Information search method and device, equipment, storage medium and program product

Abstract

The invention provides an information search method and device, equipment and a storage medium. Relates to the financial science and technology field or other related fields. The method comprises the steps of obtaining an initial question input by a user; preprocessing the initial problem to obtain a plurality of problem keywords, a target problem and semantic vectors of the initial problem; according to the plurality of question keywords, the target question and the semantic vector, obtaining a plurality of answers corresponding to the initial question in a knowledge base through a plurality of recall modes; determining the correlation degree between each answer and the initial question, the information type of each answer and the recall mode of each answer; determining a plurality of answers to be selected in the plurality of answers according to the association degree between each answer and the initial question, the information type of each answer and the recall mode of each answer; the initial question and the to-be-selected answers are processed through the preset model, the target answer corresponding to the initial question is obtained, and the information searching efficiency is improved.

Description

Technical Field

[0001] The present application relates to the field of financial technology or other related fields, and in particular to an information search method, device, equipment, storage medium and program product. Background Art

[0002] In some scenarios, you need to search for relevant information, for example, you need to collect financial information about the user.

[0003] In the related art, information search is usually performed manually. For example, when it is necessary to query a user's financial related information, the staff needs to search in multiple databases or in paper documents. After the staff obtains multiple information through search, it is also necessary to perform multiple sorting operations such as deduplication on the searched information to obtain the user's financial related information. However, in the above process, since it is necessary to manually search multiple times in multiple places (databases, files, etc.) and manually sort the searched information, the efficiency of information search is low. Summary of the invention

[0004] The present application provides an information search method, apparatus, device, storage medium and program product to solve the problem of poor efficiency of information search.

[0005] In a first aspect, the present application provides an information search method, comprising:

[0006] Get the initial question from the user;

[0007] Preprocessing the initial question to obtain a plurality of question keywords, a target question, and a semantic vector of the initial question, wherein the target question is the question after error correction processing is performed on the initial question;

[0008] According to the multiple question keywords, the target question and the semantic vector, obtaining multiple answers corresponding to the initial question in the knowledge base through multiple recall methods;

[0009] Determine the relevance between each answer and the initial question, the information type of each answer, and the recall method of each answer, wherein the information type includes user information type, case information type, and processing result type;

[0010] Determining a plurality of answers to be selected from the plurality of answers according to the relevance between each answer and the initial question, the information type of each answer, and the recall method of each answer;

[0011] The initial question and the multiple answers to be selected are processed through a preset model to obtain a target answer corresponding to the initial question.

[0012] In a second aspect, the present application provides an information search device, including: a first acquisition module, a processing module, a second acquisition module and a determination module, wherein:

[0013] The first acquisition module is used to obtain an initial question input by a user;

[0014] The processing module is used to pre-process the initial question to obtain a plurality of question keywords, a target question and a semantic vector of the initial question, wherein the target question is the question after error correction processing is performed on the initial question;

[0015] The second acquisition module is used to acquire multiple answers corresponding to the initial question in the knowledge base through multiple recall methods according to the multiple question keywords, the target question and the semantic vector;

[0016] The determination module is used to determine the relevance between each answer and the initial question, the information type of each answer and the recall method of each answer, wherein the information type includes user information type, case information type and processing result type;

[0017] The determination module is further used to determine a plurality of answers to be selected from the plurality of answers according to the relevance between each answer and the initial question, the information type of each answer and the recall method of each answer;

[0018] The processing module is also used to process the initial question and the multiple answers to be selected through a preset model to obtain a target answer corresponding to the initial question.

[0019] In a third aspect, an embodiment of the present application provides an information search device, comprising: at least one processor and a memory; the memory stores computer execution instructions; at least one processor executes the computer execution instructions stored in the memory, so that the at least one processor executes the information search method involved in the first aspect and any item of the first aspect above.

[0020] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer execution instructions are stored. When a processor executes the computer execution instructions, the information search method involved in the first aspect and any item of the first aspect is implemented.

[0021] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the information search method involved in the first aspect and any item of the first aspect.

[0022] The information search method, device, equipment, storage medium and program product provided by the present application, when information search is needed, after obtaining the initial question input by the user, the initial question is first pre-processed to obtain multiple question keywords, target questions and semantic vectors of the initial question, and multiple answers are obtained in the knowledge base through recall. Then, based on the correlation between each answer obtained and the initial question, the information type of each answer and the recall method, multiple answers to be selected are determined, and the initial question and multiple answers to be selected are processed through a preset model to obtain the target answer. In the above process, relevant answers can be searched in the knowledge base through multiple recall methods, and the target answer can be quickly and accurately determined through the preset model, thereby improving the efficiency of information search. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0024] Figure 1 A schematic diagram of the system architecture provided for an embodiment of the present application;

[0025] Figure 2 A flowchart of an information search method provided in an embodiment of the present application;

[0026] Figure 3 A schematic diagram of a visual operation window provided in an embodiment of the present application;

[0027] Figure 4 A schematic diagram of a process for determining multiple answers to be selected provided in an embodiment of the present application;

[0028] Figure 5 A schematic diagram of another process of determining multiple answers to be selected provided in an embodiment of the present application;

[0029] Figure 6 A schematic diagram of a process for determining a target answer provided in an embodiment of the present application;

[0030] Figure 7 A schematic diagram of another process for determining a target answer provided in an embodiment of the present application;

[0031] Figure 8 A schematic diagram of a process for updating a knowledge base and a preset model provided in an embodiment of the present application;

[0032] Fig. 9 A schematic diagram of a visual operation window for user evaluation provided in an embodiment of the present application;

[0033] Fig.10 A flowchart of another information search method provided in an embodiment of the present application;

[0034] Fig.11 A schematic diagram of the structure of an information search device provided in an embodiment of the present application;

[0035] Fig.12 A schematic diagram of the structure of another information search device provided in an embodiment of the present application;

[0036] Fig.13 A schematic diagram of the structure of an information search device provided in an embodiment of the present application.

