answered based on the second question sample to be answered and the semantically relevant knowledge sample by using a second model to be trained;

[0029] According to the predicted answer, the answer sample and the second loss function, the second model to be trained is trained to obtain the question-answering model.

[0030] In a possible implementation, the semantic relevance determination model is trained in the following manner:

[0031] Obtain a third question sample to be answered, k candidate knowledge samples corresponding to the third question sample to be answered, and n semantically relevant knowledge samples semantically relevant to the third question sample to be answered; the k candidate knowledge samples are k knowledge samples with the highest similarity to the third question sample to be answered in a knowledge base associated with the third question sample to be answered, where k>0; the n semantically relevant knowledge samples belong to the k candidate knowledge samples, where n≤k, and n>0;

[0032] Generate, by means of a third model to be trained, a predicted semantically relevant knowledge sample semantically relevant to the third question sample to be answered based on the third question sample to be answered and the k candidate knowledge samples;

[0033] According to the predicted semantically relevant knowledge sample, the n semantically relevant knowledge samples and the third loss function, the third model to be trained is trained to obtain the semantically relevant determination model.

[0034] In a second aspect, an embodiment of the present application provides a question-and-answer device for insurance business, the device comprising:

[0035] A question acquisition module is used to acquire the user's unanswered questions; the unanswered questions are related to insurance business;

[0036] A target knowledge base acquisition module is used to obtain a target knowledge base associated with the question to be answered based on the question to be answered through a knowledge base association model; the knowledge base association model is obtained by training multiple knowledge bases, each of which stores multiple pieces of knowledge related to the insurance business;

[0037] A similarity calculation module is used to calculate the similarity between the plurality of pieces of knowledge in the target knowledge base and the question to be answered, and determine K pieces of candidate knowledge with the highest similarity to the question to be answered; wherein K>0;

[0038] The answer acquisition module is used to obtain the answer to the question to be answered based on the K pieces of candidate knowledge and the question to be answered.

[0039] In a third aspect, an embodiment of the present application provides a question-and-answer device for insurance business, the device comprising a memory and a processor:

[0040] The memory is used to store a computer program and transmit the computer program to the processor;

[0041] The processor is used to execute the computer program so that the device executes the question-and-answer method for insurance business described in the first aspect.

[0042] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed, the device executing the computer program implements the question-and-answer method for insurance business described in the first aspect above.

[0043] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0044] The embodiment of the present application provides a question-answering method and related device for insurance business, in which the user's unanswered question is first obtained, and the unanswered question is related to the insurance business; then, through the knowledge base association model, based on the unanswered question, a target knowledge base associated with the unanswered question is obtained, wherein the knowledge base association model is obtained based on the training of multiple knowledge bases, and each knowledge base stores multiple pieces of knowledge related to the insurance business; then, the multiple pieces of knowledge in the target knowledge base are respectively calculated with the unanswered question for similarity, and the K pieces of candidate knowledge with the highest similarity to the unanswered question are determined, K>0; finally, based on the K pieces of candidate knowledge and the unanswered question, the answer to the unanswered question is obtained. In this way, the pre-constructed knowledge base is fully utilized, and the semantics of the user's unanswered question is fully mined in combination with the vector similarity, and K pieces of candidate knowledge with strong semantic association can be determined for the unanswered question. Therefore, based on these K pieces of candidate knowledge, a more accurate answer to the unanswered question can be obtained, which is conducive to further improving the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in this embodiment or 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 application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 A flowchart of a question-and-answer method for insurance business provided in an embodiment of the present application;

[0047] Figure 2 A schematic diagram of a question-and-answer method for insurance business provided in an embodiment of the present application;

[0048] Figure 3 A schematic diagram of the structure of a question-and-answer device for insurance business provided in an embodiment of the present application. DETAILED DESCRIPTION

[0049] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0050] At present, taking the insurance field as an example, the existing question-answering method is: questions related to insurance business raised by users will be answered based on predefined rules or templates. However, the questions raised by users may be diverse, and predefined rules or templates are difficult to cover all questions that users may ask. They are not flexible enough and it is difficult to meet user needs, which easily leads to low accuracy of the final generated answers.

