Data processing method and device, model training method and device, equipment, medium and program

Abstract

The invention provides a data processing method and device, a model training method and device, equipment, a medium and a program, and relates to the technical field of computer application, in particular to artificial intelligence technologies such as knowledge maps and natural language process.The data processing method comprises the steps that current question data input by a current user to a service large model is obtained; determining a user role of the current user according to the current question data; wherein the user role comprises at least one of service personnel and service consulting personnel; and determining current answer data matched with the current question data according to the user role of the current user. According to the embodiment of the invention, the personalized answering service of the large service model to the user can be flexibly adjusted, so that the intelligence and service precision of the large service model are improved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer application technology, and specifically to artificial intelligence technologies such as knowledge graphs and natural language processing. Background Art

[0002] With the rapid development and widespread application of information technology, smart service systems have become an important means for enterprises and institutions to improve service levels and public satisfaction. In the face of the basic needs of "Internet + smart services", it is necessary to further improve the service space gridding, service experience intelligence and service precision of smart service systems. Summary of the invention

[0003] The disclosed embodiments provide a data processing and model training method, apparatus, device, medium and program, which can flexibly adjust the personalized answer service of the service big model to the user, thereby improving the intelligence of the service big model and the precision of the service.

[0004] In a first aspect, an embodiment of the present disclosure provides a data processing method, including:

[0005] Get the current question data input by the current user to the service model;

[0006] Determine the user role of the current user according to the current question data; wherein the user role includes at least one of a service staff and a service consultant;

[0007] The current answer data matching the current question data is determined according to the user role of the current user.

[0008] In a second aspect, the present disclosure provides a model training method, including:

[0009] Obtain sample data of questions asked by users with different roles;

[0010] Determine the target service reference data that matches the question sample data;

[0011] Determine answer sample data for the sample data of questions asked by users of different roles according to the target service reference data;

[0012] Based on the user role, construct question-answer sample pairs according to the question sample data and the answer sample data;

[0013] Training the service model based on sample pairs of answers to the questions;

[0014] Among them, the trained service big model is used to execute the data processing method described in any embodiment of the present disclosure.

[0015] In a third aspect, an embodiment of the present disclosure provides a data processing device, including:

[0016] The current question data acquisition module is used to obtain the current question data input by the current user to the service model;

[0017] A user role determination module, configured to determine the user role of the current user according to the current question data; wherein the user role includes at least one of a service staff and a service consultant;

[0018] The answer data determination module is used to determine the current answer data matching the current question data according to the user role of the current user.

[0019] In a fourth aspect, an embodiment of the present disclosure provides a model training device, including:

[0020] A question sample data acquisition module is used to obtain question sample data of users with different roles;

[0021] A target service reference data determination module, used to determine the target service reference data that matches the question sample data;

[0022] An answer sample data determination module, used to determine answer sample data for the question sample data of the users with different roles according to the target service reference data;

[0023] A question-answer sample pair construction module is used to construct a question-answer sample pair based on the question sample data and the answer sample data based on the user role;

[0024] A service big model training module, used for training the service big model according to the sample pairs of answers to the questions;

[0025] Among them, the trained service big model is used to execute the data processing method described in any embodiment of the present disclosure.

[0026] In a fifth aspect, an embodiment of the present disclosure provides an electronic device, including:

[0027] at least one processor; and

[0028] a memory communicatively connected to the at least one processor; wherein,

[0029] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the data processing method provided by the first aspect embodiment or the model training method provided by the second aspect embodiment.

[0030] In the sixth aspect, the embodiments of the present disclosure also provide a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the data processing method provided in the first aspect embodiment or the model training method provided in the second aspect embodiment.

[0031] In the seventh aspect, the embodiments of the present disclosure also provide a computer program product, including a computer program, which, when executed by a processor, implements the data processing method provided in the embodiment of the first aspect or the model training method provided in the embodiment of the second aspect.

[0032] The disclosed embodiment obtains the sample data of questions from users of different roles, and determines the target service reference data that matches the sample data of questions, and then determines the sample data of answers for the sample data of questions from users of different roles based on the target service reference data, thereby taking the user role as a benchmark, constructing question-answer sample pairs based on the sample data of questions and the sample data of answers, and training the service big model based on the question-answer sample pairs. After the training of the service big model is completed, the current question data input by the current user to the service big model can be obtained, and the user role of the current user such as service personnel or service consultant can be determined based on the current question data, and finally the current answer data that matches the current question data can be determined based on the user role of the current user. The above technical solution can solve the problem that the existing big service model cannot provide personalized answer services to users of different roles, and can flexibly adjust the personalized answer services of the big service model to users, thereby improving the intelligence of the big service model and the precision of services.

[0033] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0035] Figure 1 is a flow chart of a data processing method provided by an embodiment of the present disclosure;

[0036] Figure 2 is a flow chart of a data processing method provided by an embodiment of the present disclosure;

[0037] Figure 3 This is a schematic diagram of a process for a user to use a medical insurance big model service provided by an embodiment of the present disclosure;

[0038] Figure 4 is a flow chart of a service large model training method provided by an embodiment of the present disclosure;

[0039] Figure 5 It is a schematic diagram of the overall training process of a large medical model provided by an embodiment of the present disclosure;

[0040] Figure 6 is a structural diagram of a data processing device provided by an embodiment of the present disclosure;

[0041] Figure 7 is a structural diagram of a model training device provided by an embodiment of the present disclosure;

[0042] Figure 8 It is a schematic diagram of the structure of an electronic device used to implement the data processing method or model training method of the embodiment of the present disclosure. DETAILED DESCRIPTION

[0043] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0044] Large Model (also called Foundation Model) refers to a machine learning model with a large number of parameters and complex structure, which can process massive data and complete various complex tasks, such as natural language processing, computer vision, and speech recognition. Compared with small models, large models usually have more parameters and deeper layers, with stronger expression ability and higher accuracy, but also require more computing resources and time for training and reasoning. They are suitable for scenarios with large data volumes and sufficient computing resources, such as cloud computing, high-performance computing, and artificial intelligence. They can be applied to various application fields such as medical image analysis, autonomous driving, security monitoring, face recognition, image search, and robot navigation.

[0045] Large language model is a typical large model, usually a natural language processing model with large-scale parameters and computing power. These models can be trained with a large amount of data and parameters to generate human-like text or answer natural language questions. Large language models are widely used in natural language processing, text generation and intelligent dialogue. The output data of the large model may include but is not limited to text, voice, audio and video data. For example, the large language model based on neural network can realize automatic understanding, generation and inference of text by training massive language data, and is suitable for various tasks such as language translation, text summary, query professional knowledge, generation of articles that meet the requirements, sentiment analysis and dialogue system. The input of the large language model is data, such as text data, voice data or text and picture data. The large language model can obtain the corresponding word vector representation by encoding the input data, and further decode the word vector obtained by encoding, so as to automatically complete the processing of the input data and obtain the output data corresponding to the input data. Exemplarily, the large language model can be input with text, and the large language model can process and predict the input text and output the corresponding answer text of the text.

[0046] However, when big models are currently applied in various fields, because they are used to implement functions such as natural language processing, text generation, and intelligent dialogue, they only give a unified standard answer to questions raised by users, and cannot provide personalized answer services that distinguish user roles.