[0037] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0038] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0039] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with the relevant laws, regulations and standards of the relevant countries and regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0040] In addition, this application involves big data analysis of user information (including but not limited to personal biometrics, identity data, consumption data, asset data, electronic terminal operation data, etc.), and the use of artificial intelligence technology for automated decision-making, and a technical solution for making decisions that have a significant impact on personal rights and interests based on the results of automated decision-making. The application provides users with corresponding operation entrances for them to choose to agree or reject the results of automated decision-making; if the user chooses to reject, the expert decision-making process will be entered.

[0041] It should be noted that the information search methods, devices, equipment, storage media and products provided in this application can be used in the field of financial technology, and can also be used in any field other than financial technology. The application fields of the information search methods, devices, equipment, storage media and products in this application are not limited.

[0042] For ease of understanding, the following Figure 1 , the system architecture applicable to the embodiments of the present application is described.

[0043] Figure 1 This is a schematic diagram of the system architecture provided by the embodiment of this application. Figure 1 , including a terminal device 101 and an information search device 102.

[0044] The user can input questions in the terminal device 101 , and the terminal device 101 can send the questions input by the user to the information search device 102 . The information search device 102 can search for information on the questions sent by the terminal device 101 , and can also send the search results to the user's terminal device 101 .

[0045] The terminal device 101 may be an electronic device, for example, a mobile phone, a tablet computer, a laptop computer, etc. A target application may be set in the terminal device 101, and the user may input questions through the target application, for example, the target application may be a banking application, etc.

[0046] In the related art, information search is usually performed manually. For example, when it is necessary to query a user's financial related information, the staff needs to search in multiple databases or in paper documents. After the staff obtains multiple information through search, it is also necessary to perform multiple sorting operations such as deduplication on the searched information to obtain the user's financial related information. However, in the above process, since it is necessary to manually search multiple times in multiple places (databases, files, etc.) and manually sort the searched information, the efficiency of information search is low.

[0047] In response to the above technical problems, in an embodiment of the present application, when information search is required, after obtaining the user's initial question, the user's initial question is preprocessed, and based on the multiple keywords obtained by preprocessing, the target question and the semantic vector corresponding to the initial question, multiple answers corresponding to the initial question are obtained in the knowledge base through multiple recall methods, and multiple answers to be selected are determined based on the correlation between each answer and the initial question, the information type of each answer and the recall method, and then the initial question and the multiple answers to be selected are processed through a preset model to obtain the target answer. In the above process, relevant answers can be searched in the knowledge base through multiple recall methods, and the target answer can be quickly and accurately determined through the preset model, thereby improving the efficiency of information search.

[0048] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0049] Figure 2 A flowchart of an information search method provided in an embodiment of the present application. Figure 2 As shown, the method may include the following steps:

[0050] S201: Obtain an initial question input by a user.

[0051] The execution subject of the embodiment of the present application may be an information search device, or an information search apparatus arranged in the information search device. The information search apparatus may be implemented by software, or by a combination of software and hardware.

[0052] The initial question may be a question input by the user to the information search device. The user may input the initial question in the visual operation window, for example, the user may input "What is Zhang San's basic information?" in the visual operation window by manual input.

[0053] Next, combine Figure 3 , the visualization operation window is explained through specific examples.

[0054] Figure 3 For a schematic diagram of the visual operation window provided in the embodiment of the present application, see Figure 3 , including a user question input box and a "send" control. The user can enter the initial question in the user question input box, and by clicking the "send" control, the user's initial question can be sent to the information search device.

[0055] S202: Preprocess the initial question to obtain semantic vectors of multiple question keywords, target questions and initial questions.

[0056] The target problem is the problem after error correction of the initial problem.

[0057] The target question can be obtained in the following way: obtain multiple preset questions and multiple preset knowledge; construct a corresponding knowledge graph based on the multiple preset questions and multiple preset knowledge; correct the user's initial question through the knowledge graph to obtain the target question.

[0058] Preset questions refer to pre-set questions that are related or similar to the initial question. For example, if the initial question is "When was Zhang San born?", the preset questions can be "When is Zhang San's date of birth?", "When was Zhang San born?", etc.

[0059] Preset knowledge refers to knowledge points that are related or similar to the preset question. For example, assuming the preset question is "When is Zhang San's date of birth?", the preset knowledge can be "Zhang San's date of birth is December 10, 1980."

[0060] The knowledge graph is used to indicate various entities and their relationships in the real world, that is, the association relationship between multiple entities can be obtained based on the knowledge graph. Entities can be conceptual nouns such as names of people and places. For example, assuming that the entities are "Zhang San", "City A", and "City B", the relationship between "Zhang San" and "City A" can be "birth", and the relationship between "Zhang San" and "City B" can be "workplace".

