Knowledge processing method, message processing method, computing device, storage medium and program product

Abstract

The embodiment of the invention provides a knowledge processing method, a message processing method, computing equipment, a storage medium and a program product. The method comprises the following steps: taking historical customer service session data or object description information as a data source, and identifying candidate problem data and candidate response data corresponding to the candidate problem data from the data source by utilizing a knowledge identification model; querying whether a target question and answer knowledge pair which meets a first similar requirement with a candidate question and answer knowledge pair formed by the candidate question data and the candidate response data exists in a question and answer knowledge base; if not, storing the candidate question and answer knowledge pairs in a question and answer database; and if yes, evaluating the candidate question-answer knowledge pair and the target question-answer knowledge pair by utilizing the knowledge evaluation model, and updating the target question-answer knowledge pair in the question-answer database by utilizing the candidate question-answer knowledge pair under the condition that the evaluation score of the candidate question-answer knowledge pair is higher than that of the target question-answer knowledge pair. According to the technical scheme, the accuracy of the question and answer knowledge base can be improved.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of computer, and particularly relate to a knowledge processing method, a message processing method, a computing device, a storage medium and a program product. BACKGROUND

[0002] With the rapid development of Internet technology, intelligent customer service has been widely applied in various service fields. To realize intelligent customer service, it is usually necessary to pre-build a question and answer knowledge base. The question and answer knowledge base stores question and answer knowledge pairs composed of question data and answer data. When providing intelligent customer service, the matching question and answer knowledge pair can be queried from the question and answer knowledge base according to the question data consulted by the user, and then the corresponding answer data is obtained and returned to the user, so as to realize intelligent question and answer without human intervention.

[0003] However, the current question and answer knowledge base is usually built and generated by operation personnel for the to-be-served object, which leads to inaccurate question and answer knowledge base and affects the service effect of intelligent customer service. SUMMARY

[0004] Embodiments of the present application provide a knowledge processing method, a message processing method, a computing device, a storage medium and a program product to solve the technical problem of inaccurate data in the question and answer knowledge base in the prior art.

[0005] In a first aspect, a knowledge processing method is provided in embodiments of the present application, comprising:

[0006] The historical customer service conversation data or the object description information is taken as a data source;

[0007] A candidate question data and candidate answer data corresponding to the candidate question data are identified from the data source by using a knowledge recognition model;

[0008] It is queried from a question and answer knowledge base whether there is a target question and answer knowledge pair meeting a first similarity requirement with the candidate question and answer knowledge pair composed of the candidate question data and the candidate answer data; the target question and answer knowledge pair is composed of question data and answer data;

[0009] If the target question and answer knowledge pair does not exist in the question and answer knowledge base, the candidate question and answer knowledge pair is saved to the question and answer database;

[0010] If the target question and answer knowledge pair exists in the question and answer knowledge base, a knowledge evaluation model is used to evaluate the candidate question and answer knowledge pair and the target question and answer knowledge pair respectively, and in the case that the evaluation score of the candidate question and answer knowledge pair is higher than that of the target question and answer knowledge pair, the target question and answer knowledge pair in the question and answer database is updated by using the candidate question and answer knowledge pair.

[0011] Optionally, the identifying, by using the knowledge recognition model, candidate question data and candidate answer data corresponding to the candidate question data from the data source comprises:

[0012] identifying, by using the knowledge recognition model, candidate question data, candidate knowledge type corresponding to the candidate question data, and candidate answer data from the data source;

[0013] the saving, if the target question and answer knowledge pair does not exist in the question and answer knowledge base, the candidate question and answer knowledge pair into the question and answer database comprises:

[0014] the saving, if the target question and answer knowledge pair does not exist in the question and answer knowledge base, the candidate question and answer knowledge pair corresponding to the candidate knowledge type into the question and answer database;

[0015] the evaluating, if the target question and answer knowledge pair exists in the question and answer knowledge base, the candidate question and answer knowledge pair and the target question and answer knowledge pair by using the knowledge evaluation model respectively comprises:

[0016] the evaluating, if the target question and answer knowledge pair exists in the question and answer knowledge base, the candidate question and answer knowledge pair and the target question and answer knowledge pair by using the knowledge evaluation model respectively comprises:

[0017] Optionally, the identifying, by using the knowledge recognition model, candidate question data and candidate answer data corresponding to the candidate question data from the data source comprises:

[0018] identifying, by using the knowledge recognition model, candidate question data, candidate knowledge type corresponding to the candidate question data, and candidate answer data from the data source;

[0019] the querying, from the question and answer knowledge base, whether there exists a target question and answer knowledge pair meeting the first similarity requirement with the candidate question and answer knowledge pair constituted by the candidate question data and the candidate answer data comprises:

[0020] determining, from the question and answer knowledge base, a question and answer pair corresponding to the candidate knowledge type;

[0021] the querying, from the question and answer pair corresponding to the candidate knowledge type, whether there exists a target question and answer knowledge pair meeting the first similarity requirement with the candidate question and answer knowledge pair constituted by the candidate question data and the candidate answer data.

[0022] Optionally, the identifying, by using the knowledge recognition model, candidate question data and candidate answer data corresponding to the candidate question data from the data source comprises:

[0023] performing segmentation on the data source to obtain a plurality of data segments;

[0024] For any data segment, a knowledge recognition model is used to recognize candidate question data and candidate answer data corresponding to the candidate question data from the data segment.

[0025] Optionally, if the target question and answer knowledge pair does not exist in the question and answer knowledge base, the candidate question and answer knowledge pair is saved into the question and answer database, including:

[0026] If the target question and answer knowledge pair does not exist in the question and answer knowledge base, a first confirmation prompt information is sent to relevant personnel based on the candidate question and answer knowledge pair;

[0027] In response to a confirmation request or an update request for the candidate question and answer knowledge pair, the candidate question and answer knowledge pair or the candidate question and answer knowledge pair after updating is saved into the question and answer database.

[0028] Optionally, if the target question and answer knowledge pair exists in the question and answer knowledge base, a knowledge evaluation model is used to evaluate the candidate question and answer knowledge pair and the target question and answer knowledge pair respectively, and in a case where the evaluation score of the candidate question and answer knowledge pair is higher than that of the target question and answer knowledge pair, the target question and answer knowledge pair in the question and answer database is updated by using the candidate question and answer knowledge pair, including:

[0029] If the target question and answer knowledge pair exists in the question and answer knowledge base, a knowledge evaluation model is used to evaluate the candidate question and answer knowledge pair and the target question and answer knowledge pair respectively;

[0030] In a case where the evaluation score of the candidate question and answer knowledge pair is higher than that of the target question and answer knowledge pair by a first predetermined score, the target question and answer knowledge pair in the question and answer database is updated by using the candidate question and answer knowledge pair;

[0031] In a case where the evaluation score of the candidate question and answer knowledge pair is not higher than that of the target question and answer knowledge pair by the first predetermined score, a second confirmation prompt information is sent to relevant personnel based on the candidate question and answer knowledge pair and the target question and answer knowledge pair;

[0032] In response to a selection request for the candidate question and answer knowledge pair, the target question and answer knowledge pair in the question and answer database is updated by using the candidate question and answer knowledge pair.

[0033] Optionally, the query from the question and answer knowledge base whether a candidate question and answer knowledge pair constituted by the candidate question data and the candidate answer data meets a first similarity requirement includes:

[0034] query, from the question and answer knowledge base, whether there exists target question data satisfying the first similarity requirement with the candidate question data, and if so, obtain a target question and answer pair composed of the target question data and corresponding target answer data thereof;

[0035] Alternatively, from the question and answer knowledge base, query whether there exists target question data satisfying the first similarity requirement with the candidate question data, and the candidate answer data satisfies the second similarity requirement with target answer data corresponding to the target question data; if so, obtain a target question and answer pair composed of the target question data and corresponding target answer data thereof.

[0036] Optionally, if there exists target question data satisfying the first similarity requirement with the candidate question data, and the candidate question and answer data does not satisfy the second similarity requirement with the target answer data corresponding to the target question data, a third confirmation prompt information is sent to relevant personnel.

[0037] In response to a selection request for the candidate answer data, the target answer data is updated with the candidate answer data.

[0038] Optionally, the historical customer service conversation data or object description information as the data source comprises:

[0039] In response to a knowledge update instruction, historical customer service conversation data corresponding to at least one artificial customer service is obtained, and the historical customer service conversation data corresponding to any artificial customer service is taken as the data source;

[0040] Alternatively, in response to the knowledge update instruction, object description information of at least one object newly published is obtained, and the object description information of any object is taken as the data source;

[0041] Alternatively, in response to an object publishing instruction of a target object, the object description information of the target object is taken as the data source.

[0042] Optionally, the candidate question data and the candidate answer data corresponding to the candidate question data are identified from the data source by using a knowledge recognition model, comprising:

[0043] Based on the data source and the identification requirement, a first prompt instruction is generated;

[0044] The first prompt instruction is input into the knowledge recognition model to instruct the knowledge recognition model to identify candidate question data and candidate answer data corresponding to the candidate question data from the data source according to the identification requirement.

[0045] In a second aspect, the embodiments of the present application provide a knowledge processing method, comprising:

[0046] The historical customer service session data or object description information is taken as a data source;

[0047] A candidate question data and a candidate answer data corresponding to the candidate question data are identified from the data source by using a knowledge recognition model;

[0048] A candidate question and answer knowledge pair composed of the candidate question data and the candidate answer data is saved into the question and answer database.

