Data processing method and system

Abstract

The embodiment of the invention provides a data processing method and system.The data processing method comprises the steps that an interaction request submitted by a user for a consultation event is received, and an interaction data sequence corresponding to the consultation event is collected based on the interaction request; the consultation intention of the user is determined based on the interaction data sequence, the consultation intention of the user is determined based on the interaction data sequence of the user through a context transmission mechanism, and the consultation intention determination accuracy is improved. And constructing model allocation information based on the consultation intention, and determining at least one question and answer model matched with the model allocation information in a model set. And executing an event task corresponding to the consultation event by using the question and answer model to obtain consultation feedback data, and taking the consultation feedback data as a response of the interaction request. By analyzing the consultation intention of the user in real time, the question and answer model with the high matching degree with the consultation intention is selected, calling of a model irrelevant to the consultation intention of the user is avoided, and computing resources are saved.

Description

Technical Field

[0001] The embodiments in this specification relate to the field of computer technology, and in particular to data processing methods and systems. Background Technology

[0002] In the field of intelligent information services, users submitting consultation requests through terminals to obtain professional answers has become a common need. Existing consultation service systems mostly use fixed rules or a single question-and-answer model to handle user interactions. When the consultation process involves multiple rounds of dialogue or complex scenarios, it is difficult to provide users with accurate question-and-answer services, and the generated feedback information often lacks specificity, reducing service quality and user experience. Therefore, a more effective data processing method is urgently needed to solve these problems. Summary of the Invention

[0003] In view of the above, embodiments of this specification provide a data processing method. One or more embodiments of this specification also relate to a data processing system, a data processing apparatus, a computing device, a computer-readable storage medium, and a computer program product, to address the technical deficiencies existing in the prior art.

[0004] According to a first aspect of the embodiments of this specification, a data processing method is provided, comprising: Receive user interaction requests for consultation events, and collect the interaction data sequence corresponding to the consultation event based on the interaction requests; The user's consultation intent is determined based on the interaction data sequence, and a model is constructed based on the consultation intent to allocate information; In the model set, at least one question-answering model that matches the model allocation information is determined. The question-answering model is used to execute the event task corresponding to the consultation event to obtain consultation feedback data, and the consultation feedback data is used as the response to the interaction request.

[0005] According to a second aspect of the embodiments of this specification, a data processing system is provided, including a client and a server; The client is used to receive interaction requests submitted by users in response to consultation events and send the interaction requests to the server. The server is configured to collect an interaction data sequence corresponding to the consultation event based on the interaction request; determine the user's consultation intent based on the interaction data sequence, and construct model allocation information based on the consultation intent; determine at least one question-and-answer model in the model set that matches the model allocation information; use the question-and-answer model to execute the event task corresponding to the consultation event, obtain consultation feedback data, and feed back the consultation feedback data as a response to the interaction request to the client.

[0006] According to a third aspect of the embodiments of this specification, a data processing apparatus is provided, comprising: The receiving module is configured to receive interaction requests submitted by users in response to consultation events, and to collect the interaction data sequence corresponding to the consultation event based on the interaction requests. The construction module is configured to determine the user's consultation intent based on the interaction data sequence, and to construct a model allocation information based on the consultation intent; The execution module is configured to determine at least one question-answering model in the model set that matches the model allocation information, use the question-answering model to execute the event task corresponding to the consultation event, obtain consultation feedback data, and use the consultation feedback data as a response to the interaction request.

[0007] According to a fourth aspect of the embodiments of this specification, a computing device is provided, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the above-described data processing method.

[0008] According to a fifth aspect of the embodiments of this specification, a computer-readable storage medium is provided that stores computer-executable instructions, which, when executed by a processor, implement the steps of the data processing method described above.

[0009] According to a sixth aspect of the embodiments of this specification, a computer program product is provided, including a computer program or instructions that, when executed by a processor, implement the steps of the data processing method described above.

[0010] This specification provides a data processing method in one embodiment, which receives an interaction request submitted by a user regarding a consultation event and collects an interaction data sequence corresponding to the consultation event based on the interaction request. The user's consultation intent is determined based on the interaction data sequence through a context-passing mechanism, improving the accuracy of intent determination. Model allocation information is constructed based on the consultation intent, and at least one question-and-answer model matching the allocation information is determined from the model set. The question-and-answer model is used to execute the event task corresponding to the consultation event, obtaining consultation feedback data, which is then used as a response to the interaction request. By analyzing the user's consultation intent in real time, a question-and-answer model with a high degree of matching is selected, avoiding the invocation of models unrelated to the user's consultation intent and saving computational resources. Precise data processing based on the question-and-answer model improves the accuracy of consultation feedback data and enhances the user's consultation experience. Attached Figure Description

[0011] Figure 1This is a flowchart illustrating a data processing method provided in one embodiment of this specification; Figure 2 This is an interactive schematic diagram of a data processing method provided in one embodiment of this specification; Figure 3 This is a schematic diagram of the structure of a data processing system provided in one embodiment of this specification; Figure 4 This is a schematic diagram of the structure of a data processing apparatus provided in one embodiment of this specification; Figure 5 This is a structural block diagram of a computing device provided in one embodiment of this specification. Detailed Implementation

[0012] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0013] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of this specification. The singular forms “a,” “described,” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0014] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0015] Furthermore, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties. Moreover, the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0016] The technical solutions provided in this application can employ deep learning models with relatively large parameter scales. However, this large model is merely an example; this application does not limit the number of model parameters supported by the deep learning model used, aiming to meet actual needs. The deep learning models involved in this application can be artificial intelligence-based language models (LM) or multimodal models (MM).

[0017] First, the terms and concepts used in one or more embodiments of this specification will be explained.

[0018] AI-managed services: Building upon the traditional human customer service model, this approach introduces large-scale models to assist human staff in answering messages and can automatically execute some tools.

[0019] Agent: An intelligent agent is a core concept in the field of artificial intelligence. It usually refers to a software or system entity that can perceive the environment, make autonomous decisions, and take actions to achieve specific goals.