[0061] In actual application, the error information in the initial question can be corrected according to the knowledge graph to obtain the target question. For example, assuming the initial question is "Is Zhang San's date of departure December 10, 1980?", if the relationship between "Zhang San" and "December 10, 1980" in the knowledge graph is "date of birth", then the target question can be "Is Zhang San's date of birth December 10, 1980?".

[0062] S203. According to multiple question keywords, target questions and semantic vectors, multiple answers corresponding to the initial question are obtained in the knowledge base through multiple recall methods.

[0063] The recall method refers to the method of obtaining answers in the knowledge base. For example, the recall method can be a keyword retrieval method, a question-answer pair retrieval method, and a semantic vector retrieval method.

[0064] Multiple answers corresponding to the initial question can be obtained in the following ways: determine the knowledge base; obtain multiple first answers in the knowledge base through keyword retrieval based on the question keywords; obtain multiple second answers in the knowledge base through question-answer pair retrieval based on the target question; obtain multiple third answers in the knowledge base through semantic vector retrieval based on the semantic vector; determine that the multiple answers include multiple first answers, multiple second answers and multiple third answers.

[0065] Among them, the knowledge base includes multiple historical case information, multiple first question-answer pairs, and data vectors. The multiple first question-answer pairs include actual question-answer pairs and question-answer pairs generated by a preset model. The data vectors are obtained by vectorizing multiple historical case information and multiple question-answer pairs.

[0066] Historical case information refers to case information within a historical period. The historical case information includes multiple preset case information, for example, the preset case information may be case type, case summary, case details, case result, etc. In actual application, the historical case information may be determined from the preset case library. The preset case library includes multiple case documents, for example, the case document may be a judgment document.

[0067] Historical case information may be obtained in the following ways: obtaining multiple case documents; disassembling the case documents to obtain multiple case information; extracting multiple case information to obtain multiple preset case information; and determining historical case information based on multiple preset case information.

[0068] The actual question-answer pair refers to the historical questions related to the initial question and the corresponding historical question-answer results. The actual question-answer pair can be pre-set. For example, the question in the actual question-answer pair can be "Where is Zhang San born?", and the answer can be "Zhang San was born in City A.".

[0069] S204: Determine the relevance between each answer and the initial question, the information type of each answer, and the recall method of each answer.

[0070] Relevance refers to the degree of match between the answer and the original question.

[0071] Among them, information types include user information type, case information type and processing result type.

[0072] User information type refers to the basic information type related to the user. For example, the user information type can be identity information type, account information type, etc.

[0073] The case information type is used to indicate the attributes of the case. For example, the attributes of a case may be transaction type, dispute type, etc. When the attribute of a case is transaction type, the case information type may be a transaction case; when the attribute of a case is dispute type, the case information type may be a dispute case.

[0074] The processing result type is used to indicate the processing status of the case. For example, the processing status of the case can be a fine, a penalty duration, etc. When the processing status of the case is a fine, the processing result type can be a fine type; when the processing result type is a penalty duration, the processing result type is a penalty duration type.

[0075] For example, the information type may be as shown in Table 1:

[0076] Table 1

[0077] User Information Type Case Information Type Processing result type Type of identity information Transactional Cases Types of fines Account Information Type Dispute cases Penalty duration type …… …… ……

[0078] S205. Determine multiple answers to be selected from multiple answers based on the relevance between each answer and the initial question, the information type of each answer, and the recall method of each answer.

[0079] Multiple answers to be selected may be determined in the following manner: a first answer set and a second answer set are determined from multiple answers based on the recall method of each answer; multiple answers to be selected are determined based on the relevance between each answer and the initial question, the information type of each answer, the first answer set and the second answer set.

[0080] Among them, the recall method corresponding to each answer in the first answer set is a keyword retrieval method or a semantic vector retrieval method, and the recall method corresponding to each answer in the second answer set is a question-answer pair retrieval method.

[0081] The answers in the first answer set can be sorted in descending order of relevance, and the first M answers can be obtained from the sorted first answer set; the answers in the second answer set can be sorted in descending order of relevance, and N answers can be obtained from the sorted second answer set; according to the information type of each answer to be selected in the first M answers and the first N answers, multiple answers to be selected can be determined in the first M answers and the first N answers, where M and N are positive integers.

[0082] For example, assuming that the first answer set includes answers 1 to 6, and the relevance decreases in sequence, that is, in a descending order of relevance, the first answer set can be shown in Table 2:

[0083] Table 2

[0084] Answer Information Type Answer 1 User Information Type Answer 2 Case Information Type Answer 3 Processing result type Answer 4 Case Information Type Answer 5 Case Information Type Answer 6 Processing result type

[0085] In the second answer set, assuming that the second answer set includes answers 7 to 12, and the relevance decreases in sequence, that is, in a descending order of relevance, the second answer set can be shown in Table 3:

[0086] Table 3

[0087]

[0088]

[0089] According to Table 2 and Table 3, if M is 5 and N is 4, answers 1 to 5 can be obtained in the first answer set, and answers 7 to 10 can be obtained in the second answer set.