[0051] Based on this, in order to solve the above problems, the embodiment of the present application provides a question-answering method and related devices for insurance business, in which the user's unanswered questions are first obtained, and the unanswered questions are related to the insurance business; then, through the knowledge base association model, based on the unanswered questions, a target knowledge base associated with the unanswered questions is obtained, wherein the knowledge base association model is obtained based on the training of multiple knowledge bases, and each knowledge base stores multiple pieces of knowledge related to the insurance business; then, the multiple pieces of knowledge in the target knowledge base are respectively calculated with the unanswered questions for similarity, and the K pieces of candidate knowledge with the highest similarity to the unanswered questions are determined, K>0; finally, based on the K pieces of candidate knowledge and the unanswered questions, the answers to the unanswered questions are obtained. In this way, the pre-constructed knowledge base is fully utilized, and the semantics of the user's unanswered questions are fully mined in combination with the vector similarity. K pieces of candidate knowledge with strong semantic association can be determined for the unanswered questions. Therefore, based on these K pieces of candidate knowledge, a more accurate answer to the unanswered questions can be obtained, which is conducive to further improving the user experience.

[0052] It should be noted that the question-and-answer method for insurance business provided in this application can be used in the field of natural language processing technology or artificial intelligence technology. The above is only an example and does not limit the application field of the question-and-answer method for insurance business provided in this application. In addition, the embodiment of the present application may not limit the executor of the question-and-answer method for insurance business. For example, the question-and-answer method for insurance business in the embodiment of the present application can be applied to data processing devices such as terminal devices or servers. Among them, the terminal device can be an electronic device such as a smart phone, a computer, or a tablet computer. The server can be an independent server, a cluster server, a cloud server, etc. This application does not specifically limit the terminal device or server mentioned above.

[0053] The specific implementation of the question-and-answer method and related devices for insurance business in the embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0054] See also Figure 1 , which is a flowchart of a question-and-answer method for insurance business provided in an embodiment of the present application, combined with Figure 1 As shown, it may specifically include:

[0055] S101: Obtain the user's questions to be answered.

[0056] Among them, the questions to be answered are related to insurance business.

[0057] In some embodiments, the insurance business may be auto insurance, life insurance, accident insurance or other insurance-related business, which is not limited in this application.

[0058] For example, the question to be answered may be what are the insurance requirements for insurance product A, what are the recommendations for health insurance, or what are the terms and conditions of insurance product B. This application does not limit this.

[0059] S102: obtaining a target knowledge base associated with the question to be answered based on the question to be answered through a knowledge base association model.

[0060] The knowledge base association model can be used to determine a target knowledge base associated with the question to be answered from multiple knowledge bases based on the question to be answered.

[0061] It should be noted that the present application does not limit the number of target knowledge bases, and the target knowledge base associated with the question to be answered can be one knowledge base or multiple knowledge bases.

[0062] The knowledge base association model is obtained by training based on multiple knowledge bases, and each knowledge base stores multiple pieces of knowledge related to the insurance business.

[0063] In one possible implementation, various knowledge related to the insurance business can be acquired in advance, such as the basic concepts, types, functions, insurance clauses, insurance procedures, claims procedures, after-sales services, and other knowledge of insurance products, insurance services, etc., and then these knowledge are processed for deduplication, redundancy, obsolescence, and invalidity to obtain multiple pieces of processed knowledge, and then a knowledge base can be constructed based on the processed multiple pieces of knowledge.

[0064] In some embodiments, multiple pieces of knowledge may be divided based on knowledge type to obtain knowledge belonging to different knowledge bases. For example, knowledge base A includes multiple pieces of knowledge related to the insurance process of various insurance products, and knowledge base B includes multiple pieces of knowledge related to insurance clauses, etc. This application does not limit this.

[0065] S103: Calculate similarity between the plurality of pieces of knowledge in the target knowledge base and the question to be answered respectively, and determine K pieces of candidate knowledge with the highest similarity to the question to be answered.

[0066] Wherein, K>0. For example, K can be 5, and accordingly, 5 candidate knowledge with the highest similarity to the question to be answered can be determined from multiple pieces of knowledge in the target knowledge base. K can also be 6 or 3, etc., which is not limited in this application.

[0067] In some embodiments, based on the above introduction, the target knowledge base may be one or more, and similarity calculations may be performed between all knowledge included in at least one target knowledge base and the question to be answered, and K candidate knowledge pieces with the highest similarity to the question to be answered may be determined from all the knowledge.