[0047] In one example, Figure 1 is a flow chart of a data processing method provided by an embodiment of the present disclosure. This embodiment is applicable to the case where personalized answer services are provided to users of different roles through a service macro model. The method can be executed by a data processing device based on a service macro model. The device can be implemented by software and / or hardware and can generally be integrated in an electronic device. The electronic device can be a terminal device or a server device. The embodiment of the present disclosure does not limit the specific device type of the electronic device. Accordingly, Figure 1 As shown, the method includes the following operations:

[0048] S110. Obtain the current question data input by the current user to the service macro model.

[0049] The current user may be a user who is currently using the service big model for human-computer dialogue. The current question data may be question data currently input by the current user to the service big model. Optionally, the current question data may be text type question data, or audio and video type question data, as long as the user's question intention can be identified. The disclosed embodiment does not limit the data type of the current question data.

[0050] Among them, the service big model can be an artificial intelligence model applied in the service field, and can be a big model based on the big model and trained with sample data related to various service-related fields. The service big model provides efficient and accurate services to enterprises and the public through technologies such as deep learning and big data analysis. The application scenarios of the service big model are wide, including but not limited to service consultation, auxiliary handling, urban governance, auxiliary decision-making, smart government affairs and security prevention, which can improve the work efficiency of enterprises, optimize services and enhance the service experience of the public. The service big model provides scientific decision-making for service objects by providing data analysis, prediction and planning, and at the same time improves the work efficiency and service level of enterprises through intelligent means, realizing the intelligent transformation of services. In addition, the service big model is also widely used in public services and public welfare fields such as education, medical care, culture and sports, and improves service quality and efficiency through intelligent means, bringing more convenient service experience to the public.

[0051] In the embodiments of the present disclosure, based on the service big model, services such as service consultation, decision support, and security prevention can be provided to users. Exemplarily, the current question data can be service consultation data. Service consultation can include but is not limited to technical service consultation, policy answer consultation, software consultation, financial consultation, education consultation, enterprise management consultation, logistics consultation, legal consultation, and career consultation, etc. As long as there is a consultation demand, the embodiments of the present disclosure do not limit the specific types of service consultation that the service big model can provide.

[0052] In an optional embodiment of the present disclosure, the service big model may include but is not limited to at least one of a medical insurance big model, an education big model, a real estate big model or a medical big model.

[0053] Among them, the medical insurance big model can provide users with medical insurance service functions, such as medical insurance policy consultation, medical insurance service work assessment and medical insurance reimbursement processing. The application of the medical insurance big model can also be extended to other aspects of the medical field, such as clinical decision support, fundus screening, smart medical records, intelligent prescription review, chronic disease management and medical big data services, providing AI (artificial intelligence) medical solutions for hospitals, government agencies, patients and medical device companies.

[0054] At present, medical insurance services are mainly manual, with low service efficiency, and are unable to cope with the massive consultations from the public. By introducing the medical insurance big model to provide medical insurance service functions, it is possible to provide convenient and efficient services to insured persons, empower medical insurance-related service personnel, and promote the effective implementation and execution of medical insurance policies.

[0055] Among them, the educational big model can be an artificial intelligence model that is suitable for educational scenarios, has ultra-large-scale parameters, and integrates general knowledge and professional knowledge training. It is an integration of big model technology, knowledge base technology and various intelligent education technologies, and can promote the two-way construction of human learning and machine learning.

[0056] The educational big model can promote the digital transformation and intelligent upgrading of education from multiple aspects such as learning space, learning resources, teacher roles and parent roles, and form a new educational ecology of human-machine collaborative symbiosis. Based on the general big model, the educational big model can continuously train the educational scenario model by connecting various educational digital applications, and continuously improve the ability to solve educational professional tasks. With the support of the educational big model, learners can obtain learning support through human-computer interaction, create learning results, construct personal learning space and collective learning space, and form a learning scene where physical space and cyberspace are mutually integrated, so that all learners can obtain any information they need anywhere and at any time. The educational big model can replace repetitive and inefficient educational labor, improve the scientificity and creativity of educational and teaching work, and promote teachers from "experts in teaching" to "experts in learning", and provide personalized support for each learner through creative teaching design. At the same time, the educational big model can also provide parents with auxiliary education and education policy consultation functions, such as providing parents with interpretation of admission conditions in different regions, university major consultation, and college entrance examination volunteer filling consultation. At the same time, for teachers and students, the educational big model can also provide them with teaching question-and-answer services.

[0057] Among them, the real estate big model can refer to a model or system that uses large-scale data and complex algorithms to conduct market analysis, forecasting, evaluation and decision support in the real estate field.

[0058] The real estate big model can combine various data sources, including historical data of the real estate market, geographic information, economic indicators, demographic data, policies and regulations, etc., to help stakeholders such as developers, investors and government agencies better understand market trends and risks and make more informed decisions.

[0059] In addition, service big models can also include medical big models. Medical big models refer to complex and highly accurate mathematical models obtained through deep learning training of medical data. These models combine deep learning and big data technologies and can be used in various medical application scenarios, such as disease prediction, medical image recognition, simulation, treatment plan formulation, and auxiliary diagnosis. The basic principle of the medical big model is based on deep learning neural networks. It has a large number of parameters and hierarchical structures, can handle complex data relationships, provide more accurate prediction and diagnosis results, and can provide personalized medical services to doctors and patients respectively.

[0060] It can be seen from this that no matter whether it is the medical insurance big model, the education big model, the real estate big model or the medical big model, any type of service big model can provide the corresponding personalized services required by users of different roles.

[0061] S120. Determine a user role of the current user according to the current question data; wherein the user role includes at least one of a service staff and a service consultant.

[0062] Among them, the service personnel may be the staff who provide corresponding types of services to users. The service consultant may be the person who has consultation needs for service information.

[0063] Correspondingly, after the current user inputs the current question data to the service big model, the service big model can parse the acquired current question data to automatically determine the user role of the current user according to the parsing result.

[0064] For example, if the service model is a medical insurance model, the service personnel in the user role may include but are not limited to medical insurance business handling personnel and hospital medical insurance office window service personnel. The service consultant in the user role may include the insured. It should be noted that if the hospital medical insurance office window service personnel provides medical insurance consulting services to the window consultant, the hospital medical insurance office window service personnel can also be used as the user role of the service consultant.

[0065] For example, if the service macromodel is an education macromodel, the service personnel in the user role may include but are not limited to teachers and other types of educators, etc. The service consultants in the user role may include students and parents, etc.

[0066] For example, if the service macromodel is a real estate macromodel, the service personnel in the user role may include but are not limited to real estate developers, real estate management-related workers and agents, etc. The service consultants in the user role may include home buyers, etc.

[0067] S130: Determine current answer data that matches the current question data according to the user role of the current user.

[0068] Among them, the current answer data can be the answer data provided by the service big model to the current question data input by the current user.

[0069] After the service macro model determines the user role of the current user, it can determine the current answer data that matches the current question data according to the user role of the previous user. It is understandable that due to different user roles, even if the content of the current question data raised by different user roles is the same, the content of the current answer data given by the service macro model may be different.