[0090] In actual application, multiple answers can be determined according to the weight of the information type. Taking Table 2 and Table 3 as examples, assuming that a total of 6 answers are needed and the weight of each information type is equal, the number of answers corresponding to each information type is 2, and the corresponding number of answers can be selected from answers 1 to 5 and answers 7 to 10.

[0091] S206: Process the initial question and multiple answers to be selected through a preset model to obtain a target answer corresponding to the initial question.

[0092] The preset model refers to a pre-trained model. The initial question and multiple candidate answers can be processed by the LLM large model to obtain the target answer.

[0093] In actual application, the target answer can be compared with the information. If there is wrong information in the target answer, the wrong information is corrected. If there is missing information in the target answer, the missing information is supplemented to the target answer.

[0094] In an embodiment of the present application, after obtaining the user's initial question, the user's initial question is preprocessed, and based on the multiple keywords obtained by the preprocessing, the target question and the semantic vector corresponding to the initial question, multiple answers corresponding to the initial question are obtained in the knowledge base through multiple recall methods, and multiple answers to be selected are determined based on the correlation between each answer and the initial question, the information type of each answer and the recall method, and then the initial question and the multiple answers to be selected are processed through a preset model to obtain the target answer. In the above process, relevant answers can be searched in the knowledge base through multiple recall methods, and the target answer can be quickly and accurately determined through the preset model, which improves the efficiency of information search.

[0095] Figure 4 A schematic diagram of the process of determining multiple answers to be selected provided in the embodiment of the present application. Figure 4 , the method may include the following steps:

[0096] S401: Obtain an initial question input by a user.

[0097] It should be noted that the execution process of the above S401 can refer to S201 and will not be described in detail here.

[0098] S402: Preprocess the initial question to obtain semantic vectors of multiple question keywords, target questions and initial questions.

[0099] The target problem is the problem after error correction of the initial problem.

[0100] The semantic vector is used to indicate the transformation of the initial question from text form to numerical vector form.

[0101] For example, assuming the initial question input by the user is "Does Zhang San work in City A?", the question keywords may be "Zhang San", "City A", and "work", and the target question may be "Is Zhang San's workplace City A?".

[0102] S403: Determine the knowledge base.

[0103] Among them, the knowledge base includes multiple historical case information, multiple first question-answer pairs, and data vectors. The multiple first question-answer pairs include actual question-answer pairs and question-answer pairs generated by a preset model. The data vectors are obtained by vectorizing multiple historical case information and multiple question-answer pairs.

[0104] In actual application, data cleaning, deduplication and disassembly processing can be performed in advance on multiple historical case information and multiple first question-answer pairs in the knowledge base.

[0105] S404: Acquire multiple first answers in the knowledge base according to the question keywords by keyword retrieval.

[0106] In actual application, the first answer can be obtained in the knowledge base according to at least one question keyword. For example, assuming the question keywords are "Zhang San", "yes", "born" and "city A", the first answer can be "Zhang San was born in city A" or "Zhang San's date of birth is December 10, 1980" according to the three question keywords "Zhang San" and "born".

[0107] S405: According to the target question, multiple second answers are obtained in the knowledge base through a question-answer pair retrieval method.

[0108] S406: Acquire multiple third answers in the knowledge base according to the semantic vector by using a semantic vector search method.

[0109] According to the search semantic model, the semantic vector can be retrieved in the data vector to obtain multiple third answers.

[0110] It should be noted that after S403 , S404 , S405 and S406 may be executed in parallel.

[0111] S407: Determine the relevance between each answer and the initial question, the information type of each answer, and the recall method of each answer.

[0112] Among them, information types include user information type, case information type and processing result type.

[0113] It should be noted that the execution process of the above S407 can refer to S204, which will not be repeated here.

[0114] S408. Determine a first answer set and a second answer set from multiple answers according to the recall method of each answer.

[0115] S409 , sort the answers in the first answer set in descending order of relevance, and obtain the first M answers in the sorted first answer set.

[0116] S410 , sorting the answers in the second answer set in descending order of relevance, and obtaining N answers from the sorted second answer set.

[0117] It should be noted that after S408 , S409 and S410 may be executed in parallel.

[0118] S411. Determine the weight corresponding to each information type.

[0119] S412: For any information type, determine the number of answers corresponding to the information type according to the weight corresponding to the information type.

[0120] For example, assuming that a total of 10 answers are needed, the weight of the user information type is 0.6, the weight of the case information type is 0.2, and the weight of the processing situation type is 0.2, then the number of answers corresponding to the user information type is 6, the number of answers corresponding to the case information type is 2, and the number of answers corresponding to the processing situation type is 2.

[0121] The number of answers corresponding to each information type can be shown in Table 4:

[0122] Table 4

[0123] Information Type User Information Type Case Information Type Type of situation to be handled Weight 0.6 0.2 0.2 Number of answers 6 2 2

[0124] S413. According to the number of answers, determine at least one answer to be selected corresponding to the information type among the first M answers and the first N answers.

[0125] At least one answer to be selected may include user information, case information, case result information, etc.

[0126] S414: Determine that the multiple answers to be selected include: at least one answer to be selected corresponding to each information type.