[0068] In a possible implementation, each piece of knowledge in each knowledge base can be converted into a vector for storage through a pre-trained embedded Embedding model. For questions to be answered, the Embedding model can be used to convert the questions to be answered into vectors, and then the vector similarity of the knowledge vector and the vector of the question to be answered is calculated, and then the similarity values ​​corresponding to multiple pieces of knowledge are sorted, and the K pieces of knowledge with the highest vector similarity are determined as the K candidate knowledge.

[0069] Among them, the Embedding model can convert the input unanswered questions or knowledge into semantic vectors of fixed dimensions, so that the semantically similar unanswered questions and knowledge in the vector space are closer when the vector similarity is calculated later.

[0070] Furthermore, in some embodiments, when the number of knowledge items included in the target knowledge base is less than K, the indication included in the target knowledge base may be directly determined as candidate knowledge.

[0071] S104: Based on the K pieces of candidate knowledge and the question to be answered, an answer to the question to be answered is obtained.

[0072] In a possible implementation of the present application, S104 may specifically include: through a semantic relevance determination model, based on the K pieces of candidate knowledge and the question to be answered, obtaining N pieces of semantically relevant knowledge semantically related to the question to be answered, and the reasons why the KN pieces of candidate knowledge are semantically irrelevant to the question to be answered; then if the number of the N pieces of semantically relevant knowledge is greater than or equal to a preset number, through a question-answering model, based on the N pieces of semantically relevant knowledge and the question to be answered, obtaining the answer to the question to be answered.

[0073] The K pieces of candidate knowledge include the N pieces of semantically related knowledge and the KN pieces of candidate knowledge; 0≤N≤K.

[0074] In some embodiments, K can be 5. Assuming that the model is determined through semantic relevance, based on 5 candidate knowledge and the question to be answered, 3 semantically relevant knowledge semantically relevant to the question to be answered are obtained, as well as the reasons why the remaining 2 candidate knowledge are semantically irrelevant to the question to be answered.

[0075] For example, K pieces of knowledge may all be semantically relevant to the question to be answered. In this case, the semantic relevance model does not need to output the reason for semantic irrelevance, that is, N=K.

[0076] As another example, the K pieces of knowledge may not be semantically relevant to the question to be answered. In this case, the semantic relevance model does not need to output semantically relevant knowledge, that is, N=0.

[0077] In some embodiments, the semantic irrelevance may be caused by the mismatch between the question to be answered and the knowledge type of the target knowledge base to which the candidate knowledge belongs.

[0078] Exemplarily, the target knowledge bases to which the K pieces of candidate knowledge belong include knowledge base A and knowledge base B, and the reason for the semantic irrelevance may be that the question to be answered does not match the knowledge type of knowledge base A.

[0079] In some embodiments, the preset number may be 3 or 5, etc., which is not limited in this application. For example, assuming that the preset number is 3 and K is 8, 5 semantically relevant knowledge semantically related to the question to be answered is obtained through the semantic relevance determination model, and then the answer to the question to be answered can be obtained through the question-answering model based on the 5 semantically relevant knowledge and the question to be answered.

[0080] In this way, from the K candidate knowledge, the semantic relevance determination model can further screen out knowledge with higher semantic relevance to the question to be answered, and the answer obtained based on this is more accurate.

[0081] The above describes the case where the number of N pieces of semantically related knowledge is greater than or equal to a preset number. The following describes the case where the number of N pieces of semantically related knowledge is less than a preset number.

[0082] In a possible implementation of the present application, the insurance business question-and-answer method may further include the following steps:

[0083] A1: If the number of the N semantically related knowledge is less than the preset number, a new target knowledge base associated with the question to be answered is obtained through the knowledge base association model based on the question to be answered and the semantically irrelevant reasons.

[0084] The unanswered question and semantically irrelevant reasons can be combined into a new unanswered question. Through the knowledge base association model, a new target knowledge base associated with the unanswered question can be obtained.

[0085] In an example where the semantic irrelevance is caused by the mismatch between the knowledge type of the question to be answered and the target knowledge base to which the candidate knowledge belongs, the knowledge base association model may filter out the knowledge base with the mismatched knowledge type when determining a new target knowledge base.