[0070] In a specific example, suppose that the current question data input by the educator into the education big model is "What is the latest transfer policy for the primary school in BB District, AA City?" Since educators usually need more detailed policy content, the current answer data given by the education big model for the current question data can be the content data of the current latest transfer policy of BB District, AA City, such as the latest transfer policy officially released by BB District, AA City, that is, "Families who acquire real estate in BB District, AA City after the implementation of this notice, if the purchased real estate is an existing house (second-hand house), must hold the property for two years before they can apply for transfer; the two-year holding period is until August 31 (inclusive) of the year in which the transfer application is made. If the purchased property is a newly built commercial house (first-hand house) within the scope of the eco-city, it will not be affected." Assume that the current question data input by parents to the education big model is "What is the latest transfer policy for primary schools in BB District, AA City?" Since parents usually need concise and easy-to-understand policy content, the current answer data given by the education big model for the current question data can be relevant policy data that is further summarized and refined after interpreting the current latest transfer policy in BB District, AA City. For example, it can be the latest transfer policy officially released by BB District, AA City, that is, "students who apply to transfer to primary schools in BB District, AA City after September 1, 2024, if their families have acquired second-hand houses in BB District, AA City after September 1, 2024, can only apply for transfer after three years. Families who purchase first-hand houses are not affected."

[0071] It can be seen from this that by determining the current answer data that matches the current question data based on the user role of the current user, the service big model can generate corresponding answer content according to the characteristics and needs of different user roles, thereby flexibly adjusting the language style and expression of the answer data to meet the personalized answer needs of users with different roles, and further improving the intelligence of the service big model and the precision of the service.

[0072] Optionally, the ability to interpret relevant service content (such as policy content in related fields or corporate product white papers, etc.) of the service big model can be encapsulated into a standard API (Application Programming Interface) interface for different users to call in the client system.

[0073] The disclosed embodiment obtains the current question data input by the current user to the service big model, and determines the user role of the current user such as service personnel or service consultant according to the current question data, and finally determines the current answer data matching the current question data according to the user role of the current user. The above technical solution can solve the problem that the existing service big model cannot provide personalized answer services to users with different roles, and can flexibly adjust the personalized answer service of the service big model to users, thereby improving the intelligence of the service big model and the precision of the service.

[0074] In one example, Figure 2 It is a flow chart of a data processing method provided by an embodiment of the present disclosure. Based on the technical solutions of the above embodiments, the embodiment of the present disclosure has been optimized and improved, and multiple specific optional implementation methods are provided for determining the user role of the current user and determining the current answer data matching the current question data according to the user role of the current user.

[0075] like Figure 2 A data processing method based on a service big model is shown, comprising:

[0076] S210: Obtain the current question data input by the current user to the service macro model.

[0077] S220: Analyze the current question data to obtain a current question data analysis result.

[0078] The current question data analysis result may be the analysis result obtained by analyzing the current question data by the service big model.

[0079] It is understandable that the data type of the current question data is different, and the service model analyzes the current question data in different ways.

[0080] In an optional embodiment of the present disclosure, the current question data may include text question data; parsing the current question data to obtain the current question data parsing result may include: performing text parsing on the text question data to obtain user association information of the current user; extracting the current user's role association information based on the current user's user association information to obtain the current question data parsing result.

[0081] The user-related information may be relevant information about the user, and the role-related information may be relevant information that can reflect the user's role.

[0082] Optionally, the current question data may be text-type question data. Accordingly, if the current user inputs text question data to the service macro model, the service macro model may directly perform text analysis on the text question data, thereby obtaining user association information of the current user according to the text analysis result.

[0083] For example, assuming that the service model is an education model, and the current question data is in text form: "We bought a house in District AA in July 2024 and registered our household registration. Are there any restrictions on our primary school enrollment?", the education model can determine that the user association information of the current user is: homebuyer in District AA. Further, based on the user association information of the current user "homebuyer", the role association information of the current user can be extracted as a parent, and the current question data analysis result can be obtained.

[0084] For example, assuming that the service model is a medical insurance model, and the current question data is in text form: "My daughter was born in July 2024. Can I apply for medical insurance for her for 24 years?", the medical insurance model can determine that the user association information of the current user is: a newborn born in 24 years. Further, based on the user association information of the current user "a newborn born in 24 years", the role association information of the current user can be extracted as a newborn, and the current question data analysis result can be obtained.

[0085] The above technical solution can automatically extract user role information based on the text analysis results by performing text analysis on text question data, thereby further improving the intelligence of service large model data processing.

[0086] In an optional embodiment of the present disclosure, the current question data may include audio and video question data; the parsing of the current question data to obtain the current question data parsing result may include: performing voiceprint analysis on the audio and video question data to obtain the audio fingerprint information of the current user; determining the role association information of the current user based on the audio fingerprint information of the current user to obtain the current question data parsing result.

[0087] The audio and video question data may be audio and video type question data.

[0088] Optionally, the current question data may also be audio and video type question data. It is understandable that the audio fingerprints corresponding to different users are also different. Accordingly, if the current user inputs audio and video question data to the service big model, the service big model can directly perform voiceprint analysis on the audio and video question data, thereby obtaining the audio fingerprint information of the current user according to the voiceprint analysis result. A mapping relationship can be established between the audio fingerprint information of each user and the corresponding user role. Exemplarily, the audio fingerprint information of each user can establish a mapping relationship with the identity, position, etc. of the corresponding user. Further, the service big model can obtain the identity and position of the corresponding user and other related information according to the audio fingerprint information of the current user, and then extract the role association information of the current user, thereby further generating the current question data analysis result according to the role association information. Exemplarily, the service big model obtains the current user's job position as a medical insurance business handler according to the audio fingerprint information of the current user, and then determines that the role association information of the current user is a medical insurance business handler. The service big model obtains the current user's job position as a school teacher according to the audio fingerprint information of the current user, and then determines that the current user's role association information is a teacher.

[0089] The above technical solution, by performing voiceprint analysis on audio and video question data, can automatically extract user role information based on audio fingerprint information, further improving the intelligence of service large model data processing.

[0090] It should be noted that if the service model determines the role-related information of the current user based on both the text analysis results and the voiceprint analysis results, and there is a contradiction or conflict between the role-related information of the current user determined by the two methods, then the current question data input by the user can be further accurately identified, such as by using intent analysis and other methods to perform role identification, and obtain accurate analysis results of the current question data. Alternatively, the accurate role information of the current user can be obtained by asking questions to the user. Alternatively, the role-related information of the current user determined by the two methods can be used as two analysis results, that is, corresponding answer data can be provided for multiple user roles analyzed at the same time. The disclosed embodiments are not limited to this.

[0091] S230: When it is determined that the current question data analysis result includes a user role, use the user role included in the current question data analysis result as the user role of the current user.

[0092] In the disclosed embodiment, if the service big model determines that the current question data parsing result includes the user role, such as directly including the user role information, or implicitly including the user role information through the current user's role association information, the user role included in the current question data parsing result can be used as the user role of the current user, without the need to guide the user to provide the user role information to the service big model in the form of human-computer dialogue through repeated questioning.

[0093] The above technical solution can automatically obtain user role information by parsing the user's question data, thereby realizing imperceptible determination of user role identity and further improving the service intelligence of the service big model.

[0094] S240: Determine current answer data that matches the current question data according to the user role of the current user.

[0095] In an optional embodiment of the present disclosure, if the user role is the service personnel, then determining the current answer data that matches the current question data based on the user role of the current user may include: determining a plurality of first-classified answer data that matches the current question data; wherein the first-classified answer data includes standard answer data and concise answer data; and using the standard answer data as the current answer data that matches the current question data.

[0096] The first category of answer data may be answer data distinguished by standardized style and simplified style. Standard answer data may be official answer data in a standard format provided by reference to relevant service data (such as a policy or product white paper, etc.). Simplified answer data may be answer data in a concise format provided by reference to relevant service data and easy to understand.