[0127] In an embodiment of the present application, after obtaining the user's initial question, the initial question is preprocessed to obtain multiple question keywords, target questions, and semantic vectors corresponding to the initial question, and multiple answers are obtained in the knowledge base through keyword retrieval, question-answer pair retrieval, and semantic vector retrieval. According to the correlation between each answer and the initial question and the recall method, the first M answers of the first answer set and the first N answers of the second answer set are determined, and according to the weight corresponding to the information type, the corresponding number of answers is determined, thereby determining multiple answers to be selected. In the above process, multiple recall methods can be used to determine multiple answers to be selected, combined with the correlation and the weight of the information type, to ensure that the answers to be selected are highly relevant to the business needs, thereby improving the efficiency of information search. Furthermore, the answers to be selected can be close to the business needs and the initial question to a large extent, making the search results more comprehensive and accurate.

[0128] Next, combine Figure 5 , the process of determining multiple candidate answers is explained.

[0129] Figure 5 Another schematic diagram of the process of determining multiple answers to be selected provided in the embodiment of the present application. Figure 5 After obtaining the user's initial question, the initial question is preprocessed. For example, the preprocessing may include: extracting question keywords from the initial question, processing the initial question through the knowledge graph to obtain the target question, and determining the semantic vector corresponding to the initial question.

[0130] After obtaining the question keywords, target question and semantic vector, multi-way recall processing can be performed based on the question keywords, target question and semantic vector. For example, the multi-way recall processing may include: recalling answers by keyword retrieval based on the question keywords, recalling answers by question-answer pair retrieval based on the target question, and recalling answers by semantic vector retrieval based on the semantic vector by searching the semantic model in the vector knowledge base. After performing the multi-way recall processing, multiple answers can be obtained.

[0131] After obtaining multiple answers, the multiple answers can be sorted by relevance, and multiple candidate answers can be obtained according to the information type. The candidate answers can be processed through the LLM large model to obtain the target answer corresponding to the initial question.

[0132] Next, combine Figure 6 , the process of determining the target answer is explained.

[0133] Figure 6 Schematic diagram of the process of determining the target answer provided in the embodiment of the present application. Figure 5 , the method may include the following steps:

[0134] S601: Obtain an initial question input by a user.

[0135] S602: Determine multiple answers to be selected.

[0136] It should be noted that the execution process of the above S602 can refer to S401 to S414, which will not be described in detail here.

[0137] S603: Process the initial question and multiple candidate answers through a preset model to obtain a target answer.

[0138] The target answer can be determined based on the correspondence between the initial question and multiple answers to be selected.

[0139] In an embodiment of the present application, after obtaining the initial question and multiple answers to be selected, the initial question and the multiple answers to be selected are input into a preset model, and the preset model obtains the target answer based on the correspondence between the initial question and the multiple answers to be selected, and performs information comparison on the target answer to obtain the final target answer. In the above process, the target answer can be quickly and accurately determined through the preset model, thereby improving the efficiency of information search. Furthermore, the target answer can be compared with information to ensure the high correlation between the target answer and the initial question, thereby improving the accuracy of the target answer.

[0140] Next, combine Figure 7 , through specific examples, Figure 6 The illustrated embodiment is described.

[0141] Figure 7 A schematic diagram of another target answer determination process provided in an embodiment of the present application. Figure 7 After obtaining the user's initial question and the answers to be selected (including user information, case information and case result information), the initial question and the answers to be selected can be processed in various ways through the preset model, such as multimodal analysis and recognition, text reading and reasoning, element recognition and extraction, user information verification, and case information verification (integrity verification, validity verification and compliance verification) to obtain the target answer.

[0142] Based on any of the above embodiments, in the actual application process, in order to improve the accuracy of the question and answer of the preset model, the knowledge base and the preset model can be updated periodically. Figure 6The illustrated embodiment is described.

[0143] Figure 8 A schematic diagram of the process of updating the knowledge base and the preset model provided in the embodiment of the present application. Figure 8 , the method may include the following steps:

[0144] S801. Obtain a set of question-answer pairs and the accuracy of each question-answer pair in the set of question-answer pairs.

[0145] The question-answer pair set refers to the initial question and the corresponding target answer within a preset historical time period.

[0146] The preset historical time period is a pre-set period before the current time, for example, the preset historical time period may be the past day, the past month, etc.

[0147] Among them, accuracy is used to indicate the degree of match between the question and answer and the answer and the question.

[0148] In actual application, the accuracy of each question-answer pair can be determined based on the user's evaluation result of the answer. For example, assuming that the evaluation result can include satisfaction and dissatisfaction, if the user's evaluation result of the answer is satisfied, the accuracy of the question-answer pair is high; if the user's evaluation result of the answer is dissatisfied, the accuracy of the question-answer pair is low.

[0149] S802. Determine a plurality of second question-answer pairs and a plurality of third question-answer pairs in the question-answer pair set according to the accuracy of each question-answer pair in the question-answer pair set.

[0150] Among them, the accuracy of the second question and answer pair is greater than or equal to the preset threshold, and the accuracy of the third question and answer pair is less than the preset threshold.

[0151] S803: Correct the answers in the plurality of third question-answer pairs to obtain a plurality of fourth question-answer pairs.

[0152] S804: Update the plurality of second question-answer pairs and the plurality of fourth question-answer pairs to the knowledge base.

[0153] S805. Update the preset model according to the plurality of second question-answer pairs and the plurality of fourth question-answer pairs.