[0020] To address the aforementioned technical problems, this specification provides a data processing method. This specification also relates to a data processing apparatus, a computing device, a computer-readable storage medium, and a computer program product, which will be described in detail in the following embodiments.

[0021] With the rapid development of large-scale model technology, large-scale models are increasingly being applied to consultation scenarios, aiming to free up manpower and provide consultation services to users, answering their questions. Existing large-scale model-based consultation services typically use a single, general-purpose large-scale model to handle all consultation scenarios, allocating and processing tasks through keyword matching or intent assignment. A single, general-purpose large-scale model cannot comprehensively cover all consultation scenarios; when a new consultation scenario is added, the large-scale model needs to be retrained, resulting in low model iteration efficiency. This specification provides a data processing method in one embodiment, which receives user interaction requests for consultation events and collects corresponding interaction data sequences based on the interaction requests. The user's consultation intent is determined based on the interaction data sequence through a contextualization mechanism, improving the accuracy of intent determination. Model allocation information is constructed based on the consultation intent, and at least one question-and-answer model matching the model allocation information is determined from the model set. The question-and-answer model is used to execute the event task corresponding to the consultation event, obtaining consultation feedback data, which is then used as the response to the interaction request. By analyzing the user's consultation intent in real time, a question-and-answer model with a high degree of matching with the consultation intent is selected, avoiding the invocation of models unrelated to the user's consultation intent and saving computational resources. By using a question-and-answer model for precise data processing, the accuracy of consultation feedback data can be improved, thereby enhancing the user's consultation experience.

[0022] See Figure 1 , Figure 1 A flowchart of a data processing method according to an embodiment of this specification is shown, which specifically includes the following steps.

[0023] Step 102: Receive the user's interaction request for the consultation event, and collect the interaction data sequence corresponding to the consultation event based on the interaction request.

[0024] Specifically, users can be those who utilize the consulting services provided by the consulting platform. These consulting services can be offered by service providers in any field or scenario. Service providers offer consulting services to users through the consulting platform. These services can be offered in various scenarios such as shopping, legal consultation, education, and Q&A, by different service providers. The service providers use the consulting platform to provide solutions to users' problems and answer their questions through communication. Users are the recipients of these consulting services. In a shopping scenario, users can be consumers; in a legal consultation scenario, users can be lawyers or those seeking legal assistance; and in an education scenario, users can be students, teachers, or parents.

[0025] A consultation event refers to the event in which a user currently uses the consultation service to inquire. Consultation events can be of various types, such as after-sales events, logistics tracking events, legal interpretation events, and math problem solving events. An interaction request can be a computer command submitted by the user through the consultation platform. Users can submit interaction requests by inputting and submitting text content, or through touch buttons on the consultation interface. An interaction request can also be a request from a user to switch to a human customer service representative while currently communicating with a non-human representative. The interaction data sequence includes the dialogue sequence generated by the user with non-human customer service before submitting the interaction request. It can also include historical dialogue sequences generated by the user on the self-selected platform, which can be the dialogue sequence of previous consultation events corresponding to the current consultation event.

[0026] Based on this, users can submit interaction requests for consultation events either by submitting text content to the consultation platform or by using touch buttons within the consultation interface. Upon receiving a user's interaction request, the server responds by collecting the corresponding interaction data sequence. This data sequence is then used to process the user's consultation event.

[0027] Furthermore, users can submit interaction object transformation information in at least one way to generate interaction requests, as detailed below: Collect the interaction object transformation information submitted by the user in response to the consultation event, and generate the interaction request based on the interaction object transformation information.

[0028] Specifically, when a user requests an interaction object conversion, they can submit interaction object conversion information through the consultation service. This information is used to change the interaction object currently interacting with the user. The interaction object can be a human interaction object (human customer service) or a non-human interaction object (non-human customer service, such as AI-powered customer service based on a large model).

[0029] Based on this, the system collects interaction object transformation information submitted by users in response to consultation events, and generates interaction requests based on this information. The interaction object transformation information submitted by users in response to consultation events can be text or voice information input by the user expressing the meaning of the object transformation, or it can be generated by the user through touch operations on controls, where the controls are the operable controls corresponding to the interaction object transformation information. Touch operations can be clicking on a control or hovering the mouse over a control, triggering the generation and submission of the interaction object transformation information.

[0030] For example, in a transaction scenario, the user is the consumer, using the trading platform to trade goods. Simultaneously, the user can enjoy the platform's consultation services. Any questions encountered by the user within the platform can be raised through these services to resolve issues. In the case of order logistics inquiries, users can interact with customer service via text or voice input on the customer service page. The trading platform typically assigns a non-human customer service representative to the user upon their first inquiry. The user can submit an interaction request by typing "transfer to human" or by using the corresponding control, aiming to change the current interaction partner. To further improve the consultation service, the trading platform's server responds to the interaction request by collecting data on the current inquiry event, specifically the interaction data sequence. This includes the dialogue content between the user and the non-human customer service representative regarding the inquiry event, as well as the dialogue content of historical inquiry events.

[0031] In summary, by generating interaction requests based on interaction object conversion information, and allowing users to submit interaction object conversion information through at least one method, the system enables timely acquisition of users' object conversion needs.

[0032] Furthermore, considering that user inquiries can be handled by AI-managed systems, but AI-managed systems have limitations in their event handling capabilities—inquiries exceeding these limitations cannot be processed by AI—the event attributes of the inquiries can be detected before collecting the interaction data sequence. The specific implementation is as follows: Collect user data for the user and determine the event attribute corresponding to the consultation event based on the user data; if the event attribute set contains the event attribute, perform the step of collecting the interaction data sequence corresponding to the consultation event based on the interaction request.

[0033] Specifically, user data includes, but is not limited to, user dialogue context, user intent, and user historical behavior data. The event attribute corresponding to the consultation event can be the event type. The event attribute set corresponds to the event processing scope. If the event attribute belongs to the event attribute set, it means that the event type of the consultation event is within the AI-managed event processing scope. In this case, the consultation event can be handled by AI without being transferred to manual processing, and the AI-managed mode can be activated subsequently, along with the collection of interaction data sequences.