[0086] A2: Calculate the similarity between the plurality of pieces of knowledge in the new target knowledge base and the question to be answered respectively, and determine K new candidate pieces of knowledge having the highest similarity with the question to be answered.

[0087] A3: Through the semantic relevance determination model, based on the K new candidate knowledge and the question to be answered, M new semantically relevant knowledge semantically relevant to the question to be answered and the reasons why the KM new candidate knowledge are irrelevant to the new semantics of the question to be answered are obtained.

[0088] The K new candidate knowledge pieces include the M new semantically related knowledge pieces and the KM new candidate knowledge pieces; 0≤M≤K.

[0089] A4: If the sum of the N pieces of semantically-related knowledge and the M pieces of new semantically-related knowledge is greater than or equal to the preset number, then through the question-answering model, based on the N pieces of semantically-related knowledge, the M pieces of new semantically-related knowledge and the question to be answered, the answer to the question to be answered is obtained.

[0090] It should be noted that the implementation method of A1-A4 can refer to the introduction of the specific implementation method of S102-S104, which will not be repeated here.

[0091] A5: If the sum of the N semantically-related knowledge and the M new semantically-related knowledge is less than the preset number, the latest semantically-related knowledge is determined based on the question to be answered and the reason why the new semantics are irrelevant, until the number of semantically-related knowledge meets the preset number, and the answer to the question to be answered is obtained.

[0092] When the sum of the number of N semantically-related knowledge and the number of M new semantically-related knowledge is less than the preset number, A1-A3 can be repeatedly executed based on the question to be answered and the new semantically irrelevant reasons until the sum of the number of all semantically-related knowledge such as the obtained semantically-related knowledge and the new semantically-related knowledge is greater than or equal to the preset number. Then, the question-answering model can be used to obtain the answer to the question to be answered based on all the semantically-related knowledge and the question to be answered.

[0093] In a possible implementation of the present application, S104 may also include: obtaining an answer to the question to be answered based on the K pieces of candidate knowledge and the question to be answered through a question-answering model.

[0094] After obtaining K pieces of candidate knowledge, the K pieces of candidate knowledge and the question to be answered can also be directly used through the question-answering model to obtain the answer to the question to be answered. This application does not limit the implementation method of S104.

[0095] Next, we will continue to introduce the training methods of each model mentioned above.

[0096] In a possible implementation of the present application, the knowledge base association model can be trained in the following manner: first, obtain a first question sample to be answered and a knowledge base associated with the first question sample to be answered; then, through the first model to be trained, based on the first question sample to be answered, obtain a predicted knowledge base associated with the first question sample to be answered; then, according to the predicted knowledge base, the knowledge base associated with the first question sample to be answered and a first loss function, train the first model to be trained to obtain the knowledge base association model.

[0097] In some embodiments, the first sample of questions to be answered may be obtained from historical questions of the user, and the knowledge base associated with the first sample of questions to be answered may be selected by business personnel from multiple knowledge bases, which is not limited in this application.

[0098] In some embodiments, the first model to be trained can be a large model, or a BERT model, etc., which is not limited in this application.

[0099] The first to-be-answered question sample is input into the first to-be-trained model, and the first to-be-trained model can output a predicted prediction knowledge base associated with the to-be-answered question sample.

[0100] In some embodiments, the difference between the predicted knowledge base and the knowledge base associated with the first question sample to be answered is determined through a first loss function. When the preset training end condition is met, a trained knowledge base association model can be obtained. When the preset training end condition is not met, the above training process can continue to be iterated until the preset training end condition is met.

[0101] Exemplarily, the preset training end condition may be that the number of training times of the knowledge base association model reaches a preset number threshold; or the model performance of the knowledge base association model meets preset requirements, such as the difference between the predicted knowledge base and the knowledge base associated with the first sample of the question to be answered meets a preset difference condition, etc. This application does not limit this.

[0102] In a possible implementation of the present application, the question-answering model is trained in the following manner: obtaining a second question sample to be answered, a semantically-related knowledge sample semantically related to the second question sample to be answered, and an answer sample of the second question sample to be answered; generating a predicted answer to the second question sample to be answered based on the second question sample to be answered and the semantically-related knowledge sample through a second model to be trained; and training the second model to be trained according to the predicted answer, the answer sample, and a second loss function to obtain the question-answering model.