[0154] Next, combine Fig. 9 ,The visual operation window of user evaluation is explained through specific examples.

[0155] Fig. 9 For a schematic diagram of a visual operation window for user evaluation provided in an embodiment of the present application, see Fig. 9, including a user question input box, an answer generation box, a "Send" control, a "Copy" control and an "Evaluation" control. The user can enter the initial question in the user question input box. By clicking the "Send" control, the user's initial question can be sent to the information search device. The information search device generates a target answer based on the initial question and displays it through the answer generation box.

[0156] Users can evaluate the accuracy of the generated answers through the "Evaluation" module, and can also click the "Copy" control to copy the answer text.

[0157] Next, combine Fig.10 , the process of the information search method is explained.

[0158] Fig.10 This is a flowchart of another information search method provided in an embodiment of the present application. Fig.10 , a knowledge base can be built based on historical case information, actual question-answer pairs, and annotated data (business demand data), and this information can be vectorized to obtain data vectors. Optionally, a question-answer pair can be generated based on a preset model, and the question-answer pair generated by the preset model can be added to the knowledge base.

[0159] In actual application, after obtaining the user's initial question, the initial question is searched in combination with the knowledge base (through keyword search, question-answer pair search and semantic vector search) to obtain multiple answers, and multiple candidate answers are selected from the multiple answers. The initial question and the candidate answers are processed and information is compared through a preset model to obtain the target answer.

[0160] It is also possible to periodically obtain a set of question-answer pairs output by a preset model, as well as the corresponding accuracy (which can be determined by user evaluation), and perform correction processing on the question-answer pairs according to the accuracy to obtain multiple question-answer pairs. The knowledge base can be updated based on multiple question-answer pairs and question-answer pairs generated by the preset model, and the preset model can be updated to improve the efficiency of information search.

[0161] In an embodiment of the present application, after obtaining a set of question-answer pairs and the accuracy of each question-answer pair in the set of question-answer pairs, it is determined whether the accuracy of the question-answer pair is greater than or equal to a preset threshold. If the accuracy of the question-answer pair is greater than or equal to the preset threshold, the question-answer pair is determined as the second question-answer pair; if the accuracy of the question-answer pair is less than the preset threshold, the question-answer pair is determined as the third question-answer pair, and the third question-answer pair is corrected to obtain a fourth question-answer pair. Thereafter, the knowledge base and the preset model are updated based on multiple second question-answer pairs and multiple fourth question-answer pairs. In the above process, the knowledge base and the preset model can be updated periodically to ensure that the knowledge base and the preset model can be adjusted and optimized according to changes in the question-answer pairs, thereby improving the efficiency of information search.

[0162] Fig.11 This is a schematic diagram of the structure of an information search device provided in an embodiment of the present application. Fig.11 The information search device 10 includes: a first acquisition module 11, a processing module 12, a second acquisition module 13 and a determination module 14, wherein:

[0163] The first acquisition module 11 is used to acquire an initial question input by a user;

[0164] The processing module 12 is used to pre-process the initial question to obtain a plurality of question keywords, a target question and a semantic vector of the initial question, wherein the target question is the question after error correction processing is performed on the initial question;

[0165] The second acquisition module 13 is used to acquire multiple answers corresponding to the initial question in the knowledge base through multiple recall methods according to the multiple question keywords, the target question and the semantic vector;

[0166] The determination module 14 is used to determine the relevance between each answer and the initial question, the information type of each answer and the recall method of each answer, wherein the information type includes user information type, case information type and processing result type;

[0167] The determination module 14 is further configured to determine a plurality of answers to be selected from the plurality of answers according to the relevance between each answer and the initial question, the information type of each answer, and the recall method of each answer;

[0168] The processing module 12 is further used to process the initial question and the multiple answers to be selected through a preset model to obtain a target answer corresponding to the initial question.

[0169] The information search device provided in the embodiment of the present application can execute the method shown in the above method embodiment, and its implementation principle and beneficial effects are similar, which will not be repeated here.

[0170] In a possible implementation manner, the determining module 14 is specifically configured to:

[0171] Determine a first answer set and a second answer set from the multiple answers according to a recall method of each answer, wherein the recall method corresponding to each answer in the first answer set is a keyword search method or a semantic vector search method, and the recall method corresponding to each answer in the second answer set is a question-answer pair search method;

[0172] The plurality of answers to be selected are determined according to the relevance between each answer and the initial question, the information type of each answer, the first answer set, and the second answer set.

[0173] In a possible implementation manner, the determining module 14 is specifically configured to:

[0174] Sort the answers in the first answer set in descending order of relevance, and obtain the first M answers in the sorted first answer set, where M is a positive integer;

[0175] Sorting the answers in the second answer set in descending order of relevance, and obtaining N answers from the sorted second answer set, where N is a positive integer;

[0176] The plurality of answers to be selected are determined from the first M answers and the first N answers according to the information type of each answer to be selected from the first M answers and the first N answers.

[0177] In a possible implementation manner, the determining module 14 is specifically configured to:

[0178] Determine the weight corresponding to each information type;

[0179] For any information type, determine the number of answers corresponding to the information type according to the weight corresponding to the information type, and determine at least one answer to be selected corresponding to the information type from among the top M answers and the top N answers according to the number of answers;

[0180] Determine that the multiple answers to be selected include: at least one answer to be selected corresponding to each information type.