[0034] Based on this, user data such as dialogue context, user intent, and historical behavior are collected, and the event attributes corresponding to the consultation event are determined based on this data. If the event attribute set contains the specified event attribute, it indicates that the event type of the consultation event falls within the scope of AI-managed event processing. The consultation event does not need to be handled by human customer service but can be handled by AI. Once in AI-managed mode, the interaction data sequence corresponding to the consultation event is collected based on the interaction request to facilitate AI-managed processing of the user's consultation event. If the event attribute set does not contain the specified event attribute, it indicates that the event type of the consultation event is not within the scope of AI-managed event processing, and a human customer service representative can be assigned to handle the consultation event.

[0035] Continuing with the previous example, when the server receives a user's "transfer to human" interaction request, it can perform an AI-managed access assessment before transferring the request to a human customer service representative. Based on the analysis of the user's historical behavior, current dialogue context, and user intent, it determines whether the type of the inquiry falls within the preset scope of scenarios suitable for AI-managed processing. If the user's inquiry pertains to a shipped order, and the semantics of the conversation between the user and the non-human customer service representative clearly involve logistics inquiries, the inquiry can be classified as a logistics scenario inquiry, falling within the scope of AI-managed events. Therefore, it can be handled by AI instead of transferring to a human customer service representative. Once in AI-managed mode, the server collects the user's interaction data sequence to facilitate subsequent processing of the inquiry. The AI-managed logistics scenario solution includes an undelivered shipment status. If the order is in a delivered status, AI-managed processing is not capable of handling the inquiry, and it must be handled by a human customer service representative.

[0036] In summary, when the event attribute set contains event attributes, the interaction data sequence corresponding to the consultation event is collected based on the interaction request to ensure that subsequent user consultation events can be processed in a timely and smooth manner.

[0037] Step 104: Determine the user's consultation intent based on the interaction data sequence, and construct a model to allocate information based on the consultation intent.

[0038] Specifically, after receiving the user's interaction request for a consultation event and collecting the corresponding interaction data sequence based on the request, the user's consultation intent can be determined based on the interaction data sequence. Model allocation information can then be constructed based on this intent. The user's consultation intent represents the purpose of the consultation event and the event type. The model allocation information is used to determine the large model matching the consultation intent. This information can be a routing instruction (ROUTE_TO_LOGISTICS_QUERY_MODEL) or an instruction tag, used to record the large model type or identifier corresponding to the consultation intent.

[0039] Based on this, after receiving the user's interaction request for the consultation event and collecting the corresponding interaction data sequence based on the interaction request, the user's consultation intent is determined by analyzing the interaction data sequence, and model allocation information is constructed based on the consultation intent so that it can be routed / mapped to the corresponding large model in the future.

[0040] Furthermore, considering that the interaction data sequences collected for consultation events contain a large amount of data, directly analyzing the interaction data sequences to determine the consultation intent cannot obtain accurate consultation intent. Therefore, the interaction data sequences can be processed based on a text processing model, as follows: The interaction data sequence is extracted using a text processing model to obtain an interaction data summary corresponding to the interaction data sequence; the interaction data summary is then input into the text processing model to obtain the user's consultation intent.

[0041] Specifically, the text processing model includes a voice chat message summarization model and an intent summarization model. The interaction data summary is a dialogue summary generated by extracting key information points from the interaction data sequence; that is, it removes noise from the interaction data sequence, focuses on the core question, and then determines the interaction data summary. The interaction data summary can be the core question extracted from the interaction data sequence. The voice chat message summarization model is used to process the interaction data sequence and generate the interaction data summary, while the intent summarization model is used to summarize the intent in the interaction data summary and generate the consultation intent. The consultation intent can represent the problem that the user needs to solve and the consultation purpose for the consultation event.

[0042] Based on this, the voice chat message summarization model included in the text processing model is used to extract key information points from the interaction data sequence, and an interaction data summary corresponding to the interaction data sequence is generated based on these key information points. The interaction data summary is then input into the intent summarization model included in the text processing model to obtain the user's consultation intent.

[0043] Continuing with the previous example, the interaction data sequence, representing the dialogue between the user and a non-human customer service representative, records multiple questions raised by the user and the responses from the non-human representative, forming a multi-turn dialogue. A voice chat message summarization model can be used to extract key information points from the interaction data sequence, generating an interaction data summary corresponding to the sequence, resulting in a concise dialogue summary. Based on this summary, an intent summarization model is further used to analyze the summary and identify the user's core inquiry intent. Given that the interaction data summary is "Why hasn't my package arrived yet? It's been three days!", the identified inquiry intent is to check the logistics status.

[0044] In summary, by inputting the interaction data summary into the text processing model, the user's consultation intent is obtained, and the consultation intent is generated based on the concise interaction data summary, thus ensuring the accuracy of the consultation intent.

[0045] Furthermore, considering the possibility that user interaction data sequences may not be available, or that these sequences may not contain any data, it may be impossible to determine the user's consultation intent based on the interaction data sequences. The specific implementation is as follows: If the interaction data sequence does not meet the conditions for generating consultation intent, the consultation intent of the user is determined by interacting with the user; if the number of interactions with the user is greater than the interaction number threshold, a target question and answer service is assigned to the user.

[0046] Specifically, if the interaction data sequence does not meet the conditions for generating consultation intent, it means that consultation intent cannot be generated based on the interaction data sequence, or that the interaction data sequence does not contain processable data. Interacting with a user can involve asking the user questions, that is, proactively initiating questions to clarify the user's consultation intent regarding the consultation event. The number of interactions with the user can be considered as the number of dialogue rounds.

[0047] Therefore, if the interaction data sequence does not meet the conditions for generating consultation intent, it means that consultation intent cannot be generated based on the interaction data sequence, or the interaction data sequence does not contain processable data. In this case, the user's consultation intent cannot be determined based on the interaction data sequence. Instead, it can be determined through interaction with the user; that is, by proactively asking the user questions related to their intent and clarifying the user's consultation intent through dialogue. If the event type corresponding to the consultation intent determined through user interaction is not in the event attribute set, it means that the event type of the consultation event is not within the scope of AI-managed event processing. A target Q&A service needs to be assigned to the user, which can be a human customer service representative. If the number of interactions with the user exceeds the interaction threshold and the consultation intent has not been determined, it means that a clear consultation intent has still not been determined after multiple rounds of dialogue, and a target Q&A service is assigned to the user.