[0103] The training process of the question-answering model can be found in the training process of the knowledge base association model introduced above, and will not be repeated here.

[0104] It should be noted that the second question sample to be answered may be the same sample as the first question sample to be answered, or may be a different sample, and this application does not limit this.

[0105] In a possible implementation of the present application, the semantic relevance determination model is trained in the following manner: obtaining a third question sample to be answered, k candidate knowledge samples corresponding to the third question sample to be answered, and n semantically relevant knowledge samples semantically relevant to the third question sample to be answered; the k candidate knowledge samples are the k knowledge samples with the highest similarity to the third question sample to be answered in the knowledge base associated with the third question sample to be answered, and k>0; the n semantically relevant knowledge samples belong to the k candidate knowledge samples, n≤k, and n>0; through the third model to be trained, based on the third question sample to be answered and the k candidate knowledge samples, a predicted semantically relevant knowledge sample semantically relevant to the third question sample to be answered is generated; according to the predicted semantically relevant knowledge samples, the n semantically relevant knowledge samples and a third loss function, the third model to be trained to obtain the semantic relevance determination model.

[0106] The training process of the semantic relevance determination model can refer to the training process of the semantic relevance determination model introduced above, and will not be repeated here.

[0107] It should be noted that the third question sample to be answered may be the same sample as the first question sample to be answered, or may be a different sample, and this application does not impose any limitation on this.

[0108] As an example, see Figure 2 , which is a schematic diagram of a question-and-answer method for insurance business provided in an embodiment of the present application. Figure 2 As shown, business personnel can build multiple knowledge bases including knowledge base 1-knowledge base n, and then process each knowledge in the knowledge base through the Embedding model to obtain the vector corresponding to the knowledge, and obtain knowledge vector base 1-knowledge vector base n that corresponds one to one with knowledge base 1-knowledge base n.

[0109] like Figure 2 As shown, users can ask questions and obtain user questions (i.e., the questions to be answered as described above). User questions can be input into the Embedding model to obtain the vectors corresponding to the user questions. The knowledge vector library 6 and the knowledge vector library 8 (i.e., the target knowledge library described above) can be determined through the large model service (i.e., the trained knowledge base association model). Then, the vector similarity is calculated for each knowledge vector of the knowledge vector library 6 and the knowledge vector library 8 and the vector of the user question, and the K contexts with the highest similarity (i.e., the K candidate knowledge described above) are determined.

[0110] like Figure 2 As shown, through the large model service (that is, the trained semantic relevance determination model), it is possible to determine whether K contexts can answer user questions (that is, determining N semantically relevant knowledge as described above, and the reasons why the KN candidate knowledge are semantically irrelevant to the question to be answered), store the N semantically relevant knowledge in the candidate context record table, and store the reasons for semantic irrelevance in the reason table for not being able to answer.

[0111] like Figure 2 As shown, it can be determined whether the number of candidate contexts reaches a preset value (that is, whether the number of semantically relevant knowledge introduced above is greater than or equal to the preset number). If so, the subsequent steps can be continued, otherwise the previous steps can be repeated based on user questions and semantically irrelevant reasons.

[0112] When the number of candidate contexts reaches a preset value, the user's original question (ie, the user question) and the candidate context records can be merged, and the answer can be output through the large model service (ie, the trained question-answering model).

[0113] To sum up, the question-and-answer method for insurance business provided in the embodiment of the present application can fully explore the semantics of the questions to be answered based on the semantic analysis capabilities of models such as similarity comparison and semantic relevance determination models, better understand the questions to be answered, and help provide more accurate answers.

[0114] In addition, the embodiments of the present application can use the similarity comparison and various models to realize automatic semantic analysis and answering of the questions to be answered, which greatly improves the efficiency of question answering and the accuracy of the output answers. At the same time, it can extract key information (i.e. candidate knowledge) from the knowledge base including a large amount of knowledge, and then provide more accurate answers.

[0115] The above are some specific implementations of the insurance business question-answering method provided in the embodiment of the present application. Based on this, the present application also provides a corresponding insurance business question-answering device. The following will introduce the insurance business question-answering device provided in the embodiment of the present application from the perspective of functional modularization.