[0049] In a third aspect, a message processing method is provided in the embodiments of the present application, and the method comprises:

[0050] A user message sent by a user terminal based on a session connection of a customer service terminal is acquired;

[0051] A historical session message in the session connection is acquired;

[0052] The user message and the historical session message are counted by using a message processing model, so as to extract target question data;

[0053] At least one target answer data matched with the target question data is found from a question and answer knowledge base; the question and answer knowledge base is updated based on question and answer pairs extracted from a data source by using a knowledge recognition model; and the data source comprises historical customer service session data or object description information;

[0054] A reply message corresponding to the user message is generated according to the at least one target answer data.

[0055] Optionally, the finding of the at least one target answer data matched with the target question data from the question and answer knowledge base comprises:

[0056] A plurality of candidate question data satisfying a third similarity requirement with the target question data is found from the question and answer knowledge base;

[0057] A matching degree of a plurality of candidate answer data corresponding to the plurality of candidate question data with the target question data is calculated respectively;

[0058] At least one target answer data satisfying a matching requirement in the matching degree is determined.

[0059] Optionally, the counting of the user message and the historical session message by using the message processing model to extract the target question data comprises:

[0060] A fourth prompt instruction is generated based on the user message, the historical session message and an extraction requirement;

[0061] input the fourth prompt instruction into the message processing model to instruct the message processing model to extract target question data from the user message and the historical session message according to extraction requirements.

[0062] Optionally, the generating, according to the at least one target response data, of the reply message corresponding to the user message comprises:

[0063] generating, by using a response generation model, the reply message corresponding to the user message according to the at least one target response data.

[0064] In a fourth aspect, an embodiment of the present application provides a message processing method, comprising:

[0065] obtaining a user message sent by a user end based on a session connection with a customer end;

[0066] determining user feature data corresponding to the user end;

[0067] obtaining historical session messages corresponding to the user message;

[0068] performing statistics on the user message and the historical session messages by using a message processing model in combination with the user feature data, to extract target question data;

[0069] finding at least one target response data matched with the target question data from a question and answer knowledge base;

[0070] generating, according to the at least one target response data, a reply message corresponding to the user message.

[0071] Optionally, the customer end comprises a customer robot; and the obtaining of the user message sent by the user end based on the session connection with the customer end comprises:

[0072] establishing the session connection between the user end and the customer robot in response to a customer request sent by the user end;

[0073] obtaining the user message sent by the user end to the customer robot based on the session connection with the customer end;

[0074] The method further comprises:

[0075] sending the reply message to the user end.

[0076] Optionally, the method further comprises:

[0077] sending, based on the reply message, recommended prompt information to the customer end;

[0078] sending, in response to a recommended confirmation request sent by the customer end, the reply message to the user end.

[0079] In a fifth aspect, an embodiment of the present application provides a computing device, comprising a processing component and a storage component;

[0080] The storage component stores a computer program; the computer program is used to be invoked and executed by the processing component, so as to implement the knowledge processing method in the first aspect above or the knowledge processing method in the second aspect above or the message processing method in the third aspect above or the message processing method in the fourth aspect above.

[0081] In a sixth aspect, an embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processing component to implement the knowledge processing method in the first aspect above or the knowledge processing method in the second aspect above or the message processing method in the third aspect above or the message processing method in the fourth aspect above.

[0082] In a seventh aspect, an embodiment of the present application provides a computer program product, comprising a computer program or instructions, and the computer program or instructions are executed by a processing component to implement the knowledge processing method in the first aspect above or the knowledge processing method in the second aspect above or the message processing method in the third aspect above or the message processing method in the fourth aspect above.

[0083] The embodiments of the present application utilize the knowledge recognition model to automatically extract the question and answer knowledge pairs from the historical customer service conversation data or the object description information, and save the question and answer knowledge pairs to the question and answer knowledge base. Compared with the manual generation mode, the method is more objective and accurate, so as to improve the richness and comprehensiveness of the question and answer knowledge base. In addition, the extracted candidate question and answer knowledge pairs can be compared with the existing question and answer knowledge pairs in the question and answer knowledge base, so as to update the question and answer knowledge base by using the question and answer knowledge pairs with higher evaluation scores, and realize the supplement and update of the question and answer knowledge base, thereby further improving the accuracy of the question and answer knowledge pairs in the question and answer knowledge base.

[0084] These aspects or other aspects of the present application will be more apparent in the following description of the embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0085] The accompanying drawings, which are included to provide a further understanding of the present application, constitute a part of the present application, and the illustrative embodiments of the present application and their description serve to explain the present application, and do not limit the present application in any way. In the drawings:

[0086] Figure 1 A method flow chart of one embodiment of the knowledge processing method provided by the present application is shown;

[0087] Figure 2A method flow chart of another embodiment of the knowledge processing method provided by the present application is shown;

[0088] Figure 3 A method flow chart of an embodiment of the message processing method provided by the present application is shown;

[0089] Figure 4 A method flow chart of another embodiment of the message processing method provided by the present application is shown;

[0090] Figure 5 A scene interaction schematic diagram of an actual application of the embodiments of the present application is shown;

[0091] Figure 6 A structural schematic diagram of an embodiment of the knowledge processing apparatus provided by the present application is shown;

[0092] Figure 7 A structural schematic diagram of another embodiment of the knowledge processing apparatus provided by the present application is shown;

[0093] Figure 8 A structural schematic diagram of an embodiment of the message processing apparatus provided by the present application is shown;

[0094] Figure 9 A structural schematic diagram of another embodiment of the message processing apparatus provided by the present application is shown;

[0095] Figure 10 A structural schematic diagram of an embodiment of the computing device provided by the present application is shown. DETAILED DESCRIPTION

[0096] In order to make the objectives, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described below in connection with the embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without any creative work fall within the scope of protection of the present application.

[0097] It should be noted that in the case of the user information involved in the embodiments of the present application, the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the embodiments of the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation portal for user to choose authorization or refusal. In addition, the various models (including but not limited to language models or large models) involved in the present application are in line with relevant legal and standard regulations.

[0098] The technical solutions of the embodiments of the present application can be applied to the scene of intelligent question answering.

[0099] As described in the background, the question and answer knowledge base queried by the intelligent customer service has the problem of being not accurate enough.

[0100] In view of the technical problem of the inaccuracy of the existing question and answer knowledge base, the embodiments of the present application provide a solution, and the basic idea is: taking historical customer service conversation data or object description information as a data source; using a knowledge recognition model to recognize candidate question data and candidate answer data corresponding to the candidate question data from the data source; querying from the question and answer knowledge base whether there is a candidate question and answer knowledge pair composed of the candidate question data and the candidate answer data to meet the first similarity requirement Target question and answer knowledge pair; the target question and answer knowledge pair is composed of question data and answer data; if the target question and answer knowledge pair does not exist in the question and answer knowledge base, save the candidate question and answer knowledge pair to the question and answer database; if the target question and answer knowledge pair exists in the question and answer knowledge base, use a knowledge evaluation model to evaluate the candidate question and answer knowledge pair and the target question and answer knowledge pair respectively, and in the case that the evaluation score of the candidate question and answer knowledge pair is higher than that of the target question and answer knowledge pair, use the candidate question and answer knowledge pair to update the target question and answer knowledge pair in the question and answer database.

[0101] Among them, the knowledge recognition model is used to automatically extract the question and answer knowledge pair from the historical customer service conversation data or the object description information to save to the question and answer knowledge base, which is more objective and accurate than the manual generation method, thereby improving the richness and comprehensiveness of the question and answer knowledge base. In addition, the extracted candidate question and answer knowledge pair can be compared with the existing question and answer knowledge pair in the question and answer knowledge base, so as to update the question and answer knowledge base by using the question and answer knowledge pair with higher evaluation score, realize the supplement and update of the question and answer knowledge base, and further improve the accuracy of the question and answer knowledge pair in the question and answer knowledge base.

[0102] With reference to the drawings and embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0103] The implementation details of the technical solutions of the embodiments of the present application will be described in detail below.

[0104] Figure 1 A flowchart of an embodiment of a knowledge processing method provided by the present application, the method can include the following steps:

[0105] 101: Use historical customer service conversation data or object description information as a data source.

[0106] The historical customer service conversation data can be real historical conversation data of a user and an artificial customer service.

[0107] Using historical customer service conversation data as a data source can include: in response to a knowledge update instruction, obtaining historical customer service conversation data corresponding to at least one artificial customer service within a predetermined time, and using the historical customer service conversation data corresponding to any artificial customer service as a data source.

[0108] The object description information can refer to product description information, and the product can include articles, software products, or services, etc. The product description information can include product functions, characteristics, usage methods, and technical specifications, etc. The product description information can be in the form of product specification documents, etc.

[0109] Using object description information as a data source can include: in response to a knowledge update instruction, obtaining object description information of at least one object newly published, and using the object description information of any object as a data source.

[0110] Alternatively, in response to an object publishing instruction of a target object, using object description information of the target object as a data source.

[0111] For example, in an e-commerce scenario, the object can refer to a commodity, and the object publishing can refer to publishing a commodity to an e-commerce platform.

[0112] 102: Use a knowledge recognition model to identify candidate question data and candidate answer data corresponding to the candidate question data from the data source.

[0113] 103: Save a candidate question and answer knowledge pair composed of the candidate question data and the candidate answer data to a question and answer database.

[0114] The target question and answer knowledge pair can be composed of question data and answer data.

[0115] In this embodiment, the knowledge recognition model is used to automatically extract the question and answer knowledge pair from the historical customer service conversation data or object description information and save it to the question and answer knowledge base. Compared with the manual generation method, it is more objective and accurate, thereby improving the accuracy of the question and answer knowledge base.