[0048] Continuing with the previous example, if a user hasn't spoken with a non-human customer service representative before submitting an interaction request, the interaction data sequence won't contain data that clarifies the user's intent regarding the inquiry. In this case, the user's intent can be clarified by asking questions or providing options. Once AI-managed, if the user's intent cannot be determined based on the interaction data sequence, the Agent initiates the conversation by asking, "What question would you like to ask?" to clarify the user's intent. If the user clearly stated their intent before speaking with the Agent (e.g., "The logistics are too slow, I want a refund"), the intent can be further clarified with, "Would you like to ask about 'logistics' and get a refund?". If, after multiple rounds of conversation, the user's intent still cannot be clarified, human customer service takes over, and the AI-managed process ends.

[0049] In summary, when the interaction data sequence does not meet the conditions for generating consultation intent, the user's consultation intent can be determined by interacting with the user. This ensures that the user's consultation intent is clear, which is then used to match the question-and-answer model based on the user's consultation intent.

[0050] Step 106: Determine at least one question-answering model in the model set that matches the model allocation information, use the question-answering model to execute the event task corresponding to the consultation event, obtain consultation feedback data, and use the consultation feedback data as the response to the interaction request.

[0051] Specifically, after determining the user's consultation intent based on the interaction data sequence and constructing model allocation information based on the consultation intent, at least one question-and-answer model matching the model allocation information can be determined from the model set. The question-and-answer model is then used to execute the event task corresponding to the consultation event, obtain consultation feedback data, and use this feedback data as a response to the interaction request. The model set can be a pluggable model pool containing various types of large models, each used to solve different types of problems. Each large model in the model pool is an independent execution unit used to handle different specific consultation events, and each model can be independently deactivated and deployed. Based on the consultation event, a dedicated model for handling the consultation event, i.e., a question-and-answer model, can be selected from the model set. The event task corresponding to the consultation event can be a solution provision task, i.e., providing a solution to the user's problem; the event task can also be a query task, such as querying orders or logistics information. The consultation feedback data is the answer data output by the question-and-answer model for the consultation event, processed using the event task. The consultation feedback data can also be the solution for handling the consultation event.

[0052] Based on this, after determining the user's consultation intent based on the interaction data sequence and constructing model allocation information based on the consultation intent, at least one question-answering model matching the model allocation information is determined in the model set. This question-answering model is specifically designed to handle consultation events. The question-answering model is used to execute the event task corresponding to the consultation event, obtain consultation feedback data, and use this feedback data as a response to the interaction request. The consultation feedback data is then displayed to the user to address the user's consultation event and provide consultation services for that event.

[0053] In practical applications, after constructing model allocation information based on the consultation intent, at least one question-and-answer model matching the model allocation information can be determined from the model set. If the at least one question-and-answer model matching the model allocation information is a single model, the event task corresponding to the consultation event can be executed using that single model. If the at least one question-and-answer model matching the model allocation information is two or more models, the event task corresponding to the consultation event can be executed collaboratively using these two or more models. After generating consultation feedback data, it can be determined whether to exit AI hosting. If yes, the AI ​​hosting exit process is triggered, i.e., AI hosting is exited, and the consultation feedback data is sent to the user's client for display. If no, the generated consultation feedback data is sent to the user's terminal for display, and the system awaits the user's next instruction or the next round of input.

[0054] Furthermore, the model set contains large models for handling different event types. Based on the model allocation information, the question-answering model for handling consultation events can be determined from the model set. The specific implementation is as follows: Determine the allocation label corresponding to the model allocation information; determine the target model information in the model information record table based on the allocation label, and determine the at least one question-answering model that matches the model allocation information in the model set based on the target model information.

[0055] Specifically, the allocation tags corresponding to the model allocation information represent routing instructions or instruction tags, used for tag matching in the model information record table to map them to the question-answering models in the model set. The model information table records the model identifiers of large models that handle different event types; the model identifier can be the name or ID of the large model. The model information table also records the instruction tags corresponding to the model identifiers. By searching the model information table based on the instruction tags, the question-answering model corresponding to the instruction tags can be determined. The determination of target model information and the search and matching of question-answering models can both be performed by the model distributor. As the routing decision center, the model distributor receives model allocation information, searches the model information record table based on the model allocation information, and maps the model allocation information to the question-answering models in the model set.

[0056] Based on this, the allocation tag corresponding to the model allocation information is determined, and the allocation tag is forwarded to the model distributor. The model distributor searches the model information record table based on the allocation tag, determines the target model information in the model information record table, and then determines at least one question-answering model in the model set that matches the model allocation information based on the target model information, thus achieving the mapping of question-answering models and quickly determining the question-answering model in the model set. When determining at least one question-answering model matching the model allocation information in the model set, this at least one question-answering model can be one, two, or more question-answering models.

[0057] Following the previous example, after determining the model allocation information, it can be forwarded to the model distributor. The model distributor is a lightweight routing decision center. It receives routing instructions (model allocation information) generated by the "instruction allocation model" and, according to a preset routing table or configuration file (model information record table), maps the instruction to one or more specific problem-solving models in the "large model array" (model set). The distributor itself does not contain complex logic; its core function is table lookup and forwarding, ensuring fast and efficient routing. The model set contains multiple problem-solving models, each corresponding to a different event type: "Specific problem-solving model 1: specifically handles logistics inquiries; Specific problem-solving model 2: specifically handles refund applications; Specific problem-solving model 3: specifically handles invoice issuance; Specific problem-solving model 4: specifically handles product recommendations." After determining the question-and-answer model, user-related data and consultation event-related data can be sent to the question-and-answer model to execute the event tasks for the consultation event. If the model allocation information matches specific problem-solving model 1, then specific problem-solving model 1 is taken as the determined question-answering model; if the model allocation information matches specific problem-solving model 1 and specific problem-solving model 2 at the same time, then specific problem-solving model 1 and specific problem-solving model 2 are taken as the determined question-answering model. Specific problem-solving model 1 and specific problem-solving model 2 constitute a model combination, and specific problem-solving model 1 and specific problem-solving model 2 collaborate to process user-related data and consultation event-related data.

