Task processing method and device, electronic equipment and readable storage medium

Abstract

The invention discloses a task processing method and device, electronic equipment and a readable storage medium. The method comprises the steps that a task query request of a client side is obtained, and the task query request is used for representing a to-be-queried task of the client side; based on the task query request, identifying a task category to which the to-be-queried task belongs; an intelligent agent corresponding to the task category is called, a multi-round dialogue mode is triggered, query information of the to-be-queried task is obtained, and the query information is used for representing the query requirement of the to-be-queried task; querying the query information by using the intelligent agent to obtain a task query result; and sending the task query result to the client. The technical problem that the task processing effect is poor is solved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and more specifically, to a task processing method, apparatus, electronic device, and readable storage medium. Background Technology

[0002] Against the backdrop of digital transformation, enterprise intelligent service assistants have become one of the key technologies for improving internal operational efficiency and optimizing employee experience. Enterprise intelligent service assistants rely on rule engine systems, retrieval-based question-and-answer databases, and traditional machine learning models to automate tasks related to human resources, finance, information technology (IT) operations, and laws and regulations.

[0003] In related technologies, enterprise intelligent service assistants require the pre-setting of numerous rules to match possible user questions in practical applications, which constitutes a huge workload and high human resource costs in the initial stage. Moreover, enterprise intelligent service assistants often cannot effectively identify and handle non-standard expressions or long-tail questions from users, relying on preset rules and frequently asked questions (FAQs), resulting in limited coverage and an inability to adapt to diverse user needs, leading to technical problems such as poor task processing performance.

[0004] There is currently no effective solution to the aforementioned technical problem of poor task processing performance. Summary of the Invention

[0005] This application provides a task processing method, apparatus, electronic device, and readable storage medium to at least solve the technical problem of poor task processing performance.

[0006] According to one aspect of the embodiments of this application, a task processing method is provided, the method comprising: obtaining a task query request from a client, wherein the task query request is used to characterize a task to be queried by the client; identifying the task category to which the task to be queried belongs based on the task query request; invoking an agent corresponding to the task category to trigger a multi-turn dialogue mode and obtaining query information of the task to be queried, wherein the query information is used to characterize the query requirements of the task to be queried; using the agent to query the query information to obtain a task query result; and sending the task query result to the client.

[0007] Optionally, based on the task query request, identifying the task category to which the queried task belongs includes: converting the task query request into a vector representation to obtain a task query vector corresponding to the task query request, wherein the task query vector is used to characterize the semantic features corresponding to the task query request; comparing the similarity between the task query vector and at least one task vector in a preset vector database to obtain a comparison result, wherein the task vector corresponds one-to-one with a preset task category; and identifying the task category to which the queried task belongs based on the comparison result.

[0008] Optionally, based on the comparison results, the task category to which the query task belongs is identified, including: in response to the comparison results indicating that the similarity between the task query vector and the target task vector is greater than a similarity threshold, the target task category corresponding to the target task vector is identified as the task category to which the query task belongs.

[0009] Optionally, the agent corresponding to the task category is invoked to trigger a multi-turn dialogue mode and obtain query information, including: invoking the agent corresponding to the task category from the agent database based on the identification information corresponding to the task category; inputting the task query request into the agent for processing to obtain the query keywords corresponding to the task query request; and triggering a multi-turn dialogue mode based on the query keywords to obtain query information.

[0010] Optionally, an intelligent agent is used to process the query information to obtain task query results, including: using the intelligent agent to query the query information in the task database to obtain query results, wherein the task database is the database associated with the intelligent agent; and combining the query results to obtain task query results.

[0011] Optionally, the method further includes: updating the task database based on the updated task rules in response to an update of the task rules corresponding to the task category.

[0012] Optionally, the task query request can be in voice or text format.

[0013] According to another aspect of the embodiments of this application, a task processing apparatus is also provided, the apparatus comprising: an acquisition unit, configured to acquire a task query request from a client, wherein the task query request is used to characterize a task to be queried by the client; an identification unit, configured to identify the task category to which the task to be queried belongs based on the task query request; a triggering unit, configured to invoke an agent corresponding to the task category to trigger a multi-turn dialogue mode and acquire query information of the task to be queried, wherein the query information is used to characterize the query requirement of the task to be queried; a querying unit, configured to use the agent to query the query information to obtain a task query result; and a sending unit, configured to send the task query result to the client.