[0097] In the disclosed embodiment, for the same question data, different types of answer data can be stored according to the user role. Optionally, if the type of the answer data is the first category answer data, the service macro model can store the corresponding standard answer data and concise answer data for the same question data at the same time, that is, store in a one-to-many manner with questions and answers. Alternatively, the service macro model can also store the corresponding standard answer data and concise answer data for the same question data. For example, question A and the corresponding standard answer data constitute a question-answer pair for storage, and question A and the corresponding concise answer data constitute a question-answer pair for storage, that is, store in a one-to-one manner with questions and answers. When storing question-answer pairs, each type of answer data needs to be labeled with the applicable user role. For example, the standard answer data corresponds to the user role marked as service personnel, and the concise answer data corresponds to the user role marked as service consultant.

[0098] Optionally, if it is determined that the user role of the current user is a service personnel, and service personnel are usually highly professional, efficient in business processing, and have great responsibilities, therefore, when service personnel use the service big model to handle certain businesses or certain processes, the service big model is often required to provide professional, standard, and official answer services. At this time, the service big model can first determine multiple first-category answer data that match the current question data, and further select standard answer data from each first-category answer data as the current answer data that matches the current question data based on the user role of the service personnel. The advantage of this setting is: providing service personnel with formal, rigorous, well-organized, and highly practical answer services through standard answer data, helping service personnel quickly solve actual business problems.

[0099] In an optional embodiment of the present disclosure, if the user role is a service consultant, then determining the current answer data that matches the current question data based on the user role of the current user may include: determining multiple first-classified answer data that match the current question data; wherein the first-classified answer data includes standard answer data and concise answer data; and using the concise answer data as the current answer data that matches the current question data.

[0100] Optionally, if it is determined that the user role of the current user is a service consultant, and service consultants usually have the characteristic of quickly obtaining service interpretation content, therefore, when service personnel use the service big model to handle certain business or consult a certain service, the service big model is often required to provide a concise and easy-to-understand answer service. At this time, the service big model can first determine multiple first-category answer data that match the current question data, and further select concise answer data from each first-category answer data as the current answer data that matches the current question data based on the user role of the service consultant. The advantage of this setting is: providing concise and easy-to-understand answer services to service consultants through concise answer data, helping service consultants quickly obtain key service interpretations and practical information.

[0101] In an optional embodiment of the present disclosure, the service consultants include underage consultants and adult consultants; the determining of the current answer data matching the current question data based on the user role of the current user may include: determining a plurality of second-classified answer data matching the current question data; wherein the second-classified answer data include concise answer data for adult roles and concise answer data for underage roles; if the user role is an adult consultant, the concise answer data for adult roles is used as the current answer data matching the current question data; if the user role is a underage consultant, the concise answer data for underage roles is used as the current answer data matching the current question data.

[0102] Among them, the second category answer data can be answer data distinguished by adult user style and underage user style. Similar to the first category answer data, a one-to-many or one-to-one storage method can be adopted between questions and second category answer data. When storing question-answer pairs, each type of answer data needs to be labeled with the applicable user role. Underage consultants can be underage users with service consultation needs, such as elementary school students, junior high school students or children. Adult consultants can be adult users with service consultation needs. Adult role concise answer data can be answer data applicable to adult consultants. Underage role concise answer data can be answer data applicable to underage consultants.

[0103] It is understandable that for some types of service big models, such as education big models and medical insurance big models, the role distinction between underage users and adult users will be involved. Therefore, for these types of service big models, service consultants can be further divided into underage consultants and adult consultants. Correspondingly, the service big model can first determine the multiple second-classified answer data that match the current question data, and further select the answer data of the corresponding style type from each second-classified answer data according to the user role that is refined and distinguished for the service consultant as the current answer data that matches the current question data. Specifically, if the service big model recognizes that the current user is an adult consultant, the concise answer data of the adult role can be used as the current answer data that matches the current question data; if the service big model recognizes that the current user is a minor consultant, the concise answer data of the minor role can be used as the current answer data that matches the current question data. Exemplarily, the concise answer data of the minor role can be in a style of both pictures and texts, that is, in addition to including concise text content, pictures or video animations can also be used to further help the minor consultant understand the content of the answer data. The advantage of this setting is that by further distinguishing between user roles such as underage consultants and adult consultants, personalized answer services can be provided to users from the perspective of their understanding of the answer data, helping different service consultants to quickly resolve actual consulting problems.

[0104] In a specific example, assuming that the service big model is an education big model, when the teacher and the primary school student ask a math question to the education big model as the current question data, considering that the teacher needs to provide comprehensive teaching and answering to the students, the education big model can provide standard and comprehensive answer data for the math questions raised by the teacher for reference, so that the teacher can obtain the comprehensive knowledge points included in the math questions. Considering that the primary school students have poor understanding of the answers, the education big model can provide concise and easy-to-understand answer data for the math questions raised by the primary school students for reference. If necessary, the teaching big model can also use pictures and texts to provide concise and easy-to-understand answer data for the math questions raised by the primary school students for reference, so that the primary school students can quickly understand the solution ideas of the questions.

[0105] In an optional embodiment of the present disclosure, the service personnel may include medical insurance service personnel, the service consultants may include insured persons, and the insured persons may include urban insured persons and employee insured persons; if the user role is a service consultant, then determining the current answer data matching the current question data based on the user role of the current user may include: determining multiple third-classified answer data matching the current question data; wherein the third-classified answer data includes urban concise answer data and employee concise answer data; if the insured person is an urban insured person, then the urban concise answer data is used as the current answer data matching the current question data; if the insured person is an employee insured person, then the employee concise answer data is used as the current answer data matching the current question data.

[0106] Among them, medical insurance service personnel may be personnel who provide medical insurance services to the masses. Insured persons are persons who enjoy medical insurance services. Urban insured persons are persons who participate in urban medical insurance services. Employee insured persons are persons who participate in employee medical insurance services. The third category answer data may be answer data distinguished by different insurance styles. Similar to the first category answer data and the second category answer data, a one-to-many or one-to-one storage method may be used between questions and the third category answer data. When storing question-answer pairs, each type of answer data needs to be labeled with the applicable user role. Urban concise answer data may be concise and easy-to-understand answer data provided in accordance with urban medical policies. Employee concise answer data may be concise and easy-to-understand answer data provided in accordance with employee medical policies.

[0107] In a specific example, the service big model may be a medical insurance big model. Figure 3 is a flow chart of a user using a medical insurance big model service provided by an embodiment of the present disclosure, such as Figure 3 As shown in the figure, the medical insurance model can provide medical insurance policy interpretation services to different types of users. When facing different user roles, the medical insurance model can deliver differentiated information according to their characteristics and needs to ensure accurate understanding and effective implementation of medical insurance policies. The medical insurance model can flexibly adjust the language style and expression when generating answer data for different user roles.

[0108] Among them, for medical insurance business handlers, as medical insurance service personnel, they have the following characteristics. (1) Strong professionalism: Familiar with the details of medical insurance policies, they need to master the latest policy trends and specific operating procedures. (2) Efficiency: Handling a large number of businesses, they need to interpret and implement policies quickly and accurately. (3) Heavy responsibility: Directly affecting the rights and interests of insured persons, it is necessary to ensure the accuracy of policy implementation. Therefore, for medical insurance business handlers, the medical insurance big model can determine that the content style of the corresponding answer data can be: (1) Formal and rigorous: Use professional terms to ensure the accuracy and authority of information. (2) Clear and organized: The policy interpretation is clear and organized, which is convenient for handlers to quickly understand and implement. (3) Strong practicality: Provide detailed operation guides and case analysis to help handlers solve practical problems.