[0058] In summary, the model dispatcher determines at least one question-answering model in the model set that matches the model allocation information based on the target model information, thereby improving the matching efficiency of question-answering models.

[0059] Furthermore, when adding new consulting services targeting specific problem types, a target problem handling model can be trained for that target problem type and stored in a model set to enable the online deployment of consulting services for that target problem type. The specific implementation is as follows: Identify the new target question types for the consulting service, train the target question processing model corresponding to the target question type; store the target question processing model in the model set, and store the target question processing model information corresponding to the target question processing model in the model information record table.

[0060] Specifically, the target problem type can be a newly added consultation service type that can resolve consultation events of the target problem type. The target problem processing model is a large language model trained specifically for consultation services of the target problem type. The target problem processing model is dedicated to processing consultation events of the event type corresponding to the target problem type. The target problem processing model information includes the model identifier of the target problem processing model and the consultation event type (event attribute) corresponding to the target problem type.

[0061] Based on this, the target question types newly added to the consulting service are determined. The newly added target question types correspond to newly added service types in the consulting service. A target question processing model is trained for each target question type; this model is specifically designed to handle consulting events corresponding to the target question type. The target question processing model is stored in a model set, and its corresponding information is stored in a model information record table. This allows the system to provide consulting services corresponding to the target question type when a user has a consulting event. In other words, if a user's consulting event matches the consulting type corresponding to the target question type, the system can search the model information record table to match the user with the appropriate target question processing model, thus providing accurate consulting services.

[0062] Continuing with the previous example, when a shopping platform needs to support a new consultation service (adding an "after-sales repair application" service), the platform's administrators can independently develop and train a new "specific problem-solving model," namely, the repair application model (target problem handling model), for the new after-sales repair application service. This repair application model is stored in the model set. A rule is added to the "instruction allocation model" to enable it to recognize the user's "repair application" intent and generate a new routing instruction: "ROUTE_TO_REPAIR_SERVICE_MODEL". A mapping is added to the routing table of the "model distributor" to point the new routing instruction to the newly launched repair application model, thus providing the after-sales repair application service to the user. Adding the repair application service does not require modifying or redeploying the core main link code; only incremental deployment of the new model and updating the configuration file are needed to complete the launch of the new consultation service, achieving extremely high iteration efficiency. The voice chat distribution recognition model and the model distributor constitute a stable main link, while the specific problem-solving models in the large model array (model set) are pluggable, independent execution units. The main link and the execution unit communicate through standardized routing instructions, achieving logical decoupling.

[0063] In summary, by storing the target problem handling model in a model set and storing the target problem handling model information in a model information record table, the consulting services corresponding to the target problem type can be launched without updating the existing large models in the model set, thus enabling rapid launch of consulting services corresponding to the target problem type and improving iteration efficiency.

[0064] Furthermore, when deleting pending issue types in the consulting service, the pending issue model can be taken offline by deleting the pending issue model corresponding to the pending issue type from the model set. The specific implementation is as follows: Identify the types of pending issues to be deleted for the consulting service, delete the pending issue model corresponding to the pending issue type from the model set, and delete the pending issue model information corresponding to the pending issue model from the model information record table.

[0065] Specifically, the type of problem to be processed can be a service type that has been taken offline, indicating that the service type has been discontinued or replaced by another service type. The problem model corresponding to the problem type is the large model of the consultation event for the user to handle the problem type. The problem model information is the mapping information recorded in the model information record table, used to map to the problem models in the model set.

[0066] Based on this, the pending issue types to be deleted from the consulting service are identified. Deletion of a pending issue type from the consulting service indicates that the corresponding consulting service type has been taken offline or replaced by another consulting service type. The pending issue model corresponding to the pending issue type is deleted from the model set, the pending issue model information corresponding to the pending issue model is deleted from the model information record table, and the routing instruction conversion rule corresponding to the pending issue type is deleted from the instruction allocation model.

[0067] Following the previous example, when a consulting service type corresponding to the type of problem to be processed is taken offline, or needs to be replaced by another consulting service type, the corresponding mapping rules can be removed or modified in the routing table (model information record table) of the "Model Distributor," and the old model service (problem model to be processed) can be selectively taken offline. The model taking offline process has no impact on existing consulting services.

[0068] In summary, deleting the information of the pending problem model corresponding to the pending problem model in the model information record table enables flexible management of the models in the model set. It allows for convenient deletion of models in the model set, achieving the purpose of taking models offline and consulting event types offline or updating, without affecting the functionality of other models.

[0069] This specification provides a data processing method in one embodiment, which receives an interaction request submitted by a user regarding a consultation event and collects an interaction data sequence corresponding to the consultation event based on the interaction request. The user's consultation intent is determined based on the interaction data sequence through a context-passing mechanism, improving the accuracy of intent determination. Model allocation information is constructed based on the consultation intent, and at least one question-and-answer model matching the allocation information is determined from the model set. The question-and-answer model is used to execute the event task corresponding to the consultation event, obtaining consultation feedback data, which is then used as a response to the interaction request. By analyzing the user's consultation intent in real time, a question-and-answer model with a high degree of matching is selected, avoiding the invocation of models unrelated to the user's consultation intent and saving computational resources. Precise data processing based on the question-and-answer model improves the accuracy of consultation feedback data and enhances the user's consultation experience.

[0070] The following is in conjunction with the appendix Figure 2 Taking the application of the data processing method provided in this specification in an intelligent customer service scenario as an example, the data processing method will be further explained. Among other things, Figure 2 An interactive schematic diagram of a data processing method provided in one embodiment of this specification is shown, which specifically includes the following steps.