[0014] According to another aspect of the embodiments of this application, an electronic device is also provided, including: a memory storing an executable program; and a processor for running the program, wherein the program executes the task processing methods of various embodiments of this application when it runs.

[0015] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, which includes a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to execute the task processing methods of various embodiments of this application.

[0016] According to another aspect of the embodiments of this application, a computer program product is also provided, including a computer program that, when executed by a processor, implements the task processing methods of various embodiments of this application.

[0017] According to another aspect of the embodiments of this application, a computer program product is also provided, including a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, implements the task processing methods of various embodiments of this application.

[0018] According to another aspect of the embodiments of this application, a computer program is also provided, which, when executed by a processor, implements the task processing methods of various embodiments of this application.

[0019] In this embodiment, a task query request from the client is obtained, whereby the task query request represents the client's query task. Based on the task query request, the task category to which the query task belongs is identified. The agent corresponding to the task category is invoked to trigger a multi-turn dialogue mode, obtaining query information for the query task, whereby the query information represents the query requirement of the query task. The agent is used to query the query information to obtain the task query result. The task query result is then sent to the client. In other words, in this embodiment, based on the client's task query request, the task category to which the query task belongs is identified, and then the agent corresponding to the task category is invoked to trigger a multi-turn dialogue mode to supplement and improve the query information. Even with incomplete initial information, the query task can be successfully completed, greatly optimizing the user experience, improving task processing efficiency, and thus solving the technical problem of poor task processing efficiency. Attached Figure Description

[0020] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0021] Figure 1This is a flowchart of a task processing method according to an embodiment of this application;

[0022] Figure 2 This is a schematic diagram of a task processing flow according to an embodiment of this application;

[0023] Figure 3 This is a schematic diagram of an intent recognition framework according to an embodiment of this application;

[0024] Figure 4 This is a schematic diagram of a business processing flow according to an embodiment of this application;

[0025] Figure 5 This is a flowchart of a method for automating business travel processing according to an embodiment of this application;

[0026] Figure 6 This is a flowchart of an intelligent questioning and decision-making method according to an embodiment of this application;

[0027] Figure 7 This is a schematic diagram of a task processing device according to an embodiment of this application. Detailed Implementation

[0028] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0029] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0030] According to an embodiment of this application, an embodiment of a task processing method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0031] Figure 1 This is a flowchart of a task processing method according to an embodiment of this application, which can be applied to enterprise intelligent service assistants (e.g., enterprise digital humans). Figure 1 As shown, the method includes the following steps:

[0032] Step S101: Obtain the client's task query request.

[0033] In the technical solution provided in step S101 of this application, the client can be a personal device or platform used by the user, such as a personal computer (PC), enterprise applications on mobile devices, enterprise-level instant messaging tools, etc. Users can use these clients to issue task query requests in the form of text, voice, or even images. This task query request is used to represent the task to be queried by the client, such as rule consultation, business processing status query, data analysis request, etc.

[0034] In this embodiment, after a user submits a task query request through the client, the enterprise intelligent assistant can obtain the task query request submitted by the client. This task query request supports multimodal input; for example, the task query request can be in multiple forms such as text, voice, and images.

[0035] Optionally, after obtaining a task query request, the enterprise intelligent assistant can use the natural language processing capabilities of the large model to perform in-depth analysis of the task query request, identify user intent, and accurately capture user query needs through contextual understanding and semantic expansion, even if the task query request is not precise enough.

[0036] Step S102: Based on the task query request, identify the task category to which the task to be queried belongs.

[0037] In the technical solution provided in step S102 of this application, semantic analysis of user-input text or speech can be performed using natural language processing (NLP) and deep learning models to identify the underlying intent and specific task category behind the task query request.

[0038] In this embodiment, upon receiving a task query request, the enterprise intelligent assistant first converts the request into parsable structured data. Then, using a fine-tuned large language model (Qwen-plus), semantic understanding is performed on this structured data. Through context awareness and knowledge retrieval enhancement (RAG), key information and potential needs within the task query request can be captured. This fine-tuned large language model is trained to understand the specific context within the enterprise.

[0039] Alternatively, the large language model (Qwen-plus) can identify the query intent of a task query request, such as consultation, inquiry, complaint, or handling specific business, based on Prompt engineering and Chain of Thought (COT) technology.

[0040] Optionally, NLP techniques, such as named entity recognition, relation extraction, and text classification, can be used to further refine the identified intent and determine the specific task category, such as querying travel reimbursement rules, seeking IT support, or consulting on tax compliance issues.