[0109] Among them, the window service personnel of the hospital medical insurance office, because they are medical insurance service personnel, they also need to connect with the insured persons to provide services to them. Therefore, they can provide them with both standard answer data and concise answer data. The window service personnel of the hospital medical insurance office have the following specific characteristics: (1) Directly facing the insured persons: they need to have good communication skills and service awareness. (2) Policy explanation: they often need to explain the medical insurance policy to the insured persons and help them understand the reimbursement process. (3) Patient and meticulous: facing various questions from the insured persons, they need to patiently answer them to ensure satisfaction. Therefore, for the window service personnel of the hospital medical insurance office, the medical insurance big model can determine that the content style of the corresponding answer data can be: (1) Easy to understand: explain complex medical insurance policies in simple and easy-to-understand language to avoid misunderstandings caused by professional terms. (2) Warm and considerate: reflect humanistic care, and use warm language and attitude to relieve the anxiety of the insured persons. (3) Operational guidance: provide a clear reimbursement process and a list of required materials to facilitate the preparation of the insured persons.

[0110] Among them, insured persons as service consultants have the following characteristics: (1) Diversity: they have different ages, occupations and educational levels, and their understanding of medical insurance policies is different. (2) Focus on their own interests: they are concerned about the reimbursement scope and reimbursement ratio and other matters that are directly related to their own interests. Diversified needs: some insured persons may have special needs, such as chronic disease management or medical treatment in other places.

[0111] Therefore, for insured persons, the medical insurance big model can determine that the content style of the corresponding answer data can be: (1) Concise: highlight the key points and directly answer the questions that the insured persons are most concerned about. (2) Graphical: use charts and schematic diagrams (such as relevant certificates or material sample pictures) and other auxiliary explanations to improve readability. (3) Personalized push: push relevant policy and service information based on the insured person's actual situation (such as age and health status).

[0112] In summary, the medical insurance model can flexibly adjust the language style and expression method when generating content according to the characteristics and needs of different user roles, and push differentiated information on medical insurance policies. That is, medical insurance business handlers can use formal and rigorous language styles to provide detailed policy interpretations and operation guidelines to ensure the accuracy and operability of information, thereby assisting handlers to quickly understand the process and policy of handling a certain business, and thus quickly complete the handling of the business. The medical insurance model can explain the cost composition and calculation logic to the medical insurance office service personnel according to the medical insurance settlement form. The answer data provided by the window service personnel of the hospital medical insurance office can focus on the affinity and understandability of the language, explain the policy in simple and clear language, provide intimate service suggestions, and enhance the trust and satisfaction of the insured. The answer data of the insured can present the policy content in a concise and graphic way, highlight the key points and practical information, and consider the personalized push needs at the same time, improve the pertinence and attractiveness of the information, and quickly provide the required information or answers to the insured, so as to answer all questions and respond in time.

[0113] In a specific example, the medical insurance big model can be used to assist the medical insurance business handling process, such as helping service personnel improve the processing efficiency of insurance registration and change, fee settlement and reimbursement, out-of-town medical record, personal account query and withdrawal, and relationship transfer and continuation. The medical insurance big model can also be used to interpret policy summaries, thereby helping service personnel to master the latest medical insurance policy information. The medical insurance big model can also provide an explanation of the medical insurance reimbursement calculation logic, specifically providing the medical insurance reimbursement calculation logic according to system data or document data. At the same time, the medical insurance big model can also provide material guidance services, providing insured persons with a comprehensive and accurate list of materials and handling guides to ensure that they can complete the handling of medical insurance business smoothly and quickly. Furthermore, the medical insurance big model can also provide a popular policy interpretation function, thereby converting complex and professional medical insurance policy content into concise and easy-to-understand language, helping insured persons better understand their own medical insurance rights and benefits. Finally, the medical insurance big model can also predict the medical insurance issues that they may be concerned about based on the specific circumstances of different insured persons, such as age, occupation, health status and historical medical records, and give targeted answers to medical insurance questions.

[0114] It should be noted that medical insurance policies usually distinguish between urban medical insurance and employee medical insurance. Therefore, for insured persons, it is possible to further distinguish whether they are urban insured persons or employee insured persons. Accordingly, for urban insured persons, the medical insurance big model can generate urban concise answer data based on urban medical insurance policies, and use the urban concise answer data as the current answer data; for employee insured persons, the medical insurance big model can generate employee concise answer data based on employee medical insurance policies, and use the employee concise answer data as the current answer data.

[0115] Through the above-mentioned differentiated content generation strategy, the medical insurance big model can better serve different user roles and promote the effective implementation of medical insurance policies.

[0116] In an optional embodiment of the present disclosure, the method may further include: determining an associated region of the current user based on the current question data; and determining current answer data matching the current question data based on the user role of the current user and the associated region of the current user.

[0117] The associated region area may be an area related to the user role, for example, including but not limited to the current user's insurance area, school enrollment area, or home purchase area, etc. That is, the type of the associated region area is directly related to the specific application field of the service macro model, and the embodiment of the present disclosure does not limit the specific region area type of the associated region area.

[0118] In a specific example, for government policies, the same government policy may be different in different regions. For example, the admission policy in BB District of AA City may be different from that in CC District of AA City. The medical insurance policies in AA City and DD City may be different. Therefore, after the service big model obtains the current question data, it can further parse the current question data to obtain the current user's associated region, and then determine the current answer data that matches the current question data in combination with the current user's user role and the current user's associated region. The advantage of this setting is that it can further distinguish regional differences, thereby further improving the flexibility and precision of the service big model's answer service.

[0119] In an optional embodiment of the present disclosure, determining the associated regional area of ​​the current user based on the current question data may include: parsing the current question data to obtain a current question data parsing result, and determining the associated regional area of ​​the current user based on the current question data parsing result; and / or determining the driving mode for the current user to call the service big model; determining the associated regional area of ​​the current user by the driving mode for the current user to call the service big model.

[0120] It is understandable that the current question data analysis results obtained by parsing part of the current question data may include relevant information about the region of the current user. Therefore, the associated region of the current user can be directly determined based on the current question data analysis results obtained by parsing the current question data. In addition, the service big model generally needs to access the third-party platform in a corresponding manner to provide corresponding services to users. Therefore, in the process of users using the service big model, the service big model can also determine the driving method of the current user calling the service big model. Exemplarily, the service big model can access the third-party platform through various available methods such as links, interfaces, and capability plug-ins. Correspondingly, the service big model can also determine the associated region of the current user through the specific driving method of the current user calling the service big model. For example, when the service big model accesses the third-party platform through an interface, after the current user calls the service big model through the interface of the third-party platform, the service big model can determine the regional area information of the current user through the relevant information of the interface as the associated region of the current user.

[0121] The above technical solution integrates other third-party professional service function modules with the service big model driver in the form of links, interfaces and capability plug-ins, and proactively provides precise services in combination with business needs. Taking the medical insurance big model as an example, embedding the medical insurance big model into the existing medical insurance information system can help service personnel reduce the business consulting pressure caused by the complexity of medical insurance policies and insufficient knowledge reserves of staff. If the medical insurance big model is connected to the mini program and WeChat public account of medical insurance services, the medical insurance big model can quickly provide the required information or answers to the insured persons in the form of text responses, so that all questions are answered and responded to in a timely manner. Through the display form of text responses, the question and answer display is more intuitive, the answers to questions are more accurate, the answers are clearer to read and can be viewed repeatedly, which further improves the consultation efficiency and effectiveness of medical insurance policies.