[0116] The saving of the candidate question and answer knowledge pair composed of the candidate question data and the candidate answer data to the question and answer database can include: querying from the question and answer knowledge base whether there is a target question and answer knowledge pair that meets the first similarity requirement with the candidate question and answer knowledge pair composed of the candidate question data and the candidate answer data; if the target question and answer knowledge pair does not exist in the question and answer knowledge base, saving the candidate question and answer knowledge pair to the question and answer database.

[0117] In addition, if the target question and answer knowledge pair exists in the question and answer knowledge base, the knowledge evaluation model is used to evaluate the candidate question and answer knowledge pair and the target question and answer knowledge pair respectively, and in the case that the evaluation score of the candidate question and answer knowledge pair is higher than that of the target question and answer knowledge pair, the candidate question and answer knowledge pair is used to update the target question and answer knowledge pair in the question and answer database.

[0118] For the convenience of understanding, Figure 2 A flowchart of a knowledge processing method of another embodiment provided by the present application is shown, which can include the following steps:

[0119] 201: Taking historical customer service conversation data or object description information as a data source.

[0120] 202: Using a knowledge recognition model to recognize candidate question data and candidate answer data corresponding to the candidate question data from the data source.

[0121] 203: Querying from the question and answer knowledge base whether there is a target question and answer knowledge pair that meets the first similarity requirement with the candidate question and answer knowledge pair composed of the candidate question data and the candidate answer data.

[0122] The target question and answer knowledge pair is composed of question data and answer data.

[0123] The candidate question and answer knowledge pair can be converted into a vector, and the question and answer knowledge pair in the question and answer knowledge base can be converted into a vector. The similarity between the vectors corresponding to the candidate question and answer knowledge pair and the vectors corresponding to the question and answer knowledge pairs in the question and answer knowledge base is calculated to measure the similarity, which can be calculated by cosine similarity or Euclidean distance.

[0124] 204: If the target question and answer knowledge pair does not exist in the question and answer knowledge base, saving the candidate question and answer knowledge pair to the question and answer database.

[0125] 205: If the target question and answer knowledge pair exists in the question and answer knowledge base, the knowledge evaluation model is used to evaluate the candidate question and answer knowledge pair and the target question and answer knowledge pair respectively, and in the case that the evaluation score of the candidate question and answer knowledge pair is higher than that of the target question and answer knowledge pair, the candidate question and answer knowledge pair is used to update the target question and answer knowledge pair in the question and answer database.

[0126] In the embodiment, the knowledge recognition model is used to automatically extract the question and answer knowledge pair from the historical customer service conversation data or the object description information to save to the question and answer knowledge base. Compared with the manual generation mode, the question and answer knowledge base is more objective and accurate, so that the richness and comprehensiveness of the question and answer knowledge base can be improved. In addition, the extracted candidate question and answer knowledge pair can be compared with the existing question and answer knowledge pair in the question and answer knowledge base, so that the question and answer knowledge pair with a higher evaluation score is used to update the question and answer knowledge base, the supplement and update of the question and answer knowledge base are realized, and the accuracy of the question and answer knowledge pair in the question and answer knowledge base is further improved.

[0127] In order to improve the information recognition accuracy, in some embodiments, the knowledge recognition model is used to recognize the candidate question data and the candidate answer data corresponding to the candidate question data from the data source, which can include: performing segmentation processing on the data source to obtain a plurality of data segments; for any data segment, the knowledge recognition model is used to recognize the candidate question data and the candidate answer data corresponding to the candidate question data from the data segment.

[0128] For example, the data source is a product description document, which can be segmented according to the paragraphs in the product description document. For example, the data source is historical customer service conversation data, wherein the historical customer service conversation data can be segmented according to the time interval, such as the time interval between two continuous messages exceeding a preset threshold, the two messages can be regarded as the beginning and end of different conversation paragraphs, or the historical customer service conversation data can be segmented according to specific keywords, or the natural language processing technology can be combined to recognize different conversation topics in the historical customer service conversation data, and the segmentation point is segmented when the conversation topic changes.

[0129] In some embodiments, recognizing the candidate question data and the candidate answer data corresponding to the candidate question data from the data source can include: generating a first prompt instruction based on the data source and the recognition requirement; inputting the first prompt instruction into the knowledge recognition model to instruct the knowledge recognition model to recognize the candidate question data and the candidate answer data corresponding to the candidate question data from the data source according to the recognition requirement.

[0130] In some embodiments, the knowledge recognition model is used to recognize the candidate question data and the candidate answer data corresponding to the candidate question data from the data source, which can include: using the knowledge recognition model to recognize the candidate question data, the candidate knowledge type corresponding to the candidate question data, and the candidate answer data from the data source.

[0131] The above-mentioned identification requirements may also include requirements for identifying knowledge types. Therefore, a first prompt instruction is generated based on the data source and the identification requirements, and the first prompt instruction is input into the knowledge identification model, which can instruct the knowledge identification model to identify candidate question data, candidate knowledge types corresponding to the candidate question data, and candidate answer data from the data source according to the identification requirements.

[0132] For example, the recognition requirements may include requirements on the number of candidate question-answer pairs to be identified, the format of the recognition results, and / or the level of division of knowledge types.

[0133] For example, the first prompt instruction may be: analyze the following content {AAAA} and extract the content according to the recognition requirements.

[0134] Identification requirements:

[0135] - Extract at least 10 question-answer pairs

[0136] -Generate JSON format data with the structure (question, type, answer)

[0137] -Types need to be divided into levels from top to bottom according to content

[0138] - Questions should cover the core information from different paragraphs in the content

[0139] - The answer must accurately correspond to the question and remain complete

[0140] -Type level can be up to 3 levels, separated by " / "

[0141] -Avoid the problem of generating duplicate semantics

[0142] The {AAAA} field can be filled with a data source such as a product description document.

[0143] For example, the candidate question data identified, the candidate knowledge types corresponding to the candidate question data, and the number of candidate answers

[0144] The data can be:

[0145] {

[0146] "question":"What is the maximum range of this product?",

[0147] "category":"Performance Parameters / Battery Life",

[0148] "answer":"Under standard operating conditions, the maximum range can reach 620 kilometers"

[0149] }

[0150] Optionally, the querying, from the question-answer knowledge base, whether there exists the target question-answer knowledge pair satisfying the first similarity requirement with the candidate question-answer knowledge pair composed of the candidate question data and the candidate answer data can include: determining, from the question-answer knowledge base, a question-answer pair corresponding to the candidate knowledge type; and querying, from the question-answer pair corresponding to the candidate knowledge type, whether there exists the target question-answer knowledge pair satisfying the first similarity requirement with the candidate question-answer knowledge pair composed of the candidate question data and the candidate answer data. That is, the querying can be performed according to the candidate knowledge type of the candidate question-answer pair to improve the querying efficiency, for example, whether the candidate knowledge type exists in the question-answer knowledge base can be queried. If the candidate knowledge type does not exist, it is determined that there does not exist the target question-answer knowledge pair satisfying the first similarity requirement with the candidate question-answer knowledge pair. If the candidate knowledge type exists in the question-answer knowledge base, whether there exists the target question-answer knowledge pair satisfying the first similarity requirement with the candidate question-answer knowledge pair under the candidate knowledge type can be queried.

[0151] If the target question-answer knowledge pair does not exist in the question-answer knowledge base, the candidate question-answer knowledge pair can be saved to the question-answer database, which can include: if the target question-answer knowledge pair does not exist in the question-answer knowledge base, saving the candidate question-answer knowledge pair corresponding to the candidate knowledge type to the question-answer database.

[0152] If the target question-answer knowledge pair exists in the question-answer knowledge base, the candidate question-answer knowledge pair and the target question-answer knowledge pair can be respectively evaluated by using the knowledge evaluation model, which can include: if the target question-answer knowledge pair exists in the question-answer knowledge base, evaluating the candidate question-answer knowledge pair and the target question-answer knowledge pair by using the knowledge evaluation model, obtaining an evaluation score of the candidate question-answer knowledge pair, and obtaining an evaluation score of the target question-answer knowledge pair.

[0153] In some embodiments, the candidate question-answer knowledge pair and the target question-answer knowledge pair can be respectively evaluated by using the knowledge evaluation model, which can include: generating a second prompt instruction based on the candidate question-answer knowledge pair and the evaluation rule, inputting the second prompt instruction into the knowledge evaluation model to instruct the knowledge evaluation model to generate the evaluation score of the candidate question-answer knowledge pair according to the evaluation rule; and generating a third prompt instruction based on the target question-answer knowledge pair and the evaluation rule, inputting the third prompt instruction into the knowledge evaluation model to instruct the knowledge evaluation model to generate the evaluation score of the target question-answer knowledge pair according to the evaluation rule.

[0154] For example, the evaluation rule can include scoring the question data, the answer data and the knowledge type of the question-answer knowledge pair from multiple dimensions, for example, can include scoring the accuracy dimension, the integrity dimension and the practicability dimension, different dimensions can correspond to different weights, and the question data, the answer data and the knowledge type correspond to different weights.

[0155] For example, the evaluation rule can be: score the question, the answer and the knowledge type respectively from three dimensions of accuracy (40%), integrity (30%), and practicability (30%), with a total score of 100, and the average of the scores of the question, the answer and the knowledge type is the final score.