[0041] In this step, identifying the task category of the queried task based on the task query request significantly improves the automation and accuracy of task processing, substantially reduces the need for manual intervention, and shortens task response and processing time. More importantly, it lays a solid foundation for subsequent multi-round dialogue management and application programming interface (API) calls, ensuring smooth and efficient business processes. Furthermore, by combining user profiles and historical behavioral data, it can provide a more personalized and humanized service experience, further enhancing user satisfaction and enterprise operational efficiency.

[0042] Step S103: Invoke the agent corresponding to the task category, trigger a multi-turn dialogue mode, and obtain the query information of the task to be queried.

[0043] In the technical solution provided in step S103 of this application, the intelligent agent corresponding to the task category identified in step S102 is invoked. Each intelligent agent is designed for a specific task category (e.g., travel, expense reimbursement, consultation, etc.) and possesses the professional knowledge and processing capabilities for that category. This process ensures a targeted response to user queries, providing the most professional and effective service in the field.

[0044] In this embodiment, if the task request is identified as being related to the "travel booking" task category, an agent specific to the travel booking business category can be invoked. This agent is familiar with all relevant rules, processes, and user needs.

[0045] Optionally, to gain a more comprehensive understanding of the user's query needs, a multi-turn dialogue mode can be automatically initiated. This mode allows the agent to interact continuously with the client, gradually collecting all the information needed to complete the query task. In the multi-turn dialogue, the agent dynamically generates new questions based on existing information and context, guiding the user to provide more details until all the key data required to perform the task or query is collected.

[0046] Optionally, through multi-turn dialogue, the intelligent agent can gradually construct a complete picture of the task to be queried. The collected query information may include, but is not limited to, date, location, budget, and personal preferences, all of which are key parameters required to perform specific business operations or retrieval. The acquisition of this information is based on user input and contextual understanding, demonstrating the advanced nature of enterprise intelligent service assistants in natural language processing and dialogue management.

[0047] For example, during the travel booking process, the system can ask for information such as departure date, destination, and preferred mode of transportation (airplane, high-speed rail, etc.) to ensure that the query or booking operation can be carried out according to the user's specific needs.

[0048] Optionally, the collected query information will be used for the next step of task execution, transformed into API call parameters for internal enterprise systems, or used as search criteria for knowledge base queries. This means that, based on understanding user intent, it can seamlessly integrate with internal enterprise systems to automatically execute tasks or provide accurate query results, thereby greatly improving business processing efficiency and accuracy.

[0049] For example, based on the business trip dates and locations provided by the user, the intelligent agent can call the enterprise's travel management system API to query eligible flight and hotel options and automatically complete the booking process.

[0050] In this step, by calling an agent that matches the task category, a multi-turn dialogue mode is initiated to meticulously collect and refine the user's query information, laying a solid foundation for subsequent business execution or knowledge retrieval. This not only improves the user experience but also significantly optimizes the enterprise's internal business processes, achieving efficient transformation from user needs to business responses.

[0051] Step S104: Use the intelligent agent to query the query information and obtain the task query results.

[0052] In the technical solution provided in step S104 of this application, after obtaining the query information, an intelligent agent can be used to query the query information to obtain the task query result.

[0053] In this embodiment, after obtaining the query information, the intelligent agent securely accesses and manipulates business data by calling the API of the enterprise's internal system, thereby obtaining the task query results. This approach not only ensures the real-time nature and accuracy of the data but also avoids the potential risks associated with directly manipulating the database.

[0054] Optionally, after obtaining the query information, the task type of the task to be queried can also be identified, where the task type includes at least simple tasks and complex tasks. Different query strategies are adopted for different types of business.

[0055] Optionally, simple business operations typically refer to those with straightforward processing logic, requiring no additional data uploads, and without complex approval processes, such as routine travel bookings and expense reimbursement applications. When processing simple business operations, the system can automatically generate corresponding API call instructions based on the query information, and then directly call the system's API to execute the business operation, such as querying and booking high-speed rail tickets.