[0071] Step 202: The client collects the interaction object transformation information submitted by users of the associated consultation service in response to consultation events, generates an interaction request based on the interaction object transformation information, and sends the interaction request to the server.

[0072] The data processing method provided in one embodiment of this specification can be applied to intelligent customer service consultation scenarios. The client includes a target platform for users to interact with intelligent customer service. The target platform can correspond to a webpage or application. The client is the user terminal, i.e., the mobile device held by the user. The consultation service is the service provided by the target platform. The target platform can be any platform that can provide consultation services to users, such as a shopping platform, learning platform, legal consultation platform, medical consultation platform, etc. In the case that the target platform is a shopping platform, the consultation event can be any event that can be handled by customer service, such as a logistics consultation event or a return consultation event. The user-initiated interaction object conversion information can be used to convert the interaction object, such as converting from a non-human customer service representative to a human customer service representative.

[0073] In practical applications, when a user connects through online channels, they are initially greeted by a standard automated customer service representative. During the conversation, if the user expresses an intention to transfer to a human agent, or if the system determines that a transfer to a human agent is necessary based on the user's actions, a "transfer to human agent" instruction is triggered, i.e., an interactive request.

[0074] Step 204: The server collects user data and performs a hosting access judgment based on the user data. If the judgment is successful, the server enters the hosting state.

[0075] Upon receiving the "transfer to human" instruction, the server does not directly assign a human customer service representative to the user. Instead, it determines whether the current scenario is suitable for AI-assisted customer service. At this point, it first collects user data such as historical behavior, current context, and user intent. Based on this data, it performs a customer service admission judgment: whether the user's inquiry exceeds the capabilities of AI-assisted customer service. If it does not, the user can enter AI-assisted customer service mode. If it does, the user is then handed over to a human customer service representative.

[0076] Step 206: The server collects the interaction data sequence corresponding to the consultation event stored on the client.

[0077] Once in AI-hosted mode, the server needs to collect user interaction data related to this consultation event, i.e., the interaction data sequence, to clarify information such as the user's orders and requests. The collected interaction data sequence can represent the user's historical communication records, i.e., the context of the user's communication.

[0078] Step 208: The server uses a text processing model to extract data from the interaction data sequence, obtaining an interaction data summary corresponding to the sequence. This summary is then input into the text processing model to obtain the user's consultation intent. The consultation intent is then converted into model allocation information.

[0079] The server can extract user inquiry intent based on the interaction data sequence. Specifically, it uses a voice chat message summary model to read and analyze lengthy, conversational dialogue history, extracting key information points and generating a concise dialogue summary to remove noise and focus on the core issue. Then, an intent summary model further analyzes the dialogue summary to identify and summarize the user's core inquiry intent. For example, from "Why hasn't my package arrived yet? It's been three days!", the inquiry intent of "Checking the logistics status" can be summarized. Finally, an instruction assignment model converts the inquiry intent into a recognizable, standardized routing instruction (or instruction label). For example, the intent of "Checking the logistics status" can be converted into an instruction like "ROUTE_TO_LOGISTICS_QUERY_MODEL," which serves as the model's assignment information. This instruction is a crucial input to the subsequent model dispatcher.

[0080] Furthermore, when the user's consultation intent cannot be determined based on the interaction data sequence—that is, when the user's consultation intent is vague or unclear—it is necessary to guide the user through follow-up questions to obtain a clear consultation intent. This can be achieved through clarifying questions, providing options, etc., to guide the user to clarify their needs until a clear consultation intent can be found. AI-managed services can proactively initiate questions to confirm the user's order or intent. If, after multiple rounds of dialogue, the user's request is still outside the scope of AI-managed services, or if the user makes a request that AI-managed services cannot handle, the user will directly exit AI-managed services.

[0081] Step 210: The server determines the question-answering model from the model set based on the model allocation information.

[0082] Once the model allocation information in the form of the instruction is determined, the model dispatcher can be invoked. Based on a pre-defined configuration file, the instruction is mapped to one or more specific problem-solving models (i.e., question-answering models) within the large model set. The configuration file records the correspondence between the large models and the model allocation information. By searching the configuration file, a large model capable of handling consultation events can be identified within the large model set.

[0083] The large model set is a pool of models consisting of multiple specific problem-solving models. Each model is specifically trained to solve a vertical problem in a particular domain. Once the model dispatcher determines the question-answering model (e.g., calling "specific problem-solving model 1" based on instructions), it passes relevant contextual information (such as dialogue summary, order number, etc.) to that model. The selected model then executes the event task corresponding to the consultation event, engages in dialogue with the user, and generates a solution.

[0084] Each specific problem-solving model in the large model set is a pluggable, independent execution unit. The model dispatcher communicates with the corresponding task execution units of each model through standardized routing instructions, achieving logical decoupling. When a new consultation scenario needs to be supported (e.g., adding a "after-sales repair application" service), a new specific problem-solving model (repair application model) can be independently developed and trained. A rule is added to the instruction allocation model to enable it to recognize the "repair application" intent and generate a new routing instruction, such as "ROUTE_TO_REPAIR_SERVICE_MODEL". A mapping is then added to the model dispatcher's configuration file to point the new instruction to the newly deployed model. The entire process requires no modification or redeployment of the core main chain code; incremental deployment of the new model and updating the configuration file completes the deployment of new service capabilities, achieving extremely high iteration efficiency. Conversely, when a model is taken offline—that is, when a consultation scenario is no longer needed or needs to be replaced—simply remove or modify the corresponding mapping rule in the "model dispatcher" configuration file, and selectively take the old model service offline. The entire process has no impact on existing models and other services.

[0085] Step 212: The server uses the question-and-answer model to execute the event task corresponding to the consultation event and obtain consultation feedback data.

[0086] The server can call the core Agent (a managed large model service) to execute the event tasks corresponding to the consultation event using the question-and-answer model, and generate preliminary solutions (consultation feedback data) for the consultation event.

[0087] Step 214: The server sends the consultation feedback data back to the client as a response to the interaction request.

[0088] The server sends the consultation feedback data as a response to the interaction request to the client, where the user can then view the consultation feedback data in the client's dialog box.