[0156] Accuracy (40 points)

[0157] 40-35 points: completely accurate, no factual / logical errors

[0158] 34-25 points: minor errors but negligible

[0159] 24-10 points: key information errors or contradictions

[0160] 9-0 points: serious errors leading to misleading

[0161] Integrity (30 points)

[0162] 30-25 points: covers all key elements, no omissions

[0163] 24-15 points: missing secondary information but the main body is complete

[0164] 14-5 points: core information is missing

[0165] 4-0 points: seriously incomplete

[0166] Practicability (30 points)

[0167] 30-25 points: directly solves user needs, language is natural and easy to understand

[0168] 24-15 points: partially meets the needs, further explanation is needed

[0169] 14-5 points: weakly related to the needs

[0170] 4-0 points: unable to solve the problem

[0171] In some embodiments, if the target question and answer pair does not exist in the question and answer knowledge base, saving the candidate question and answer pair to the question and answer database can include: if the target question and answer pair does not exist in the question and answer knowledge base, sending a first confirmation prompt information to the relevant personnel based on the candidate question and answer pair; in response to a confirmation request or an update request for the candidate question and answer pair, saving the candidate question and answer pair or the updated candidate question and answer pair to the question and answer database.

[0172] The first determination prompt information can include prompt information prompting the related personnel to determine to save the candidate question and answer knowledge pair to the question and answer database and prompt information prompting to save the candidate question and answer knowledge pair to the question and answer database after modification. Further, the candidate question and answer knowledge pair can be saved to the question and answer database in response to a confirmation request for the candidate question and answer knowledge pair. Alternatively, the modified candidate question and answer knowledge pair can be saved to the question and answer database in response to an update request for the candidate question and answer knowledge pair.

[0173] In some embodiments, if the target question and answer knowledge pair exists in the question and answer knowledge base, the candidate question and answer knowledge pair and the target question and answer knowledge pair are respectively evaluated by using the knowledge evaluation model, and in the case that the evaluation score of the candidate question and answer knowledge pair is higher than that of the target question and answer knowledge pair, the target question and answer knowledge pair in the question and answer database can be updated by using the candidate question and answer knowledge pair, which can include: if the target question and answer knowledge pair exists in the question and answer knowledge base, the candidate question and answer knowledge pair and the target question and answer knowledge pair are respectively evaluated by using the knowledge evaluation model; in the case that the evaluation score of the candidate question and answer knowledge pair is higher than that of the target question and answer knowledge pair by a first predetermined score, the target question and answer knowledge pair in the question and answer database is updated by using the candidate question and answer knowledge pair; in the case that the evaluation score of the candidate question and answer knowledge pair is not higher than that of the target question and answer knowledge pair by the first predetermined score, second confirmation prompt information is sent to the related personnel based on the candidate question and answer knowledge pair and the target question and answer knowledge pair; and in response to a selection request for the candidate question and answer knowledge pair, the target question and answer knowledge pair in the question and answer database is updated by using the candidate question and answer knowledge pair.

[0174] The second determination prompt information can include information prompting that the evaluation score of the candidate question and answer knowledge pair is not higher than that of the target question and answer knowledge pair by the first predetermined score, and the related personnel can determine whether to update according to the second determination prompt information. In response to a selection request for the target question and answer knowledge pair by the related personnel, the target question and answer knowledge pair in the question and answer database can be updated by using the candidate question and answer knowledge pair.

[0175] In some embodiments, there can be multiple ways to determine whether there exists a target question and answer knowledge pair that meets the first similarity requirement with the candidate question and answer knowledge pair from the question and answer knowledge base.

[0176] As an optional way, it can be determined whether there exists target question data that meets the first similarity requirement with the candidate question data from the question and answer knowledge base. If so, the target question and answer knowledge pair composed of the target question data and its corresponding target answer data is obtained. That is, if there exists target question data similar to the candidate question data, it is determined that there exists a target question and answer knowledge pair similar to the candidate question and answer knowledge pair.

[0177] The similarity between the vector corresponding to the candidate question data and the vector of the question data in the question and answer knowledge base can be calculated, and the question data with a similarity greater than a first threshold is determined as target question data meeting a first similarity requirement.

[0178] As another optional mode, it can be queried from the question and answer knowledge base whether there is target question data meeting the first similarity requirement with the candidate question data, and target answer data corresponding to the target question data and the candidate answer data meet the second similarity requirement; if there is, a target question and answer knowledge pair composed of the target question data and the target answer data corresponding thereto is obtained. That is, there is target question data similar to the candidate question data, and the target answer data corresponding to the target question data is similar to the candidate answer data, and it is determined that there is a target question and answer knowledge pair similar to the candidate question and answer knowledge pair.

[0179] The similarity between the vector corresponding to the candidate answer data and the vector of the target answer data corresponding to the target question data can be calculated, and if the similarity is greater than a second threshold, the second similarity requirement is met.

[0180] In some embodiments, the method further comprises: if there is target question data meeting the first similarity requirement with the candidate question data, and the target answer data corresponding to the target question data and the candidate question and answer data do not meet the second similarity requirement, sending a third confirmation prompt information to the relevant personnel; and updating the target answer data with the candidate answer data in response to a selection request for the candidate answer data. The third confirmation prompt information can be used to prompt that there is target question data similar to the candidate question data, and the candidate answer data is not similar to the target answer data, and the relevant personnel can send a selection request for the candidate answer data, and the candidate answer data is used as the target answer data corresponding to the target question data.

[0181] Figure 3 A flowchart of an embodiment of a message processing method provided in the present application can include the following steps:

[0182] 301: Obtain a user message sent by a user terminal based on a session connection of a customer service terminal.

[0183] 302: Obtain a historical session message in the session connection.

[0184] 303: Use a message processing model to statistically process the user message and the historical session message to extract target question data.

[0185] 304: Find at least one target answer data matched with the target question data from a question and answer knowledge base.

[0186] The question and answer knowledge base can be updated based on question and answer pairs extracted from the data source by the knowledge recognition model. The data source includes historical customer service conversation data or object description information. The updating method of the question and answer knowledge base can be found in the above Figure 1 or Figure 2 As described in the above embodiments, details are not repeated here.

[0187] 305: generating a reply message corresponding to the user message according to the at least one target answer data.

[0188] In this embodiment, the message processing model can be used to statistically process the user message and the historical conversation message, accurately extract the target question data reflecting the core problem of the user, query the question and answer knowledge base with comprehensive and rich question and answer pairs, find the target answer data that can accurately match the target question data, avoid understanding deviation that may occur in artificial customer service, provide answers that are more in line with user needs in combination with the knowledge base, and improve the accuracy of the reply. This embodiment can realize automatic reply, and does not require artificial customer service to understand user problems one by one, so as to improve the reply efficiency.

[0189] In some embodiments, the message processing model can be used to statistically process the user message and the historical conversation message to extract the target question data, which can include: generating a fourth prompt instruction based on the user message, the historical conversation message, and the extraction requirement; and inputting the fourth prompt instruction into the message processing model to instruct the message processing model to extract the target question data from the user message and the historical conversation message according to the extraction requirement.

[0190] For example, the extraction requirement can be:

[0191] 1. Understand the overall conversation content {Session}, extract a question that can reflect the core content of the overall conversation, and finally output only the question

[0192] 2. Refine the expression: you need to simplify the overall conversation into a concise and clear question, maintain the integrity of the core information, and the extracted question is no more than 20 characters.

[0193] The user message and the historical conversation message can be filled into {Session} to generate the fourth prompt instruction.

[0194] In addition, the extraction requirement can also require the message processing model to play the role of the user, so that the extracted target question data is a question that meets the user role.

[0195] In some embodiments, generating a reply message corresponding to the user message according to the at least one target answer data can include: using the answer generation model to generate a reply message corresponding to the user message according to the at least one target answer data.

[0196] The fifth prompt instruction can be generated based on the at least one target response data and the generation requirement; and the fifth prompt instruction is input into the response generation model to instruct the response generation model to generate a reply message corresponding to the user message based on the at least one target response data according to the generation requirement.

[0197] For example, the generation requirement can include:

[0198] 1. Do not ask if there are other questions

[0199] 2. Do not reply to the answer that has been said

[0200] 3. Your role setting is {Role}, you should act as such a role, and give objective, accurate and natural replies according to {Answer}.

[0201] The role setting details can be filled into {Role}, and the matching degree corresponding to the at least one target response data can be filled into {Answer}, to generate the fifth prompt instruction.

[0202] For example, the role setting details can be "a 25-year-old Libra woman, 168 cm tall, 55 kg, blood type B, working as a customer service teacher in a large company, full of enthusiasm for work, willing to help others, good at listening, and always keeping friendly and patient when communicating with customers. Hometown in Guangdong, can speak fluent Mandarin and Cantonese. Understand some English, can communicate with English users simply".

[0203] The response generation model and the above-mentioned message processing model can be the same model or different models.

[0204] The knowledge recognition model, the knowledge evaluation model, the message processing model and the response generation model involved in the embodiments of the present application can be a language model (LM) or a multi-modal model (LM) based on artificial intelligence, and the present application does not limit the number of model parameters supported by the model, and the target is to meet the actual demand.

[0205] The prompt instruction (prompt) involved in the embodiments of the present application is a natural language input, similar to a command or instruction, to let the model know what it needs to do. It can be a question, a sentence or a paragraph, which is used to provide context and parameter information for the input information of the large model, and to prompt or guide the large model to give the expected output.

[0206] In some embodiments, finding the target answer data matching the target question data from the question-answer knowledge base can include: finding a plurality of candidate question data satisfying the third similarity requirement from the question-answer knowledge base based on the target question data; calculating a plurality of matching degrees of the plurality of candidate answer data corresponding to the plurality of candidate question data and the target question data respectively, and determining at least one target answer data satisfying the matching requirement based on the matching degrees.

[0207] The similarity between the vector corresponding to the target question data and the vector corresponding to the question data in the question-answer knowledge base can be calculated, and the question data with a similarity greater than a third threshold value can be used as the candidate question data satisfying the third similarity requirement.