[0181] In a possible implementation manner, the second acquisition module 13 is specifically configured to:

[0182] Determine the knowledge base, wherein the knowledge base includes a plurality of historical case information, a plurality of first question-answer pairs, and a data vector, wherein the plurality of first question-answer pairs include actual question-answer pairs and question-answer pairs generated by a preset model, and the data vector is obtained by vectorizing the plurality of historical case information and the plurality of question-answer pairs;

[0183] According to the question keywords, a plurality of first answers are obtained in the knowledge base by keyword retrieval;

[0184] According to the target question, obtaining a plurality of second answers in the knowledge base by using a question-answer pair retrieval method;

[0185] According to the semantic vector, obtaining a plurality of third answers in the knowledge base by semantic vector retrieval;

[0186] Determining the plurality of answers includes the plurality of first answers, the plurality of second answers, and the plurality of third answers.

[0187] Fig.12 A schematic diagram of the structure of another information search device provided in an embodiment of the present application. Fig.11 Based on the examples shown, see Fig.12 The information search device 10 further includes a third acquisition module 15, a correction module 16 and an update module 17, wherein:

[0188] The third acquisition module 15 is used to acquire a set of question-answer pairs and the accuracy of each question-answer pair in the set of question-answer pairs, wherein the accuracy is used to indicate the matching degree between the answer and the question in the question-answer pair;

[0189] The determination module 14 is further configured to determine, according to the accuracy of each question-answer pair in the question-answer pair set, a plurality of second question-answer pairs and a plurality of third question-answer pairs in the question-answer pair set, wherein the accuracy of the second question-answer pairs is greater than or equal to a preset threshold, and the accuracy of the third question-answer pairs is less than the preset threshold;

[0190] The correction module 16 is used to correct the answers in the plurality of third question-answer pairs to obtain a plurality of fourth question-answer pairs;

[0191] The updating module 17 is used to update the plurality of second question-answer pairs and the plurality of fourth question-answer pairs to the knowledge base.

[0192] In a possible implementation manner, the updating module 17 is further configured to:

[0193] The preset model is updated according to the plurality of second question-answer pairs and the plurality of fourth question-answer pairs.

[0194] The information search device provided in the embodiment of the present application can execute the method shown in the above method embodiment, and its implementation principle and beneficial effects are similar, which will not be repeated here.

[0195] Fig.13 This is a schematic diagram of the structure of the information search device provided in the embodiment of the present application. Fig.13 As shown, the information search device 20 may include: a transceiver 21 , a processor 22 , and a memory 23 .

[0196] The processor 22 executes the computer execution instructions stored in the memory, so that the processor 22 executes the scheme in the above embodiment. The processor 22 can be a general-purpose processor, including a central processing unit CPU, a network processor (NP), etc.; it can also be a digital signal processor DSP, an application-specific integrated circuit ASIC, a field programmable gate array FPGA or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.

[0197] The memory 23 is connected to the processor 22 via a system bus and completes communication between them. The memory 23 is used to store computer program instructions.

[0198] The transceiver 21 may be used to obtain tasks to be executed and configuration information of the tasks to be executed.

[0199] The system bus can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The system bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The transceiver is used to realize the communication between the database access device and other computers (such as clients, read-write libraries, and read-only libraries). The memory may include random access memory (RAM) and may also include non-volatile memory.

[0200] The electronic device provided in the embodiment of the present application may be the terminal device of the above embodiment.

[0201] An embodiment of the present application also provides a chip for executing instructions, which is used to execute the technical solution of the information search method in the above embodiment.

[0202] An embodiment of the present application also provides a computer-readable storage medium, in which computer instructions are stored. When the computer instructions are executed on a computer, the computer executes the technical solution of the information search method in the above embodiment.

[0203] An embodiment of the present application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium. When at least one processor executes the computer program, it can implement the technical solution of the information search method in the above embodiment.

[0204] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the present application.

[0205] It should be further noted that, although the various steps in the flowchart are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowchart may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

[0206] It should be understood that the above-mentioned device embodiments are only illustrative, and the device of the present application can also be implemented in other ways. For example, the division of units / modules in the above-mentioned embodiments is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units, modules or components can be combined, or can be integrated into another system, or some features can be ignored or not executed.

[0207] In addition, unless otherwise specified, each functional unit / module in each embodiment of the present application may be integrated into one unit / module, each unit / module may exist physically separately, or two or more units / modules may be integrated together. The above-mentioned integrated unit / module may be implemented in the form of hardware or in the form of a software program module.

[0208] If the integrated unit / module is implemented in the form of hardware, the hardware may be a digital circuit, an analog circuit, etc. The physical implementation of the hardware structure includes but is not limited to transistors, memristors, etc. If not specifically stated, the processor may be any appropriate hardware processor, such as a CPU, a GPU, an FPGA, a DSP, an ASIC, etc. If not specifically stated, the storage unit may be any appropriate magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc.

[0209] If the integrated unit / module is implemented in the form of a software program module and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or all or part of the technical solution, can be embodied in the form of a software product, which is stored in a memory and includes several instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the various embodiments of the present application. The aforementioned memory includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

[0210] In the above embodiments, the description of each embodiment has its own emphasis. For the part not described in detail in a certain embodiment, please refer to the relevant description of other embodiments. The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not 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.