[0122] The disclosed embodiment provides personalized services to users of different roles through a service big model, can accurately identify and understand various types of service knowledge through a text understanding function, learn classic service data interpretation cases through a knowledge learning function, and perform logical reasoning according to corresponding file requirements and actual cases, thereby flexibly adjusting the language style and expression method to generate corresponding answer content according to the characteristics and needs of different user roles, realizing differentiated content generation and personalized push strategies, significantly improving the dissemination efficiency and execution effect of the service, promoting the effective implementation and wide implementation of the service, and meeting the diversified interpretation service needs of different users.

[0123] In the technical solution disclosed herein, the collection, storage, use, processing, transmission, provision and disclosure of user personal information (such as user role information, voice information, etc.) are in compliance with the relevant laws and regulations and do not violate public order and good morals.

[0124] It should be noted that any arrangement and combination of the technical features in the above embodiments also falls within the protection scope of the present disclosure.

[0125] In one example, Figure 4 This is a flow chart of a service model training method provided by an embodiment of the present disclosure. This embodiment can be applied to the case of training a service model to provide personalized answering services to users with different roles. The method can be executed by a service model training device, which can be implemented by software and / or hardware and can generally be integrated in an electronic device. The electronic device can be a terminal device or a server device. The embodiment of the present disclosure does not limit the specific device type of the electronic device. Accordingly, Figure 4 As shown, the method includes the following operations:

[0126] S310: Obtain sample data of questions asked by users with different roles.

[0127] The question sample data may be sample data obtained by asking questions.

[0128] In the embodiment of the present disclosure, corresponding question data may be obtained for users with different roles as question sample data.

[0129] S320: Determine target service reference data that matches the question sample data.

[0130] The target service reference data may be data related to the corresponding service type matched with the sample question data. For example, when the service macromodel is an education macromodel, the target service reference data may be education policy data and teaching resource data, etc. When the service macromodel is a product service macromodel provided by a certain enterprise, the target service reference data may be data such as a white paper corresponding to the product developed by the enterprise.

[0131] For example, when the sample data of the question is "What is the medical insurance reimbursement process in Region AA", the target service reference data matched by the sample data of the question may be the medical insurance policy of Region AA. When the sample data of the question is "What is the current enrollment policy in Region AA", the target service reference data matched by the sample data of the question may be the education policy of Region AA.

[0132] S330: Determine answer sample data for the sample data of questions asked by users of different roles according to the target service reference data.

[0133] The answer sample data may be answer data determined for the question sample data with reference to the determined target service reference data.

[0134] It is understandable that, since different role users have different requirements for answer data, even if the question sample data raised by different role users are the same, the content of the corresponding generated answer sample data may be different.

[0135] S340: Based on the user role, construct question-answer sample pairs according to the question sample data and the answer sample data.

[0136] The format of the question-answer sample pair may be a one-to-one format of question and answer, or a one-to-one format of question and answer, and the embodiments of the present disclosure do not limit this.

[0137] Since the content of the answer sample data generated corresponding to the same question sample data raised by users with different roles may be different, question-answer sample pairs can be constructed based on the question sample data and the answer sample data based on the user role.

[0138] For example, one-to-one question-answer sample pairs can be directly constructed with different role users as units. Alternatively, a group of question-answer sample pairs can be constructed by corresponding the same question sample data to the generated answer sample data. In constructing the generated question-answer sample pairs, it is necessary to annotate the corresponding user role label for each answer sample data to distinguish different answer sample data for the same question sample data.

[0139] S350, training a service big model according to the sample pairs of answers to the question;

[0140] Among them, the trained service big model is used to execute the data processing method described in any embodiment of the present disclosure.

[0141] Correspondingly, after the question-answer sample pairs are constructed, the constructed question-answer sample pairs can be used as sample data to train the service model.

[0142] Figure 5 is a schematic diagram of the overall training process of a medical large model provided by the embodiment of the present disclosure. In a specific example, Figure 5As shown, the medical big model is used as a service big model as an example. When training the model, a general big model can be selected for pre-training so that the general big model has the accuracy of the benchmark. In the process of pre-training the general big model, the medical knowledge graph generated by the medical knowledge graph construction and update technology based on the medical entity chain finger technology can be combined to guide the pre-training process of the general big model. When the pre-training is completed, it can be further combined with the specific medical field for post-pre-training. After the post-training is completed, the instructions of the trained big model can be fine-tuned to obtain the final medical big model. The trained medical big model can construct a vector database with multi-type knowledge extracted from the medical knowledge graph, which is used to match the vectors of the questions raised by the user to obtain the final recommended answers. The medical big model can be connected to the medical platform application, and the human-computer dialogue method of questions (prompt) and answers (response) can be used to provide accurate, professional and personalized medical services that can automatically distinguish user roles. For example, the medical big model can provide professional and comprehensive medical consulting services to doctors, and provide efficient and easy-to-understand medication consulting services to patients.

[0143] Correspondingly, the trained service model can be used to execute the data processing method described in any of the above embodiments.

[0144] The disclosed embodiment obtains the sample data of questions from users of different roles, and determines the target service reference data that matches the sample data of questions, and then determines the sample data of answers for the sample data of questions from users of different roles based on the target service reference data, thereby taking the user role as a benchmark, constructing question-answer sample pairs based on the sample data of questions and the sample data of answers, and training the service big model based on the question-answer sample pairs. After the training of the service big model is completed, the current question data input by the current user to the service big model can be obtained, and the user role of the current user such as service personnel or service consultant can be determined based on the current question data, and finally the current answer data that matches the current question data can be determined based on the user role of the current user. The above technical solution can solve the problem that the existing big service model cannot provide personalized answer services to users of different roles, and can flexibly adjust the personalized answer services of the big service model to users, thereby improving the intelligence of the big service model and the precision of services.

[0145] In one example, Figure 6 is a structural diagram of a data processing device provided by an embodiment of the present disclosure. The embodiment of the present disclosure is applicable to the case where personalized answer services are provided to users of different roles through a service macro model. The device is implemented by software and / or hardware and is specifically configured in an electronic device. The electronic device can be a terminal device or a server device. The embodiment of the present disclosure does not limit the specific device type of the electronic device.

[0146] like Figure 6 The data processing device 400 shown in the figure comprises: a current question data acquisition module 410, a user role determination module 420 and an answer data determination module 430.

[0147] The current question data acquisition module 410 is used to acquire the current question data input by the current user to the service macro model;

[0148] A user role determination module 420 is used to determine the user role of the current user according to the current question data; wherein the user role includes at least one of a service staff and a service consultant;

[0149] The answer data determination module 430 is used to determine the current answer data matching the current question data according to the user role of the current user.

[0150] The disclosed embodiment obtains the current question data input by the current user to the service big model, and determines the user role of the current user such as service personnel or service consultant according to the current question data, and finally determines the current answer data matching the current question data according to the user role of the current user. The above technical solution can solve the problem that the existing service big model cannot provide personalized answer services to users with different roles, and can flexibly adjust the personalized answer service of the service big model to users, thereby improving the intelligence of the service big model and the precision of the service.