[0208] The matching degree of the plurality of candidate answer data corresponding to the plurality of candidate question data and the target question data can be calculated using a re-ranking model, and the plurality of candidate answer data can be sorted based on the matching degrees. The plurality of candidate answer data and the target question data can be input into the re-ranking model, the plurality of candidate answer data and the target question data can be deeply semantically encoded, semantic features can be extracted, the semantic features of the target question data and the plurality of candidate answer data can be interacted through a cross-attention mechanism (Cross-Attention), the semantic association between the two can be captured, based on the features after the interaction, the matching scores of the target question data and the plurality of candidate answer data can be obtained through a multi-layer perception (MLP) or a classification head, and the plurality of candidate answer data can be sorted based on the matching scores. The candidate answer data ranked in the top of the predetermined number can be used as the at least one target answer data.

[0209] The re-ranking model can be trained based on sample query data, data highly related to the sample query data in semantics as positive samples, and data not related to the sample query data in semantics as negative samples.

[0210] In some embodiments, the customer service end can include a customer service robot.

[0211] The user message sent by the user end based on the session connection with the customer service end can include: in response to the customer service request sent by the user end, establishing a session connection between the user end and the customer service robot; and obtaining the user message sent by the user end to the customer service robot based on the session connection with the customer service end.

[0212] The method can further include: sending the reply message to the user end. The reply message can be sent to the customer service robot, and the customer service robot can send the reply message to the user end.

[0213] In some embodiments, the method can further include: in response to the customer service request sent by the user end, establishing a session connection between the user end and the customer service end;

[0214] The method may further include: sending recommendation prompt information to the client terminal based on the reply message; and sending a reply message to the user terminal in response to the recommendation confirmation request sent by the client terminal.

[0215] Among them, the recommendation prompt information can be used to prompt the manual customer service of the customer service end whether to send a reply message to the user end, and the recommendation confirmation request can be sent by the manual customer service of the customer service end.

[0216] In order to deeply understand user intentions and further improve the accuracy of extracting target question data so that the target question data can accurately reflect the user's core needs, the above-mentioned use of the message processing model to count user messages and historical conversation messages to extract target question data may include: using the message processing model, combined with user feature data, to count user messages and historical conversation messages to extract target question data.

[0217] For example, a user message might inquire about how to use a certain cosmetic product. User profile data might include past purchases, including skincare product purchases. The target question data extracted using the message processing model might include how to use the cosmetic product with other skincare products. Reply messages generated based on this target question data can improve the practicality of responses and enhance the user experience.

[0218] Among them, the above-mentioned fourth prompt instruction can be generated based on user feature data, user messages, historical conversation messages, and extraction requirements; the fourth prompt instruction is input into the message processing model to instruct the message processing model to extract target question data from user feature data, user messages and historical conversation messages according to the extraction requirements.

[0219] For ease of understanding, Figure 4 This is a flowchart of a knowledge processing method according to another embodiment of the present application. The method may include the following steps:

[0220] 401: Gets user messages sent by the client based on the session connection with the client.

[0221] 402: Determine user feature data corresponding to the user terminal.

[0222] User feature data can include user behavior data and user attribute data. User attribute data can include the user's region, age, gender, and registration time. User behavior data can refer to user interaction behavior data. For example, in e-commerce scenarios, user behavior data can include browsing behavior data, add-to-cart behavior data, search behavior data, and purchase behavior data.

[0223] 403: Get the historical session message corresponding to the user message.

[0224] 404: Use the message processing model to count the user message and the historical conversation message in combination with the user feature data to extract the target question data.

[0225] The specific implementation of extracting the target question data can be referred to Figure 3 In the embodiment shown, details are not described here.

[0226] 405: Find at least one target answer data matched with the target question data from the question and answer knowledge base.

[0227] 406: Generate a reply message corresponding to the user message according to the at least one target answer data.

[0228] In this embodiment, the user feature data is combined to help the message processing model deeply understand the user's intention, further improve the accuracy of extracting the target question data, so that the target question data can reflect the user's core problem, and then query the question and answer knowledge base with comprehensive and rich question and answer pairs, and find the target answer data that can accurately match the target question data. Combined with the knowledge base, a more user-demand-oriented answer can be provided, avoiding the understanding deviation that may occur in artificial customer service and improving the accuracy of the reply. And this embodiment can realize automatic reply, without artificial customer service to understand user problems one by one, which can improve the reply efficiency.

[0229] In some embodiments, the customer service end can include a customer service robot.

[0230] Obtaining the user message sent by the user end based on the conversation connection with the customer service end can include: in response to the customer service request sent by the user end, establishing a conversation connection between the user end and the customer service robot; and obtaining the user message sent by the user end to the customer service robot based on the conversation connection with the customer service end.

[0231] The method can further include: sending the reply message to the user end. Wherein the reply message can be sent to the customer service robot, and the customer service robot sends the reply message to the user end.

[0232] In some embodiments, the method can further include: in response to the customer service request sent by the user end, establishing a conversation connection between the user end and the customer service end;

[0233] The method can further include: based on the reply message, sending recommendation prompt information to the customer service end; and in response to the recommendation confirmation request sent by the customer service end, sending the reply message to the user end.

[0234] Wherein, the recommendation prompt information can be used to prompt the artificial customer service of the customer service end whether to send the reply message to the user end, and the recommendation confirmation request can be sent by the artificial customer service of the customer service end.

[0235] In order to facilitate understanding, Figure 5A scene interaction schematic diagram in an actual application of an embodiment of the present application is shown. As shown in Figure 5 The user terminal 501 and the customer service terminal 502 can respectively establish a connection with the service terminal 503 through a network.

[0236] The user terminal 501 or the customer service terminal 502 can be a browser, an APP (Application), or a web application such as an H5 (HyperText Markup Language 5) application, or a light application (also known as a small program, a lightweight application program), or a cloud application, etc. The user terminal 501 or the customer service terminal 502 can be deployed in an electronic device and needs to depend on the device or some app in the device for running, etc. The electronic device can have a display screen and support information browsing, etc., such as a personal mobile terminal such as a mobile phone, a tablet computer, a personal computer, a desktop computer, a smart speaker, a smart watch, etc. For ease of understanding, Figure 5 The user terminal 501 or the customer service terminal 502 is mainly represented by a device image in the present application. Various other types of applications can also be configured in the electronic device, such as man-machine dialogue applications, model training applications, text processing applications, web browser applications, shopping applications, search applications, instant messaging tools, mailbox clients, social platform software, etc. The electronic device can be a device used by a user and having functions of computing, networking, and communication required by the user, such as a mobile phone, a tablet computer, a personal computer, a wearable device, etc. The electronic device can generally include at least one processing component and at least one storage component. The electronic device can also include a network card chip, an IO bus, an audio and video component, etc. The present application does not limit this. Optionally, according to the implementation form of the electronic device, some peripheral devices can also be included, such as a keyboard, a mouse, an input pen, a printer, etc. The present application does not limit this.

[0237] The service terminal 503 can include servers providing various services, such as servers supporting model training, servers processing information sent by the user terminal, etc. It should be noted that the service terminal 503 can be implemented as a distributed server cluster composed of multiple servers or as a single server. The server can also be a server of a distributed system or a server combined with a blockchain. The server can also be a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content distribution networks (CDNs), and basic cloud computing services of big data and artificial intelligence platforms, etc. or an intelligent cloud computing server or intelligent cloud host with artificial intelligence technology.

[0238] The customer service end 502 can include a customer service robot, the service end 503 can establish a session connection between the user end 501 and the customer service robot in response to a customer service request sent by the user end 501, the user end 501 can send a user message to the customer service robot based on the session connection, the customer service robot can send the user message to the service end 503,

[0239] The service end 503 can determine user feature data corresponding to the user end 501, obtain historical session messages corresponding to the user message, use a message processing model 504, combine the user feature data, and statistically process the user message and the historical session messages to extract target problem data, find at least one target response data matched with the target problem data from a question and answer knowledge base 505, and use a response generation model 506 to generate a reply message corresponding to the user message according to the at least one target response data. The service end 503 can send the reply message to the customer service robot, and the customer service robot can send the reply message to the user end 501.

[0240] In this embodiment, the service end combines the user feature data to extract the target problem data, which helps the message processing model to deeply understand the user intent and further improves the accuracy of extracting the target problem data, so that the target problem data can reflect the core problem of the user, and then the question and answer knowledge base with comprehensive and rich question and answer pairs can be queried to find the target response data that can accurately match the target problem data. Combined with the knowledge base, a more user-demand-oriented answer can be provided, avoiding the understanding deviation that may occur in artificial customer service and improving the accuracy of the reply. Moreover, this embodiment can realize automatic reply, without the need for artificial customer service to understand user problems one by one, and can improve the reply efficiency.

[0241] It should be noted that the technical solutions of the embodiments of the present application are applicable to a network virtual environment, and the described user generally refers to a "virtual user". A real user can register a user account in the service end by registration to obtain a user identity in the network environment. The same user account can be logged in to the service end through different types of user ends, so that the service end can identify the same user.

[0242] The interaction operation between the service end and the user can be realized based on the user account, and the corresponding data received or sent by the service end to the user is also realized based on the user account, which is actually received or sent by the user end corresponding to the user account to the service end. In addition, the users can also realize communication and the like through the user account. Among them, the user can refer to an individual, or an institution such as an enterprise, and the present application does not make specific limitations.

[0243] The detailed implementation and beneficial effects of each step in the method of the embodiment have been described in detail in the foregoing embodiments, and will not be described in detail here.