[0089] Once the AI-managed process completes its task (e.g., the question has been answered, the user has no further questions) or triggers other exit rules during execution (e.g., the user expresses strong dissatisfaction or requests human intervention), the process ends and the AI-managed process exits.

[0090] In summary, the data processing method provided in this embodiment achieves multi-level parsing (message summarization → intent recognition → command generation → model routing) through a "voice chat message summary model" + "model distributor," supporting accurate routing in complex scenarios and solving the flexibility problem of traditional static routing. The general model is split into specialized models for vertical scenarios (such as logistics query models and refund models), with each model undergoing deep training for a specific domain, improving the accuracy of vertical scenario processing and reducing redundant computations in the general model. Model distribution (matching) and task execution (specialized models) are completely decoupled, supporting independent development, deployment, and decommissioning of models. This reduces the time to launch new models from weeks to hours, significantly lowering operational costs. The complete dialogue history is passed through context, and the task execution result is fed back after Agent execution, forming a multi-turn dialogue loop, improving the success rate of multi-turn dialogues and avoiding erroneous responses caused by context breaks. Dynamic judgments are made at the hosting entry point (admission) and execution process (exit) to determine in real time whether to exit AI hosting. This reduces the need for manual intervention for complex issues and improves user experience.

[0091] Corresponding to the above method embodiments, this specification also provides data processing system embodiments. Figure 3 A schematic diagram of the structure of a data processing system according to one embodiment of this specification is shown. Figure 3 As shown, the data processing system 300 includes a client 310 and a server 320. The client 310 is used to receive an interaction request submitted by a user for a consultation event and send the interaction request to the server 320. The server 320 is used to: collect an interaction data sequence corresponding to the consultation event based on the interaction request; determine the user's consultation intent based on the interaction data sequence and construct model allocation information based on the consultation intent; determine at least one question-and-answer model matching the model allocation information in the model set; use the question-and-answer model to execute the event task corresponding to the consultation event, obtain consultation feedback data, and feed the consultation feedback data back to the client 310 as a response to the interaction request.

[0092] In practical applications, the client can generate interaction requests based on user actions; that is, it receives interaction requests submitted by the user regarding a consultation event and sends these requests to the server. The server collects the interaction data sequence corresponding to the consultation event based on the interaction requests. Based on the interaction data sequence, the server determines the user's consultation intent. Through a context-passing mechanism, the server improves the accuracy of intent determination. Based on the consultation intent, the server constructs model allocation information and identifies at least one question-and-answer model that matches the allocation information from the model set. The server then uses the question-and-answer model to execute the event task corresponding to the consultation event, obtains consultation feedback data, and sends this feedback data back to the client as a response to the interaction request. By analyzing the user's consultation intent in real time, the server selects a question-and-answer model with a high degree of matching, avoiding the invocation of models unrelated to the user's intent and saving computational resources. Precise data processing based on the question-and-answer model improves the accuracy of consultation feedback data and enhances the user's consultation experience.

[0093] The above is an illustrative scheme of a data processing system according to this embodiment. It should be noted that the technical solution of this data processing system and the technical solution of the data processing method described above belong to the same concept. For details not described in detail in the technical solution of the data processing system, please refer to the description of the technical solution of the data processing method described above.

[0094] Corresponding to the above method embodiments, this specification also provides data processing apparatus embodiments. Figure 4 A schematic diagram of the structure of a data processing apparatus according to one embodiment of this specification is shown. Figure 4 As shown, the device includes: The receiving module 402 is configured to receive an interaction request submitted by a user in response to an inquiry event, and to collect an interaction data sequence corresponding to the inquiry event based on the interaction request. The construction module 404 is configured to determine the user's consultation intent based on the interaction data sequence, and to construct a model allocation information based on the consultation intent; The execution module 406 is configured to determine at least one question-answering model in the model set that matches the model allocation information, use the question-answering model to execute the event task corresponding to the consultation event, obtain consultation feedback data, and use the consultation feedback data as a response to the interaction request.

[0095] In an optional embodiment, the receiving module 402 is further configured to: Collect the interaction object transformation information submitted by the user in response to the consultation event, and generate the interaction request based on the interaction object transformation information.

[0096] In an optional embodiment, the receiving module 402 is further configured to: Collect user data for the user and determine the event attributes corresponding to the consultation event based on the user data; If the event attribute is included in the event attribute set, the step of collecting the interaction data sequence corresponding to the consultation event based on the interaction request is executed.

[0097] In an optional embodiment, the building module 404 is further configured to: The interactive data sequence is extracted using a text processing model to obtain an interactive data summary corresponding to the interactive data sequence. The interaction data summary is input into the text processing model to obtain the user's consultation intent.

[0098] In an optional embodiment, the building module 404 is further configured to: If the interaction data sequence does not meet the conditions for generating consultation intent, the consultation intent of the user is determined by interacting with the user. If the number of interactions with the user exceeds a threshold, a target question-and-answer service is assigned to the user.

[0099] In an optional embodiment, the execution module 406 is further configured to: Determine the allocation label corresponding to the model allocation information; Based on the assigned label, the target model information is determined in the model information record table, and based on the target model information, the question-answering model that matches the model assignment information is determined in the model set.

[0100] In an optional embodiment, the execution module 406 is further configured to: Identify the new target question types for consulting services, and train the target question processing model corresponding to the target question types; The target problem handling model is stored in the model set, and the target problem handling model information corresponding to the target problem handling model is stored in the model information record table.

[0101] In an optional embodiment, the execution module 406 is further configured to: Determine the types of pending issues to be deleted for the consulting service, and delete the pending issue models corresponding to the types of pending issues from the model set; Delete the information of the problem model to be processed corresponding to the problem model to be processed from the model information record table.