[0151] Optionally, the user role determination module 420 is further configured to: parse the current question data to obtain a current question data parsing result; and when it is determined that the current question data parsing result includes a user role, use the user role included in the current question data parsing result as the user role of the current user.

[0152] Optionally, the current question data includes text question data; the user role determination module 420 is also used to: perform text parsing on the text question data to obtain user association information of the current user; extract role association information of the current user based on the user association information of the current user to obtain the current question data parsing result.

[0153] Optionally, the current question data includes audio and video question data; the user role determination module 420 is also used to: perform voiceprint analysis on the audio and video question data to obtain the audio fingerprint information of the current user; determine the role association information of the current user based on the audio fingerprint information of the current user, and obtain the analysis result of the current question data.

[0154] Optionally, if the user role is the service personnel, the answer data determination module 430 is also used to: determine multiple first-classified answer data that match the current question data; wherein the first-classified answer data include standard answer data and concise answer data; and use the standard answer data as the current answer data that matches the current question data.

[0155] Optionally, if the user role is a service consultant, the answer data determination module 430 is also used to: determine multiple first-classified answer data that match the current question data; wherein the first-classified answer data include standard answer data and concise answer data; and use the concise answer data as the current answer data that matches the current question data.

[0156] Optionally, the service consultants include underage consultants and adult consultants; the answer data determination module 430 is also used to: determine multiple second-category answer data that match the current question data; wherein the second-category answer data includes concise answer data for adult roles and concise answer data for underage roles; if the user role is an adult consultant, the concise answer data for adult roles is used as the current answer data that matches the current question data; if the user role is a underage consultant, the concise answer data for underage roles is used as the current answer data that matches the current question data.

[0157] Optionally, the service personnel include medical insurance service personnel, the service consultants include insured persons, and the insured persons include urban insured persons and employee insured persons; if the user role is a service consultant, the answer data determination module 430 is also used to: determine multiple third-category answer data that match the current question data; wherein the third-category answer data includes urban concise answer data and employee concise answer data; if the insured person is an urban insured person, the urban concise answer data is used as the current answer data that matches the current question data; if the insured person is an employee insured person, the employee concise answer data is used as the current answer data that matches the current question data.

[0158] Optionally, the answer data determination module 430 is further used to: determine the associated region area of ​​the current user according to the current question data; and determine the current answer data matching the current question data according to the user role of the current user and the associated region area of ​​the current user.

[0159] Optionally, the answer data determination module 430 is also used to: parse the current question data to obtain the current question data analysis result, and determine the associated regional area of ​​the current user based on the current question data analysis result; and / or determine the driving method for the current user to call the service big model; determine the associated regional area of ​​the current user by the driving method for the current user to call the service big model.

[0160] Optionally, the service big model includes at least one of a medical insurance big model, an education big model, a real estate big model or a medical big model.

[0161] The above data processing device can execute the data processing method provided by any embodiment of the present disclosure, and has the corresponding functional modules and beneficial effects of the execution method. For technical details not described in detail in this embodiment, please refer to the data processing method provided by any embodiment of the present disclosure.

[0162] Since the data processing device introduced above is a device that can execute the data processing method in the embodiment of the present disclosure, based on the data processing method introduced in the embodiment of the present disclosure, the technical personnel in the field can understand the specific implementation of the data processing device of the present embodiment and its various variations, so how the data processing device implements the data processing method in the embodiment of the present disclosure is not described in detail here. As long as the technical personnel in the field implement the device used by the data processing method in the embodiment of the present disclosure, it belongs to the scope of protection of the present disclosure.

[0163] In one example, Figure 7 This is a structural diagram of a model training device provided by an embodiment of the present disclosure. The embodiment of the present disclosure can be applied to the case where a large service model is trained to provide personalized answer services to users with different roles. The device is implemented by software and / or hardware and is specifically configured in an electronic device. The electronic device can be a terminal device or a server device. The embodiment of the present disclosure does not limit the specific device type of the electronic device.

[0164] like Figure 7 The model training device 500 shown in the figure includes: a question sample data acquisition module 510, a target service reference data determination module 520, an answer sample data determination module 530, a question answer sample pair construction module 540 and a service large model training module 550.

[0165] The question sample data acquisition module 510 is used to acquire question sample data of users with different roles;

[0166] A target service reference data determination module 520, configured to determine the target service reference data that matches the question sample data;

[0167] An answer sample data determination module 530 is used to determine answer sample data for the question sample data of the users with different roles according to the target service reference data;

[0168] A question-answer sample pair construction module 540 is used to construct a question-answer sample pair based on the question sample data and the answer sample data based on the user role;

[0169] A service big model training module 550 is used to train a service big model according to the question answer sample pairs;

[0170] Among them, the trained service big model is used to execute the data processing method described in any embodiment of the present disclosure.

[0171] The above-mentioned model training device can execute the model training method provided by any embodiment of the present disclosure, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not described in detail in this embodiment, please refer to the model training method provided by any embodiment of the present disclosure.

[0172] Since the model training device introduced above is a device that can execute the model training method in the embodiment of the present disclosure, based on the model training method introduced in the embodiment of the present disclosure, the technical personnel in the field can understand the specific implementation of the model training device of the present embodiment and its various variations, so how the model training device implements the model training method in the embodiment of the present disclosure is not introduced in detail here. As long as the technical personnel in the field implement the device used in the model training method in the embodiment of the present disclosure, it belongs to the scope of protection of the present disclosure.

[0173] In one example, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0174] Figure 8 A schematic block diagram of an example electronic device 600 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0175] like Figure 8As shown, the device 600 includes a computing unit 601, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the device 600 can also be stored. The computing unit 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0176] A number of components in the device 600 are connected to the I / O interface 605, including: an input unit 606, such as a keyboard, a mouse, etc.; an output unit 607, such as various types of displays, speakers, etc.; a storage unit 608, such as a disk, an optical disk, etc.; and a communication unit 609, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 609 allows the device 600 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0177] The computing unit 601 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 601 performs the various methods and processes described above, such as data processing methods or model training methods.

[0178] Optionally, the data processing method includes: obtaining current question data input by the current user into the service big model; determining the user role of the current user based on the current question data; wherein the user role includes at least one of a service personnel and a service consultant; and determining current answer data matching the current question data based on the user role of the current user.

[0179] Optionally, the model training method includes: obtaining question sample data of users with different roles; determining target service reference data that matches the question sample data; determining answer sample data for the question sample data of users with different roles based on the target service reference data; taking user roles as a benchmark, constructing question-answer sample pairs based on the question sample data and the answer sample data; training a service big model based on the question-answer sample pairs; wherein the trained service big model is used in the data processing method described in any embodiment of the present disclosure.

[0180] For example, in some embodiments, the data processing method or the model training method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 600 via the ROM 602 and / or the communication unit 609. When the computer program is loaded into the RAM 603 and executed by the computing unit 601, one or more steps of the data processing method or the model training method described above may be performed. Alternatively, in other embodiments, the computing unit 601 may be configured to perform the data processing method or the model training method in any other suitable manner (e.g., by means of firmware).

[0181] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0182] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0183] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0184] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0185] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0186] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services. The server can also be a server of a distributed system, or a server combined with a blockchain.