[0244] It should be noted that in some of the processes described in the above embodiments and the accompanying drawings, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or may be executed in parallel. The sequence numbers of the operations, such as 401, 402, etc., are only used to distinguish between different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0245] Figure 6 A schematic diagram of the structure of an embodiment of a knowledge processing device provided in an embodiment of the present application, the device comprising:

[0246] The first determining module 601 is configured to use historical customer service session data or object description information as a data source.

[0247] The first recognition module 602 is used to use the knowledge recognition model to identify candidate question data and candidate answer data corresponding to the candidate question data from the data source.

[0248] The first saving module 603 is used to save the candidate question-answer knowledge pairs consisting of the candidate question data and the candidate answer data into the question-answer database.

[0249] The target question-answer knowledge pair may be composed of question data and answer data.

[0250] Figure 6 The knowledge processing device can execute Figure 1 The implementation principle and technical effects of the knowledge processing method described in the illustrated embodiment will not be described in detail. The specific manner in which each module and unit performs operations in the knowledge processing device in the above embodiment has been described in detail in the embodiment of the method and will not be elaborated here.

[0251] Figure 7 A schematic diagram of the structure of an embodiment of a knowledge processing device provided in an embodiment of the present application, the device comprising:

[0252] The second determining module 701 is configured to use historical customer service session data or object description information as a data source;

[0253] The second identification module 702 is used to identify candidate question data and candidate answer data corresponding to the candidate question data from the data source using the knowledge identification model;

[0254] The query module 703 is configured to query, from the question and answer knowledge base, whether there is a target question and answer knowledge pair that meets the first similarity requirement with the candidate question and answer knowledge pair composed of the candidate question data and the candidate answer data.

[0255] The second saving module 704 is configured to save the candidate question and answer knowledge pair into the question and answer database if the target question and answer knowledge pair does not exist in the question and answer knowledge base.

[0256] The evaluation module 704 is configured to evaluate the candidate question and answer knowledge pair and the target question and answer knowledge pair by using a knowledge evaluation model if the target question and answer knowledge pair exists in the question and answer knowledge base, and update the target question and answer knowledge pair in the question and answer database by using the candidate question and answer knowledge pair if the evaluation score of the candidate question and answer knowledge pair is higher than that of the target question and answer knowledge pair.

[0257] In some embodiments, the second identification module can identify the candidate question data and the candidate answer data corresponding to the candidate question data from the data source by using a knowledge identification model, which can include: performing segmentation processing on the data source to obtain a plurality of data segments; and identifying the candidate question data and the candidate answer data corresponding to the candidate question data from the data segment by using the knowledge identification model for any data segment.

[0258] In some embodiments, identifying the candidate question data and the candidate answer data corresponding to the candidate question data from the data source can include: generating a first prompt instruction based on the data source and an identification requirement; and inputting the first prompt instruction into the knowledge identification model to instruct the knowledge identification model to identify the candidate question data and the candidate answer data corresponding to the candidate question data from the data source according to the identification requirement.

[0259] In some embodiments, the query module can query, from the question and answer knowledge base, whether there is a target question and answer knowledge pair that meets the first similarity requirement with the candidate question and answer knowledge pair composed of the candidate question data and the candidate answer data, which can include: determining a question and answer pair corresponding to the candidate knowledge type from the question and answer knowledge base; and querying, from the question and answer pair corresponding to the candidate knowledge type, whether there is a target question and answer knowledge pair that meets the first similarity requirement with the candidate question and answer knowledge pair composed of the candidate question data and the candidate answer data.

[0260] If the target question and answer knowledge pair does not exist in the question and answer knowledge base, the second saving module can save the candidate question and answer knowledge pair into the question and answer database, which can include: saving the candidate question and answer knowledge pair corresponding to the candidate knowledge type into the question and answer database if the target question and answer knowledge pair does not exist in the question and answer knowledge base.

[0261] If the target question and answer knowledge pair exists in the question and answer knowledge base, the evaluation module respectively evaluates the candidate question and answer knowledge pair and the target question and answer knowledge pair using the knowledge evaluation model can include: if the target question and answer knowledge pair exists in the question and answer knowledge base, evaluating the candidate question and answer knowledge pair and its candidate knowledge type using the knowledge evaluation model to obtain an evaluation score of the candidate knowledge question and answer pair, and evaluating the target question and answer knowledge pair and its target knowledge type to obtain an evaluation score of the target knowledge question and answer pair.

[0262] In some embodiments, the evaluation module respectively evaluates the candidate question and answer knowledge pair and the target question and answer knowledge pair using the knowledge evaluation model can include: generating a second prompt instruction based on the candidate question and answer knowledge pair and the evaluation rule, inputting the second prompt instruction into the knowledge evaluation model to instruct the knowledge evaluation model to generate an evaluation score of the candidate question and answer knowledge pair according to the evaluation rule; generating a third prompt instruction based on the target question and answer knowledge pair and the evaluation rule, and inputting the third prompt instruction into the knowledge evaluation model to instruct the knowledge evaluation model to generate an evaluation score of the target question and answer knowledge pair according to the evaluation rule.

[0263] In some embodiments, if the target question and answer knowledge pair does not exist in the question and answer knowledge base, the saving module saves the candidate question and answer knowledge pair into the question and answer database can include: if the target question and answer knowledge pair does not exist in the question and answer knowledge base, sending a first confirmation prompt information to the relevant personnel based on the candidate question and answer knowledge; in response to a confirmation request or an update request for the candidate question and answer knowledge pair, saving the candidate question and answer knowledge pair or the updated candidate question and answer knowledge pair into the question and answer database.

[0264] In some embodiments, if the target question and answer knowledge pair exists in the question and answer knowledge base, the evaluation module respectively evaluates the candidate question and answer knowledge pair and the target question and answer knowledge pair using the knowledge evaluation model, and in the case that the evaluation score of the candidate question and answer knowledge pair is higher than that of the target question and answer knowledge pair, the candidate question and answer knowledge pair is used to update the target question and answer knowledge pair in the question and answer database can include: if the target question and answer knowledge pair exists in the question and answer knowledge base, the evaluation module respectively evaluates the candidate question and answer knowledge pair and the target question and answer knowledge pair using the knowledge evaluation model; in the case that the evaluation score of the candidate question and answer knowledge pair is higher than that of the target question and answer knowledge pair by a first predetermined score, the candidate question and answer knowledge pair is used to update the target question and answer knowledge pair in the question and answer database; in the case that the evaluation score of the candidate question and answer knowledge pair is not higher than that of the target question and answer knowledge pair by the first predetermined score, a second confirmation prompt information is sent to the relevant personnel based on the candidate question and answer knowledge pair and the target question and answer knowledge pair; in response to a selection request for the candidate question and answer knowledge pair, the candidate question and answer knowledge pair is used to update the target question and answer knowledge pair in the question and answer database.

[0265] In some embodiments, the query module queries from the question and answer knowledge base whether there exists a target question and answer knowledge pair that meets the first similarity requirement with the candidate question and answer knowledge pair.

[0266] As an optional way, the query module can query from the question and answer knowledge base whether there exists target question data that meets the first similarity requirement with the candidate question data, and if so, obtain a target question and answer knowledge pair composed of the target question data and its corresponding target answer data. That is, if there exists target question data similar to the candidate question data, it is determined that there exists a target question and answer knowledge pair similar to the candidate question and answer knowledge pair.

[0267] As another optional way, the query module can query from the question and answer knowledge base whether there exists target question data that meets the first similarity requirement with the candidate question data, and the candidate answer data meets the second similarity requirement with the target answer data corresponding to the target question data; if so, obtain a target question and answer knowledge pair composed of the target question data and its corresponding target answer data. That is, if there exists target question data similar to the candidate question data, and the target answer data corresponding to the target question data is similar to the candidate answer data, it is determined that there exists a target question and answer knowledge pair similar to the candidate question and answer knowledge pair.

[0268] In some embodiments, the device can further include:

[0269] The prompt updating module is configured to send a third confirmation prompt information to the relevant personnel if there exists target question data that meets the first similarity requirement with the candidate question data, and the target answer data corresponding to the target question data does not meet the second similarity requirement with the candidate answer data; and update the target answer data with the candidate answer data in response to a selection request for the candidate answer data.

[0270] Figure 7 The knowledge processing device can perform Figure 2 The implementation principle and technical effects of the knowledge processing method of the embodiments are not described again. The specific operation modes of each module and unit of the knowledge processing device in the above embodiments have been described in detail in the embodiments related to the method, and will not be described in detail here.

[0271] Figure 8 An embodiment of a message processing device provided by the embodiments of the present application has a structure schematic diagram, and the device includes:

[0272] The first acquisition module 801 is configured to acquire a user message sent by a user terminal based on a session connection of a customer service terminal.

[0273] The second acquisition module 802 is configured to acquire a historical session message in the session connection.

[0274] The first extraction module 803 is configured to use the message processing model to perform statistics on the user message and the historical session message, so as to extract the target question data.

[0275] The question and answer knowledge base can be updated based on the question and answer pairs extracted from the data source by the knowledge recognition model. The data source includes historical customer service session data or object description information.

[0276] The first search module 804 is configured to search at least one target answer data matched with the target question data from the question and answer knowledge base.

[0277] The first generation module 805 is configured to generate a reply message corresponding to the user message according to the at least one target answer data.

[0278] In some embodiments, the first extraction module uses the message processing model to perform statistics on the user message and the historical session message, so as to extract the target question data can include: generating a fourth prompt instruction based on the user message, the historical session message and the extraction requirement; inputting the fourth prompt instruction into the message processing model to instruct the message processing model to extract the target question data from the user message and the historical session message according to the extraction requirement.