[0116] See also Figure 3 , which is a schematic diagram of the structure of a question-and-answer device for insurance business provided in an embodiment of the present application. The question-and-answer device 300 for insurance business may include:

[0117] The question acquisition module 310 is used to acquire the user's unanswered questions; the unanswered questions are related to insurance business;

[0118] A target knowledge base acquisition module 320 is used to obtain a target knowledge base associated with the question to be answered based on the question to be answered through a knowledge base association model; the knowledge base association model is obtained by training multiple knowledge bases, each of which stores multiple pieces of knowledge related to the insurance business;

[0119] A similarity calculation module 330 is used to calculate the similarity between the plurality of pieces of knowledge in the target knowledge base and the question to be answered, and determine K pieces of candidate knowledge with the highest similarity to the question to be answered; wherein K>0;

[0120] The answer acquisition module 340 is used to obtain the answer to the question to be answered based on the K pieces of candidate knowledge and the question to be answered.

[0121] As an implementation method, the answer acquisition module 340 may specifically include:

[0122] A first knowledge and reason acquisition unit is used to obtain N pieces of semantically relevant knowledge semantically relevant to the question to be answered and reasons why the KN pieces of candidate knowledge are not semantically relevant to the question to be answered based on the K pieces of candidate knowledge and the question to be answered through a semantic relevance determination model; the K pieces of candidate knowledge include the N pieces of semantically relevant knowledge and the KN pieces of candidate knowledge; 0≤N≤K;

[0123] The first answer acquisition unit is used to obtain the answer to the question to be answered based on the N pieces of semantically related knowledge and the question to be answered through a question-answering model if the number of the N pieces of semantically related knowledge is greater than or equal to a preset number.

[0124] As an implementation method, the insurance business question-answering device 300 may further include:

[0125] A first acquisition module is used for obtaining a new target knowledge base associated with the question to be answered based on the question to be answered and the semantic irrelevant reason through the knowledge base association model if the number of the N semantically related knowledge is less than the preset number;

[0126] A determination module is used to calculate the similarity between the plurality of pieces of knowledge in the new target knowledge base and the question to be answered, and determine K new candidate pieces of knowledge with the highest similarity to the question to be answered;

[0127] The second acquisition module is used to obtain M new semantically relevant knowledge semantically relevant to the question to be answered and reasons why the KM new candidate knowledge are not relevant to the new semantics of the question to be answered based on the K new candidate knowledge and the question to be answered through a semantic relevance determination model; the K new candidate knowledge includes the M new semantically relevant knowledge and the KM new candidate knowledge; 0≤M≤K;

[0128] A third acquisition module is used to obtain an answer to the question to be answered based on the N pieces of semantically related knowledge, the M pieces of new semantically related knowledge and the question to be answered through a question-answering model if the sum of the number of the N pieces of semantically related knowledge and the M pieces of new semantically related knowledge is greater than or equal to the preset number;

[0129] The fourth acquisition module is used to determine the latest semantically relevant knowledge based on the question to be answered and the reason why the new semantics are irrelevant if the sum of the N semantically relevant knowledge and the M new semantically relevant knowledge is less than the preset number, until the number of semantically relevant knowledge meets the preset number, thereby obtaining the answer to the question to be answered.

[0130] As an implementation manner, the answer acquisition module 340 may be specifically used for:

[0131] Through the question-answering model, based on the K pieces of candidate knowledge and the question to be answered, an answer to the question to be answered is obtained.

[0132] As an implementation method, the knowledge base association model is trained by the following units:

[0133] A first acquisition unit, configured to acquire a first sample of questions to be answered and a knowledge base associated with the first sample of questions to be answered;

[0134] A second acquisition unit, configured to obtain a prediction knowledge base associated with the first to-be-answered question sample based on the first to-be-answered question sample by using a first to-be-trained model;

[0135] The first training unit is used to train the first model to be trained to obtain the knowledge base association model according to the prediction knowledge base, the knowledge base associated with the first question sample to be answered, and a first loss function.