[0187] The disclosed embodiment obtains the sample data of questions from users of different roles, and determines the target service reference data that matches the sample data of questions, and then determines the sample data of answers for the sample data of questions from users of different roles based on the target service reference data, thereby taking the user role as a benchmark, constructing question-answer sample pairs based on the sample data of questions and the sample data of answers, and training the service big model based on the question-answer sample pairs. After the training of the service big model is completed, the current question data input by the current user to the service big model can be obtained, and the user role of the current user such as service personnel or service consultant can be determined based on the current question data, and finally the current answer data that matches the current question data can be determined based on the user role of the current user. The above technical solution can solve the problem that the existing big service model cannot provide personalized answer services to users of different roles, and can flexibly adjust the personalized answer services of the big service model to users, thereby improving the intelligence of the big service model and the precision of services.

[0188] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0189] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A data processing method, comprising: Get the current question data input by the current user to the service model; Determine the user role of the current user according to the current question data; wherein the user role includes at least one of a service staff and a service consultant; The current answer data matching the current question data is determined according to the user role of the current user.

2. The method according to claim 1, wherein: The determining the user role of the current user according to the current question data includes: Parsing the current question data to obtain a parsing result of the current question data; In the case where it is determined that the current question data analysis result includes a user role, the user role included in the current question data analysis result is used as the user role of the current user.

3. The method according to claim 2, wherein: The current question data includes text question data; the current question data is parsed to obtain the current question data parsing result, including: Performing text analysis on the text question data to obtain user association information of the current user; The role-associated information of the current user is extracted according to the user-associated information of the current user to obtain the analysis result of the current question data.

4. The method according to claim 2, wherein: The current question data includes audio and video question data; the current question data is parsed to obtain the current question data parsing result, including: Performing voiceprint analysis on the audio and video question data to obtain audio fingerprint information of the current user; The role association information of the current user is determined according to the audio fingerprint information of the current user, and the analysis result of the current question data is obtained.

5. The method according to claim 1, wherein: If the user role is the service personnel, then determining the current answer data matching the current question data according to the user role of the current user includes: Determine a plurality of first-classified answer data matching the current question data; wherein the first-classified answer data include standard answer data and concise answer data; The standard answer data is used as current answer data matched with the current question data.

6. The method according to claim 1, wherein: If the user role is a service consultant, then determining the current answer data matching the current question data according to the user role of the current user includes: Determine a plurality of first-classified answer data matching the current question data; wherein the first-classified answer data include standard answer data and concise answer data; The concise answer data is used as current answer data matched with the current question data.

7. The method according to claim 1, wherein: The service consultants include underage consultants and adult consultants; the current answer data matching the current question data is determined according to the user role of the current user, including: Determine a plurality of second-category answer data that match the current question data; wherein the second-category answer data includes concise answer data of adult roles and concise answer data of minor roles; If the user role is an adult consultant, the concise answer data of the adult role is used as the current answer data matched with the current question data; If the user role is a minor consultant, the concise answer data of the minor role is used as the current answer data matched with the current question data.

8. The method according to claim 1, further comprising: Determine the associated region of the current user according to the current question data; The current answer data matching the current question data is determined according to the user role of the current user and the associated region of the current user.

9. The method according to claim 8, wherein: The determining the associated region of the current user according to the current question data includes: Analyze the current question data to obtain a current question data analysis result, and determine the associated region of the current user according to the current question data analysis result; and / or Determine the driving mode for the current user to call the service macro model; The associated regional area of ​​the current user is determined by the driving mode of the current user calling the service macro model.

10. The method according to any one of claims 1 to 9, wherein: The service big model includes at least one of a medical insurance big model, an education big model, a real estate big model or a medical big model.

11. A model training method, comprising: Obtain sample data of questions asked by users with different roles; Determine the target service reference data that matches the question sample data; Determine answer sample data for the sample data of questions asked by users of different roles according to the target service reference data; Based on the user role, construct question-answer sample pairs according to the question sample data and the answer sample data; Training the service model based on sample pairs of answers to the questions; The trained service big model is used to execute the data processing method described in any one of claims 1-10.

12. A data processing device, comprising: The current question data acquisition module is used to obtain the current question data input by the current user to the service model; A user role determination module, configured to determine the user role of the current user according to the current question data; wherein the user role includes at least one of a service staff and a service consultant; The answer data determination module is used to determine the current answer data matching the current question data according to the user role of the current user.

13. The device according to claim 12, wherein: The user role determination module is also used for: Parsing the current question data to obtain a parsing result of the current question data; In the case where it is determined that the current question data analysis result includes a user role, the user role included in the current question data analysis result is used as the user role of the current user.

14. The device according to claim 13, wherein: The current question data includes text question data; the user role determination module is also used for: Performing text analysis on the text question data to obtain user association information of the current user; The role-associated information of the current user is extracted according to the user-associated information of the current user to obtain the analysis result of the current question data.

15. The device according to claim 13, wherein: The current question data includes audio and video question data; the user role determination module is also used for: Performing voiceprint analysis on the audio and video question data to obtain audio fingerprint information of the current user; The role association information of the current user is determined according to the audio fingerprint information of the current user, and the analysis result of the current question data is obtained.

16. The device according to claim 12, wherein: If the user role is the service personnel, the answer data determination module is further used for: Determine a plurality of first-classified answer data matching the current question data; wherein the first-classified answer data include standard answer data and concise answer data; The standard answer data is used as current answer data matched with the current question data.

17. The device according to claim 12, wherein: If the user role is a service consultant, the answer data determination module is further used to: Determine a plurality of first-classified answer data matching the current question data; wherein the first-classified answer data include standard answer data and concise answer data; The concise answer data is used as current answer data matched with the current question data.

18. The device according to claim 12, wherein: The service consultants include minor consultants and adult consultants; the answer data determination module is also used for: Determine a plurality of second-category answer data that match the current question data; wherein the second-category answer data includes concise answer data of adult roles and concise answer data of minor roles; If the user role is an adult consultant, the concise answer data of the adult role is used as the current answer data matched with the current question data; If the user role is a minor consultant, the concise answer data of the minor role is used as the current answer data matched with the current question data.

19. The device according to claim 12, wherein: The answer data determination module is also used for: Determine the associated region of the current user according to the current question data; The current answer data matching the current question data is determined according to the user role of the current user and the associated region of the current user.

20. The device according to claim 19, wherein The answer data determination module is also used for: Analyze the current question data to obtain a current question data analysis result, and determine the associated region of the current user according to the current question data analysis result; and / or Determine the driving mode for the current user to call the service macro model; The associated regional area of ​​the current user is determined by the driving mode of the current user calling the service macro model.

21. The device according to any one of claims 12 to 20, wherein: The service big model includes at least one of a medical insurance big model, an education big model, a real estate big model or a medical big model.

22. A model training device, comprising: A question sample data acquisition module is used to obtain question sample data of users with different roles; A target service reference data determination module, used to determine the target service reference data that matches the question sample data; An answer sample data determination module, used to determine answer sample data for the question sample data of the users with different roles according to the target service reference data; A question-answer sample pair construction module is used to construct a question-answer sample pair based on the question sample data and the answer sample data based on the user role; A service big model training module, used for training the service big model according to the sample pairs of answers to the questions; The trained service big model is used to execute the data processing method described in any one of claims 1-10.

23. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the data processing method described in any one of claims 1-10, or execute the model training method described in claim 11.

24. A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute the data processing method described in any one of claims 1 to 10, or execute the model training method described in claim 11.

25. A computer program product comprising a computer program / instructions, wherein: When the computer program / instruction is executed by a processor, it implements the data processing method described in any one of claims 1 to 10, or executes the model training method described in claim 11.

Images