[0279] In some embodiments, the first generation module generates a reply message corresponding to the user message based on the at least one target answer data can include: using an answer generation model to generate a reply message corresponding to the user message according to the at least one target answer data.

[0280] In some embodiments, the first generation module generates a reply message corresponding to the user message based on the at least one target answer data can include: using an answer generation model to generate a reply message corresponding to the user message according to the at least one target answer data.

[0281] In some embodiments, the first search module searches the target answer data matched with the target question data from the question and answer knowledge base can include: searching a plurality of candidate question data satisfying the third similarity requirement from the question and answer knowledge base; calculating the matching degrees of a plurality of candidate answer data corresponding to the plurality of candidate question data with the target question data respectively, and determining at least one target answer data satisfying the matching requirement.

[0282] In some embodiments, the customer service end can include a customer service robot.

[0283] The first acquisition module acquires the user message sent by the user end based on the session connection with the customer service end can include: establishing a session connection between the user end and the customer service robot in response to the customer service request sent by the user end; and acquiring the user message sent by the user end to the customer service robot based on the session connection with the customer service end.

[0284] The apparatus can further include:

[0285] The first sending module is configured to send the reply message to the user terminal.

[0286] In some embodiments, the apparatus can further include:

[0287] The first connection establishing module is configured to establish a session connection between the user terminal and the service terminal in response to the service request sent by the user terminal.

[0288] The first sending module is further configured to send recommendation prompt information to the service terminal based on the reply message, and send the reply message to the user terminal in response to a recommendation confirmation request sent by the service terminal.

[0289] The recommendation prompt information can be used to prompt the manual service of the service terminal whether to send the reply message to the user terminal, and the recommendation confirmation request can be sent by the manual service of the service terminal.

[0290] Figure 8 The message processing apparatus can perform Figure 3 The message processing method of the embodiments described above has the same implementation principles and technical effects. The specific operation modes of each module and unit of the message processing apparatus in the above embodiments have been described in detail in the embodiments related to the method, and will not be described in detail here.

[0291] Figure 9 An embodiment of a message processing apparatus provided by the embodiments of the present application has a structure as shown in the structural schematic diagram. The apparatus includes:

[0292] The third acquisition module 901 is configured to acquire a user message sent by a user terminal based on a session connection with a service terminal.

[0293] The third determination module 902 is configured to determine user feature data corresponding to the user terminal.

[0294] The fourth acquisition module 903 is configured to acquire historical session messages corresponding to the user message.

[0295] The second extraction module 904 is configured to use a message processing model to combine the user feature data to statistically process the user message and the historical session messages, so as to extract target question data.

[0296] The second search module 905 is configured to search at least one target answer data matched with the target question data from a question and answer knowledge base.

[0297] The second generation module 906 is configured to generate a reply message corresponding to the user message according to the at least one target answer data.

[0298] In some embodiments, the customer service end can include a customer service robot.

[0299] The second obtaining module can obtain the user message sent by the user end based on the session connection with the customer service end, and the method can include: establishing a session connection between the user end and the customer service robot in response to a customer service request sent by the user end; and obtaining the user message sent by the user end to the customer service robot based on the session connection with the customer service end.

[0300] The device can further include a second sending module configured to send the reply message to the user end.

[0301] In some embodiments, the device can further include a second connection establishing module configured to establish a session connection between the user end and the customer service end in response to a customer service request sent by the user end.

[0302] The second sending module can be further configured to: send recommendation prompt information to the customer service end based on the reply message; and send the reply message to the user end in response to a recommendation confirmation request sent by the customer service end.

[0303] Figure 9 The message processing device can perform the message processing method as described in the embodiments. Figure 4 The message processing method as described in the embodiments has the implementation principle and technical effects which will not be repeated here. The specific operation manner of each module and unit in the message processing device in the above embodiments has been described in detail in the embodiments related to the method, and will not be described in detail here.

[0304] Figure 10 An embodiment of a structure schematic diagram of a computing device provided in the present application is shown. As shown in the figure, Figure 10 In practice, the computing device can include a storage component 1001 and a processing component 1002.

[0305] The storage component 1001 is configured to store computer programs and can be configured to store other various data to support operations on the computing device. Examples of these data include instructions of any application program or method for operating on the computing device, data structures, contact data, phonebook data, messages, pictures, videos, etc.

[0306] The processing component 1002 is coupled to the storage component 1001 and is configured to execute the computer programs in the storage component 1001, so as to implement the knowledge processing method as described in the embodiments or Figure 1 the knowledge processing method as described in the embodiments or Figure 2 the message processing method as described in the embodiments or Figure 3 the message processing method as described in the embodiments or Figure 4 the message processing method as described in the embodiments.

[0307] Further, as shown in the figure, Figure 10As shown, the computing device also includes a communication component 1003, a display component 1004, a power supply component 1005, an audio component 1006, and / or other components. Figure 10 Some of the components are shown as being within the computing device, but it is understood that these components can be located within other computing devices that are coupled to the computing device. Figure 10 As shown, the computing device also includes a communication component 1003, a display component 1004, a power supply component 1005, an audio component 1006, and / or other components. Figure 10 The components within the dashed line are optional components that are not required for the computing device, and the computing device can include more or different components depending on the product form of the computing device. The computing device of the present embodiment can be implemented as a terminal device such as a desktop computer, a notebook computer, a smart phone, or an IOT (Internet of Things) device, or as a server device such as a general server, a cloud server, or a server array. If the computing device of the present embodiment is implemented as a terminal device such as a desktop computer, a notebook computer, or a smart phone, the computing device can include the components within the dashed line. Figure 10 If the computing device of the present embodiment is implemented as a server device such as a general server, a cloud server, or a server array, the computing device can not include the components within the dashed line. Figure 10 If the computing device of the present embodiment is implemented as a server device such as a general server, a cloud server, or a server array, the computing device can not include the components within the dashed line.

[0308] The processing component described above can include one or more processors to execute computer instructions to complete all or part of the steps of the methods described above. Of course, the processing component can also be one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components, for executing the methods described above.

[0309] The storage component described above can be implemented by any type of volatile or non-volatile storage devices or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable programmable read only memory (EPROM), programmable read only memory (PROM), read only memory (ROM), magnetic storage, flash memory, a magnetic disk or a compact disk.

[0310] The communication component is configured to facilitate wired or wireless communication between the device on which the communication component is installed and other devices. The device on which the communication component is installed can access a wireless network based on a communication standard, such as a radio communication technology, a wireless local area network (WLAN) technology, a Bluetooth (BT) technology, a near field communication (NFC) technology, and / or a global positioning system (GPS) technology. In an exemplary embodiment, the communication component receives a broadcast signal or broadcast related information from an external broadcast management system via a broadcast channel.

[0311] The display component can include a screen, which can include a Liquid Crystal Display (LCD) and a Touch Panel (TP). If the screen includes a touch panel, the screen can be implemented as a touch screen to receive an input signal from a user. The touch panel includes one or more touch sensors to sense a touch, a slide and a gesture on the touch panel. The touch sensor can not only sense a boundary of a touching or a sliding action, but also detect duration and pressure related to the touching or sliding action.

[0312] The power supply component provides power to various components of the device on which the power supply component is installed. The power supply component can include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for the device on which the power supply component is installed.

[0313] The audio component can be configured to output and / or input audio signals. For example, the audio component includes a microphone (MIC) that is configured to receive an external audio signal when the device on which the audio component is installed is in a particular mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signal can be further stored in a memory or transmitted via the communication component. In some embodiments, the audio component also includes a speaker for outputting audio signals.

[0314] Accordingly, the embodiments of the present application also provide a computer readable storage medium storing a computer program, when the computer program is executed by a processor, the processor is enabled to implement each step in the above method embodiments. The computer readable storage medium includes volatile or non-volatile or their combination, and can be removable or non-removable. Examples of the computer readable storage medium include, but are not limited to, phase-change random access memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital video disc (DVD) or other optical storage, magnetic cassette, magnetic tape disk storage or other magnetic storage devices, or any other non-transmission medium

[0315] Accordingly, the embodiments of the present application also provide a computer program product, the computer program product includes computer programs or instructions, when the computer programs or instructions are executed by a processor, the processor is enabled to implement each step in the above method embodiments. It should be understood that each process or combination of multiple processes in the above method flow can be implemented by the computer programs or instructions. In addition, these computer programs or instructions can be applied to the processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device, so that the processor of the general-purpose computer, the special-purpose computer, the embedded processor or other programmable data processing device can be implemented as a device for implementing the corresponding functions in the above method embodiments.

[0316] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the above method embodiments, which will not be described here.

[0317] It should also be noted that the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method or device. Without more limitations, the element defined by the statement "including a" does not exclude the presence of additional same elements in the process, method or device including the element.

[0318] Finally, it should be noted that the above merely illustrates the embodiments of the present application, and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the scope of claims of the present application.

Claims

1. A knowledge processing method, characterized in that: include: Use historical customer service session data or object description information as the data source; Using a knowledge recognition model, identifying candidate question data and candidate answer data corresponding to the candidate question data from the data source; From the question-answer knowledge base, query whether there is a target question-answer knowledge pair that satisfies a first similarity requirement with the candidate question data and the candidate answer data; The target question-answer knowledge pair consists of question data and answer data; If the target question-answer knowledge pair does not exist in the question-answer knowledge base, saving the candidate question-answer knowledge pair to the question-answer database; If the target question-answer knowledge pair exists in the question-answer knowledge base, the candidate question-answer knowledge pair and the target question-answer knowledge pair are evaluated separately using the knowledge evaluation model, and if the evaluation score of the candidate question-answer knowledge pair is higher than that of the target question-answer knowledge pair, the candidate question-answer knowledge pair is used to update the target question-answer knowledge pair in the question-answer database.