[0211] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0212] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. An information search method, characterized in that: include: Get the initial question from the user; Preprocessing the initial question to obtain a plurality of question keywords, a target question, and a semantic vector of the initial question, wherein the target question is the question after error correction processing is performed on the initial question; According to the multiple question keywords, the target question and the semantic vector, obtaining multiple answers corresponding to the initial question in the knowledge base through multiple recall methods; Determine the relevance between each answer and the initial question, the information type of each answer, and the recall method of each answer, wherein the information type includes user information type, case information type, and processing result type; Determining a plurality of answers to be selected from the plurality of answers according to the relevance between each answer and the initial question, the information type of each answer, and the recall method of each answer; The initial question and the multiple answers to be selected are processed through a preset model to obtain a target answer corresponding to the initial question.

2. The method according to claim 1, characterized in that Determining multiple answers to be selected from the multiple answers according to the relevance between each answer and the initial question, the information type of each answer, and the recall method of each answer, including: Determine a first answer set and a second answer set from the multiple answers according to a recall method of each answer, wherein the recall method corresponding to each answer in the first answer set is a keyword search method or a semantic vector search method, and the recall method corresponding to each answer in the second answer set is a question-answer pair search method; The plurality of answers to be selected are determined according to the relevance between each answer and the initial question, the information type of each answer, the first answer set, and the second answer set.

3. The method according to claim 2, characterized in that Determining the plurality of answers to be selected according to the relevance between each answer and the initial question, the information type of each answer, the first answer set, and the second answer set includes: Sort the answers in the first answer set in descending order of relevance, and obtain the first M answers in the sorted first answer set, where M is a positive integer; Sorting the answers in the second answer set in descending order of relevance, and obtaining N answers from the sorted second answer set, where N is a positive integer; The plurality of answers to be selected are determined from the first M answers and the first N answers according to the information type of each answer to be selected from the first M answers and the first N answers.

4. The method according to claim 3, characterized in that Determining the plurality of answers to be selected from the first M answers and the first N answers according to the information type of each answer to be selected from the first M answers and the first N answers includes: Determine the weight corresponding to each information type; For any information type, determine the number of answers corresponding to the information type according to the weight corresponding to the information type, and determine at least one answer to be selected corresponding to the information type from among the top M answers and the top N answers according to the number of answers; Determine that the multiple answers to be selected include: at least one answer to be selected corresponding to each information type.

5. The method according to any one of claims 1 to 3, characterized in that: According to the multiple question keywords, the target question and the semantic vector, multiple answers corresponding to the initial question are obtained through multiple recall methods, including: Determine the knowledge base, wherein the knowledge base includes a plurality of historical case information, a plurality of first question-answer pairs, and a data vector, wherein the plurality of first question-answer pairs include actual question-answer pairs and question-answer pairs generated by a preset model, and the data vector is obtained by vectorizing the plurality of historical case information and the plurality of question-answer pairs; According to the question keywords, a plurality of first answers are obtained in the knowledge base by keyword retrieval; According to the target question, obtaining a plurality of second answers in the knowledge base by using a question-answer pair retrieval method; According to the semantic vector, obtaining a plurality of third answers in the knowledge base by semantic vector retrieval; Determining the plurality of answers includes the plurality of first answers, the plurality of second answers, and the plurality of third answers.

6. The method according to any one of claims 1 to 5, characterized in that: The method further comprises: Obtaining a set of question-answer pairs and the accuracy of each question-answer pair in the set of question-answer pairs, wherein the accuracy is used to indicate the degree of matching between the answer and the question in the question-answer pair; Determine, according to the accuracy of each question-answer pair in the question-answer pair set, a plurality of second question-answer pairs and a plurality of third question-answer pairs in the question-answer pair set, wherein the accuracy of the second question-answer pairs is greater than or equal to a preset threshold, and the accuracy of the third question-answer pairs is less than the preset threshold; Correcting the answers in the plurality of third question-answer pairs to obtain a plurality of fourth question-answer pairs; The plurality of second question-answer pairs and the plurality of fourth question-answer pairs are updated to the knowledge base.

7. The method according to claim 6, characterized in that The method further comprises: The preset model is updated according to the plurality of second question-answer pairs and the plurality of fourth question-answer pairs.

8. An information search device, characterized in that: include: A first acquisition module, a processing module, a second acquisition module and a determination module, wherein: The first acquisition module is used to obtain an initial question input by a user; The processing module is used to pre-process the initial question to obtain a plurality of question keywords, a target question and a semantic vector of the initial question, wherein the target question is the question after error correction processing is performed on the initial question; The second acquisition module is used to acquire multiple answers corresponding to the initial question in the knowledge base through multiple recall methods according to the multiple question keywords, the target question and the semantic vector; The determination module is used to determine the relevance between each answer and the initial question, the information type of each answer and the recall method of each answer, wherein the information type includes user information type, case information type and processing result type; The determination module is further used to determine a plurality of answers to be selected from the plurality of answers according to the relevance between each answer and the initial question, the information type of each answer and the recall method of each answer; The processing module is also used to process the initial question and the multiple answers to be selected through a preset model to obtain a target answer corresponding to the initial question.

9. An information search device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 7 when executed by a processor.

Images