[0102] This specification provides a data processing apparatus in one embodiment that receives an interaction request submitted by a user regarding a consultation event and collects an interaction data sequence corresponding to the consultation event based on the interaction request. The user's consultation intent is determined based on the interaction data sequence through a context passing mechanism, improving the accuracy of intent determination. Model allocation information is constructed based on the consultation intent, and at least one question-and-answer model matching the model allocation information is determined from a model set. The question-and-answer model is used to execute the event task corresponding to the consultation event, obtaining consultation feedback data, which is then used as a response to the interaction request. By analyzing the user's consultation intent in real time, a question-and-answer model with a high degree of matching to the intent is selected, avoiding the invocation of models unrelated to the user's intent and saving computational resources. Precise data processing based on the question-and-answer model improves the accuracy of consultation feedback data and enhances the user's consultation experience.

[0103] The above is an illustrative scheme of a data processing apparatus according to this embodiment. It should be noted that the technical solution of this data processing apparatus and the technical solution of the data processing method described above belong to the same concept. For details not described in detail in the technical solution of the data processing apparatus, please refer to the description of the technical solution of the data processing method described above.

[0104] Figure 5 A structural block diagram of a computing device 500 according to one embodiment of this specification is shown. The components of the computing device 500 include, but are not limited to, a memory 510 and a processor 520. The processor 520 is connected to the memory 510 via a bus 530, and a database 550 is used to store data.

[0105] The computing device 500 also includes an access device 540, which enables the computing device 500 to communicate via one or more networks 560. Examples of these networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 540 may include one or more of any type of wired or wireless network interface (e.g., a network interface card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, or a Near Field Communication (NFC) interface.

[0106] In one embodiment of this specification, the above-described components of the computing device 500 and Figure 5 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 5 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0107] The computing device 500 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). The computing device 500 can also be a mobile or stationary server.

[0108] The processor 520 is configured to execute the following computer-executable instructions, which, when executed by the processor, implement the steps of the above-described data processing method.

[0109] The above is an illustrative scheme of a computing device according to this embodiment. It should be noted that the technical solution of this computing device and the technical solution of the data processing method described above belong to the same concept. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solution of the data processing method described above.

[0110] An embodiment of this specification also provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the above-described data processing method.

[0111] The above is an illustrative scheme of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium and the technical solution of the data processing method described above belong to the same concept. For details not described in detail in the technical solution of the storage medium, please refer to the description of the technical solution of the data processing method described above.

[0112] An embodiment of this specification also provides a computer program product, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described data processing method.

[0113] The above is an illustrative scheme of a computer program product according to this embodiment. It should be noted that the technical solution of this computer program product and the technical solution of the data processing method described above belong to the same concept. For details not described in detail in the technical solution of the computer program product, please refer to the description of the technical solution of the data processing method described above.

[0114] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0115] The computer program / instructions include computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added or removed according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0116] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments in this specification are not limited to the described order of actions, because according to the embodiments in this specification, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments in this specification.

[0117] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0118] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the embodiments described herein. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the embodiments, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. A data processing method, comprising: receiving an interaction request submitted by a user for a consultation event, and collecting an interaction data sequence corresponding to the consultation event based on the interaction request; determining a consultation intention of the user based on the interaction data sequence, and constructing model allocation information based on the consultation intention; determining at least one question and answer model matching the model allocation information in a model set, executing an event task corresponding to the consultation event by using the question and answer model, obtaining consultation feedback data, and taking the consultation feedback data as a response to the interaction request.

2. The data processing method of claim 1, wherein the receiving an interaction request submitted by a user for a consultation event comprises: collecting interaction object conversion information submitted by the user for the consultation event, and generating the interaction request based on the interaction object conversion information.

3. The data processing method of claim 1, wherein before the collecting an interaction data sequence corresponding to the consultation event based on the interaction request, the method further comprises: collecting user data for the user, and determining an event attribute corresponding to the consultation event based on the user data; in a case where the event attribute is included in an event attribute set, executing the collecting an interaction data sequence corresponding to the consultation event based on the interaction request.

4. The data processing method of claim 1, wherein the determining a consultation intention of the user based on the interaction data sequence comprises: performing data extraction on the interaction data sequence by using a text processing model, to obtain an interaction data summary corresponding to the interaction data sequence; inputting the interaction data summary into the text processing model, to obtain the consultation intention of the user.

5. The data processing method of claim 1, wherein the determining a consultation intention of the user based on the interaction data sequence comprises: in a case where the interaction data sequence does not satisfy a consultation intention generation condition, determining the consultation intention of the user by interacting with the user; in a case where a number of interactions with the user is greater than an interaction number threshold, assigning a target question and answer service to the user.

6. The data processing method of claim 1, wherein the determining at least one question and answer model matching the model allocation information in a model set comprises: determining an allocation label corresponding to the model allocation information; determining target model information in a model information record table based on the allocation label, and determining the at least one question and answer model matching the model allocation information in the model set based on the target model information.

7. The data processing method of claim 6, further comprising: determining a target problem type newly added for a consultation service, and training a target problem processing model corresponding to the target problem type; storing the target problem processing model to the model set, and storing target problem processing model information corresponding to the target problem processing model to the model information record table.

8. The data processing method of claim 7, further comprising: determine a type of the pending question to be deleted for the consulting service, and delete a pending question model corresponding to the type of the pending question in the model set; delete pending question model information corresponding to the pending question model in the model information record table. 9.A data processing system, comprising a client and a server; the client is configured to receive an interaction request submitted by a user for a consulting event, and send the interaction request to the server; the server is configured to collect an interaction data sequence corresponding to the consulting event based on the interaction request; determine a consulting intention of the user based on the interaction data sequence, and construct model allocation information based on the consulting intention; determine at least one question and answer model matching the model allocation information in a model set, execute an event task corresponding to the consulting event by using the question and answer model, obtain consulting feedback data, and feed back the consulting feedback data as a response to the interaction request to the client. 10.A computing device, comprising: a memory and a processor; the memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions, which realize the steps of the method in any one of claims 1 to 8 when executed by the processor. 11.A computer readable storage medium, which stores computer executable instructions, and the computer executable instructions realize the steps of the method in any one of claims 1 to 8 when executed by a processor. 12.A computer program product, comprising a computer program or instructions, and the computer program or instructions realize the steps of the method in any one of claims 1 to 8 when executed by a processor.

Images