[0136] As an implementation, the question-answering model is trained by the following units:

[0137] A third acquisition unit is used to acquire a second question sample to be answered, a semantically relevant knowledge sample semantically relevant to the second question sample to be answered, and an answer sample to the second question sample to be answered;

[0138] A first generating unit, configured to generate a predicted answer to the second question sample to be answered based on the second question sample to be answered and the semantically related knowledge sample by using a second model to be trained;

[0139] The second training unit is used to train the second model to be trained to obtain the question-answering model according to the predicted answer, the answer sample and the second loss function.

[0140] As an implementation method, the semantic relevance determination model is trained by the following units:

[0141] a fourth acquisition unit, configured to acquire a third question sample to be answered, k candidate knowledge samples corresponding to the third question sample to be answered, and n semantically relevant knowledge samples semantically relevant to the third question sample to be answered; the k candidate knowledge samples are k knowledge samples with the highest similarity to the third question sample to be answered in a knowledge base associated with the third question sample to be answered, where k>0; the n semantically relevant knowledge samples belong to the k candidate knowledge samples, where n≤k, and n>0;

[0142] A second generating unit is used to generate a predicted semantically relevant knowledge sample semantically relevant to the third question sample to be answered based on the third question sample to be answered and the k candidate knowledge samples through a third to-be-trained model;

[0143] The third training unit is used to train the third model to be trained to obtain the semantic relevance determination model according to the predicted semantic relevance knowledge sample, the n semantic relevance knowledge samples and a third loss function.

[0144] The embodiments of the present application also provide a corresponding insurance business question-and-answer device and a computer-readable storage medium for implementing the solution provided in the embodiments of the present application.

[0145] Among them, the question and answer device for insurance business includes a memory and a processor, the memory is used to store a computer program, and the processor is used to execute the computer program so that the device executes the question and answer method for insurance business described in any embodiment of the present application.

[0146] The computer-readable storage medium stores a computer program. When the computer program is executed, the device executing the computer program implements the question-and-answer method for insurance business described in any embodiment of the present application.

[0147] The "first" and "second" in the names such as "first" and "second" (if any) mentioned in the embodiments of the present application are only used as name identifiers and do not represent the first or second in order.

[0148] Through the description of the above implementation methods, it can be known that those skilled in the art can clearly understand that all or part of the steps in the above-mentioned embodiment method can be implemented by means of software plus a general hardware platform. Based on such an understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a readable storage medium, such as a read-only memory (ROM) / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network communication device such as a router) to execute the methods described in each embodiment of the present application or some parts of the embodiments.

[0149] It should be noted that each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The device embodiment described above is merely schematic, in which the unit described as a separate component may or may not be physically separated, and the component prompted as a unit may or may not be a physical unit, that is, it may be located in one place, or it may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative work.

[0150] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A question-and-answer method for insurance business, characterized in that: include: Get the user's questions to be answered; The questions to be answered are related to insurance business; Based on the question to be answered, a target knowledge base associated with the question to be answered is obtained through a knowledge base association model; The knowledge base association model is obtained by training based on multiple knowledge bases, each of which stores multiple pieces of knowledge related to the insurance business; Calculate the similarity between the plurality of pieces of knowledge in the target knowledge base and the question to be answered respectively, and determine K pieces of candidate knowledge with the highest similarity to the question to be answered; wherein K>0; Based on the K pieces of candidate knowledge and the question to be answered, an answer to the question to be answered is obtained.

2. The method according to claim 1, characterized in that The step of obtaining an answer to the question to be answered based on the K pieces of candidate knowledge and the question to be answered includes: By using a semantic relevance determination model, based on the K pieces of candidate knowledge and the question to be answered, N pieces of semantically relevant knowledge semantically relevant to the question to be answered and reasons why the KN pieces of candidate knowledge are not semantically relevant to the question to be answered are obtained; the K pieces of candidate knowledge include the N pieces of semantically relevant knowledge and the KN pieces of candidate knowledge; 0≤N≤K; If the number of the N pieces of semantically related knowledge is greater than or equal to a preset number, an answer to the question to be answered is obtained through a question-answering model based on the N pieces of semantically related knowledge and the question to be answered.