2. The method according to claim 1, characterized in that The step of using the knowledge recognition model to identify candidate question data and candidate answer data corresponding to the candidate question data from the data source includes: Using a knowledge recognition model, identifying candidate question data, candidate knowledge types corresponding to the candidate question data, and candidate answer data from the data source; If the target question-answer knowledge pair does not exist in the question-answer knowledge base, saving the candidate question-answer knowledge pair to the question-answer database includes: If the target question-answer knowledge pair does not exist in the question-answer knowledge base, saving the candidate question-answer knowledge pair corresponding to the candidate knowledge type in the question-answer database; If the target question-answer knowledge pair exists in the question-answer knowledge base, using the knowledge evaluation model to evaluate the candidate question-answer knowledge pair and the target question-answer knowledge pair respectively includes: If the target question-answer knowledge pair exists in the question-answer knowledge base, the knowledge evaluation model is used to evaluate the candidate question-answer knowledge pair and its candidate knowledge type to obtain the evaluation score of the candidate knowledge question-answer pair, and the target question-answer knowledge pair and its target knowledge type are evaluated to obtain the evaluation score of the target knowledge question-answer pair.

3. The method according to claim 1, characterized in that The step of using the knowledge recognition model to identify candidate question data and candidate answer data corresponding to the candidate question data from the data source includes: Using a knowledge recognition model, identifying candidate question data, candidate knowledge types corresponding to the candidate question data, and candidate answer data from the data source; The querying of the question-answer knowledge base to determine whether there is a target question-answer knowledge pair that satisfies the first similarity requirement with the candidate question data and the candidate answer data comprises: Determining question-answer pairs corresponding to the candidate knowledge type from a question-answer knowledge base; From the question-answer pairs corresponding to the candidate knowledge types, it is queried whether there is a target question-answer knowledge pair that satisfies a first similarity requirement with the candidate question data and the candidate answer data.

4. The method according to claim 1, wherein The step of using the knowledge recognition model to identify candidate question data and candidate answer data corresponding to the candidate question data from the data source includes: Segmenting the data source to obtain multiple data segments; For any data segment, the knowledge recognition model is used to identify candidate question data and candidate answer data corresponding to the candidate question data from the data segment.

5. The method according to claim 1, wherein If the target question-answer knowledge pair does not exist in the question-answer knowledge base, saving the candidate question-answer knowledge pair to the question-answer database includes: If the target question-answer knowledge pair does not exist in the question-answer knowledge base, sending a first confirmation prompt message to a relevant person based on the candidate question-answer knowledge pair; In response to a confirmation request or an update request for the candidate question-answer knowledge pair, the candidate question-answer knowledge pair or the updated candidate question-answer knowledge pair is saved in the question-answer database.

6. The method according to claim 1, characterized in that If the target question-answer knowledge pair exists in the question-answer knowledge base, using a knowledge evaluation model to evaluate the candidate question-answer knowledge pair and the target question-answer knowledge pair respectively, and if the evaluation score of the candidate question-answer knowledge pair is higher than that of the target question-answer knowledge pair, using the candidate question-answer knowledge pair to update the target question-answer knowledge pair in the question-answer database includes: If the target question-answer knowledge pair exists in the question-answer knowledge base, using a knowledge evaluation model to evaluate the candidate question-answer knowledge pair and the target question-answer knowledge pair respectively; When the evaluation score of the candidate question-answer knowledge pair is higher than a first predetermined score of the target question-answer knowledge pair, updating the target question-answer knowledge pair in the question-answer database with the candidate question-answer knowledge pair; If the evaluation score of the candidate question-answer knowledge pair is not higher than the first predetermined score of the target question-answer knowledge pair, sending a second confirmation prompt message to the relevant personnel based on the candidate question-answer knowledge pair and the target question-answer knowledge pair; In response to a selection request for the candidate question-answer knowledge pair, the target question-answer knowledge pair in the question-answer database is updated using the candidate question-answer knowledge pair.

7. The method according to claim 1, characterized in that The querying of the question-answer knowledge base to determine whether there is a target question-answer knowledge pair that satisfies the first similarity requirement with the candidate question data and the candidate answer data comprises: From the question-answer knowledge base, query whether there is target question data that meets the first similarity requirement with the candidate question data; if so, obtain a target question-answer knowledge pair consisting of the target question data and its corresponding target answer data; Alternatively, query the question and answer knowledge base to see whether there is target question data that meets the first similarity requirement with the candidate question data, and the candidate answer data and the target answer data corresponding to the target question data meet the second similarity requirement; if so, obtain a target question and answer knowledge pair consisting of the target question data and its corresponding target answer data.

8. The method according to claim 7, characterized in that Also includes: If there is target question data that meets the first similarity requirement with the candidate question data, and the candidate question and answer data and the target answer data corresponding to the target question data do not meet the second similarity requirement, a third confirmation prompt message is sent to the relevant personnel; In response to a selection request for the candidate response data, the target response data is updated using the candidate response data.

9. The method according to claim 1, characterized in that The use of historical customer service session data or object description information as a data source includes: In response to the knowledge update instruction, obtaining historical customer service session data corresponding to at least one manual customer service representative within a predetermined time period, and using the historical customer service session data corresponding to any one manual customer service representative as a data source; Alternatively, in response to the knowledge update instruction, obtaining the object description information of at least one object that is recently published, and using the object description information of any one object as a data source; Alternatively, in response to an object publishing instruction of a target object, the object description information of the target object is used as a data source.

10. The method according to claim 1, characterized in that The step of using the knowledge recognition model to identify candidate question data and candidate answer data corresponding to the candidate question data from the data source includes: Generate a first prompt instruction based on the data source and the identification requirement; The first prompt instruction is input into the knowledge recognition model to instruct the knowledge recognition model to identify candidate question data and candidate answer data corresponding to the candidate question data from the data source according to the recognition requirement.

11. A knowledge processing method, characterized in that: include: Use historical customer service session data or object description information as the data source; Using a knowledge recognition model, identifying candidate question data and candidate answer data corresponding to the candidate question data from the data source; The candidate question-answer knowledge pairs consisting of the candidate question data and the candidate answer data are saved in the question-answer database.

12. A message processing method, characterized in that: include: Get user messages sent by the client based on the session connection of the client; Obtaining historical session messages in the session connection; Utilizing a message processing model, statistics are collected on the user messages and the historical conversation messages to extract target question data; Searching for at least one target answer data matching the target question data from a question-answering knowledge base; The question-answer knowledge base is updated based on the question-answer pairs extracted from the data source by the knowledge recognition model; The data source includes historical customer service session data or object description information; Generate a reply message corresponding to the user message according to the at least one target response data.

13. The method according to claim 12, characterized in that The step of searching the question-answer knowledge base for at least one target answer data that matches the target question data includes: Searching for multiple candidate question data that meet a third similarity requirement with the target question data from a question-answering knowledge base; Calculating the matching degree between the plurality of candidate answer data corresponding to the plurality of candidate question data and the target question data; At least one target response data having a matching degree that satisfies the matching requirement is determined.

14. The method according to claim 12, characterized in that The utilizing of the message processing model to collect statistics on the user messages and the historical conversation messages to extract target question data includes: generating a fourth prompt instruction based on the user message, the historical conversation message, and the extraction requirement; The fourth prompt instruction is input into the message processing model to instruct the message processing model to extract target question data from the user message and the historical conversation message according to the extraction requirement.

15. The method according to claim 12, characterized in that Generating a reply message corresponding to the user message according to the at least one target response data includes: A response generation model is used to generate a reply message corresponding to the user message according to the at least one target response data.

16. A message processing method, characterized in that: include: Get user messages sent by the client based on the session connection with the client; Determining user characteristic data corresponding to the user terminal; Obtaining historical conversation messages corresponding to the user message; Utilizing a message processing model and combining the user feature data, statistics are collected on the user messages and the historical conversation messages to extract target question data; Searching for at least one target answer data matching the target question data from a question-answering knowledge base; Generate a reply message corresponding to the user message according to the at least one target response data.

17. The method according to claim 16, characterized in that The customer service end includes a customer service robot; and obtaining a user message sent by the user end based on a session connection with the customer service end includes: In response to a customer service request sent by a user terminal, establishing a conversation connection between the user terminal and the customer service robot; Obtaining a user message sent by the user end to the customer service robot based on the session connection with the customer service end; Also includes: The reply message is sent to the user terminal.

18. The method according to claim 16, characterized in that Also includes: Based on the reply message, sending recommendation prompt information to the customer service end; In response to the recommendation confirmation request sent by the client, the reply message is sent to the user.

19. A computing device, characterized in that including processing components and storage components; The storage component stores a computer program; the computer program is called and executed by the processing component to implement the knowledge processing method as described in any one of claims 1 to 10, or the knowledge processing method as described in claim 11, or the message processing method as described in any one of claims 12 to 15, or the message processing method as described in any one of claims 16 to 18.

20. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed by the processing component, it implements the knowledge processing method as described in any one of claims 1 to 10, the knowledge processing method as described in claim 11, the message processing method as described in any one of claims 12 to 15, or the message processing method as described in any one of claims 16 to 18.

21. A computer program product, characterized in that It includes a computer program or instruction, which, when executed by a processing component, implements the knowledge processing method according to any one of claims 1 to 10, the knowledge processing method according to claim 11, the message processing method according to any one of claims 12 to 15, or the message processing method according to any one of claims 16 to 18.

Images