3. The method according to claim 2, characterized in that Also includes: If the number of the N semantically relevant knowledge pieces is less than the preset number, a new target knowledge base associated with the question to be answered is obtained based on the question to be answered and the semantically irrelevant reason through the knowledge base association model; Calculate the similarity between the plurality of pieces of knowledge in the new target knowledge base and the question to be answered respectively, and determine K new candidate pieces of knowledge with the highest similarity to the question to be answered; Through a semantic relevance determination model, based on the K new candidate knowledge and the question to be answered, M new semantically relevant knowledge semantically relevant to the question to be answered and reasons why the KM new candidate knowledge are not relevant to the new semantics of the question to be answered are obtained; the K new candidate knowledge includes the M new semantically relevant knowledge and the KM new candidate knowledge; 0≤M≤K; If the sum of the number of the N semantically related knowledge and the M new semantically related knowledge is greater than or equal to the preset number, then, through the question-answering model, based on the N semantically related knowledge, the M new semantically related knowledge and the question to be answered, an answer to the question to be answered is obtained; If the sum of the N semantically-related knowledge and the M new semantically-related knowledge is less than the preset number, the latest semantically-related knowledge is determined based on the question to be answered and the reason why the new semantics are irrelevant until the number of semantically-related knowledge meets the preset number, thereby obtaining the answer to the question to be answered.

4. The method according to claim 1, characterized in that: The step of obtaining an answer to the question to be answered based on the K pieces of candidate knowledge and the question to be answered includes: Through the question-answering model, based on the K pieces of candidate knowledge and the question to be answered, an answer to the question to be answered is obtained.

5. The method according to any one of claims 1 to 4, characterized in that: The knowledge base association model is trained in the following way: Obtaining a first sample of questions to be answered and a knowledge base associated with the first sample of questions to be answered; Obtaining a prediction knowledge base associated with the first to-be-answered question sample based on the first to-be-answered question sample by using the first to-be-trained model; The first model to be trained is trained according to the prediction knowledge base, the knowledge base associated with the first question sample to be answered, and the first loss function to obtain the knowledge base associated model.

6. The method according to any one of claims 2 to 4, characterized in that: The question-answering model is trained in the following way: Acquire a second question sample to be answered, a semantically relevant knowledge sample semantically relevant to the second question sample to be answered, and an answer sample to the second question sample to be answered; Generate a predicted answer to the second question sample to be answered based on the second question sample to be answered and the semantically relevant knowledge sample by using a second model to be trained; According to the predicted answer, the answer sample and the second loss function, the second model to be trained is trained to obtain the question-answering model.

7. The method according to claim 2 or 3, characterized in that: The semantic relevance determination model is trained in the following manner: Obtain a third question sample to be answered, k candidate knowledge samples corresponding to the third question sample to be answered, and n semantically relevant knowledge samples semantically relevant to the third question sample to be answered; the k candidate knowledge samples are k knowledge samples with the highest similarity to the third question sample to be answered in a knowledge base associated with the third question sample to be answered, where k>0; the n semantically relevant knowledge samples belong to the k candidate knowledge samples, where n≤k, and n>0; Generate, by means of a third model to be trained, a predicted semantically relevant knowledge sample semantically relevant to the third question sample to be answered based on the third question sample to be answered and the k candidate knowledge samples; According to the predicted semantically relevant knowledge sample, the n semantically relevant knowledge samples and the third loss function, the third model to be trained is trained to obtain the semantically relevant determination model.

8. A question-and-answer device for insurance business, characterized in that: The device comprises: A question acquisition module is used to acquire the user's unanswered questions; the unanswered questions are related to insurance business; A target knowledge base acquisition module is used to obtain a target knowledge base associated with the question to be answered based on the question to be answered through a knowledge base association model; the knowledge base association model is obtained by training multiple knowledge bases, each of which stores multiple pieces of knowledge related to the insurance business; A similarity calculation module is used to calculate the similarity between the plurality of pieces of knowledge in the target knowledge base and the question to be answered, and determine K pieces of candidate knowledge with the highest similarity to the question to be answered; wherein K>0; The answer acquisition module is used to obtain the answer to the question to be answered based on the K pieces of candidate knowledge and the question to be answered.

9. A question-and-answer device for insurance business, characterized in that: The device comprises a memory and a processor: The memory is used to store a computer program and transmit the computer program to the processor; The processor is used to execute the computer program so that the device performs the steps of the question-and-answer method for insurance business as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program. When the computer program is executed, the device executing the computer program implements the steps of the question-and-answer method for insurance business as described in any one of claims 1 to 7.

Images