[0056] Optionally, complex business processes often involve multiple steps, require uploading attachments, and involve approval processes. When handling complex business processes, multi-turn dialogues can be used to guide users to upload necessary documents (e.g., scanned copies of expense reports), while maintaining the continuity of the dialogue context. The enterprise intelligent service assistant, combined with the COT (Consciousness-Oriented Thinking) technology of the Prompt project, decomposes complex business processes and automatically generates a series of API commands for use in calling multiple stages of the business system. If the business process includes an approval stage, the enterprise intelligent service assistant will automatically trigger the enterprise's internal approval mechanism, submitting the business request to the relevant responsible person, while maintaining real-time tracking of the approval status. After obtaining all necessary information and obtaining approval, the enterprise intelligent service assistant will execute the business process, such as completing a high-value expense reimbursement application, and provide feedback on the final result to the user and relevant managers, while updating the business status in the enterprise data platform.

[0057] Optionally, after the intelligent agent completes its internal data query, it will integrate all relevant information to generate intuitive and easy-to-understand task query results. These results can directly answer the user's questions, provide instructions for the user's next steps, or be business processing results that require user confirmation.

[0058] For example, when querying travel expenses, the agent not only reports the total cost but also lists the details, including transportation, accommodation, and meals, and may include cost-saving suggestions.

[0059] In this step, the use of an intelligent agent to drive the query significantly improves query efficiency and saves query time. Furthermore, automated querying eliminates errors caused by human interpretation of requirements and data manipulation, thus improving the accuracy of query results.

[0060] Step S105: Send the task query results to the client.

[0061] In the technical solution provided in step S105 of this application, after obtaining the query result, the query result can be returned to the client to convey the task query result to the user in a user-friendly way, thereby improving user experience and satisfaction.

[0062] For example, if a user initiates a travel request through the client to book a trip to a certain destination on a specific day, after parsing, intent recognition, multi-round dialogue guidance, and business execution, the system ultimately completes the travel booking and provides detailed business query results. At this point, these query results, such as booking details, train / bus information, and e-tickets, can be fed back to the user through the client. This demonstrates the completeness of the end-to-end service, ensuring that users can not only easily initiate requests but also receive results promptly and accurately, thus completing a closed-loop business process.

[0063] Through steps S101 to S105 above, based on the client's task query request, the task category to which the task to be queried belongs is identified, and then the intelligent agent corresponding to the task category is invoked to trigger a multi-turn dialogue mode to supplement and improve the query information. Even in the case of incomplete initial information, the query task can be completed smoothly, which greatly optimizes the user experience, improves the processing efficiency of the task, and thus solves the technical problem of poor task processing effect.

[0064] The method described in this embodiment will be further described below.

[0065] As an optional implementation, step S102, based on the task query request, identifies the task category to which the queried task belongs, including: converting the task query request into a vector representation to obtain the task query vector corresponding to the task query request, wherein the task query vector is used to characterize the semantic features corresponding to the task query request; comparing the similarity between the task query vector and at least one task vector in a preset vector database to obtain a comparison result, wherein the task vector corresponds one-to-one with a preset task category; and identifying the task category to which the queried task belongs based on the comparison result.

[0066] In this embodiment, after obtaining the task query request, large model technology (such as Qwen-plus) can be used to perform in-depth analysis on the language structure and semantic content of the task query request, extract key semantic features, and then map them into a multi-dimensional vector space to form a task query vector.

[0067] Optionally, the generation of the task query vector not only includes the textual content of the user's request, but also implies rich semantic information such as the emotional tone, context and potential intent of the request, which is crucial for subsequent classification and recognition.

[0068] Optionally, after obtaining the task query vector, the generated task query vector can be compared with the task vectors in a pre-built vector database for similarity. The pre-built vector database is a pre-constructed set of vectors, where each task vector represents a specific task category defined internally by the enterprise, such as travel booking, expense reimbursement review, or consultation. These task vectors are endowed with corresponding semantic features from the outset, and through large-scale training and optimization, it is ensured that the task vectors accurately represent the typical problems and needs within their respective task categories. Text vectorization techniques, such as word embeddings, sentence vectors, or paragraph vectors, can be used to convert the predefined task category information into vector representations, facilitating subsequent calculations and comparisons.

[0069] Optionally, cosine similarity, Euclidean distance, or other suitable similarity calculation methods can be used to measure the similarity between the task query vector and the task vectors in the database. This step determines the final matching accuracy of the task category.

[0070] After comparing the similarity between the task query vector and the task vector, the task category that best matches the task query vector can be identified based on the comparison results. This identification process, based on the ranking of the comparison results and the setting of thresholds, ensures that only the semantically most matching task category is selected, thereby avoiding false matches and expanding the processing scope.

[0071] In this step, compared to traditional rule-based methods, vector representation and similarity comparison effectively address the long-tail problem and non-standardized expressions of users, significantly improving the accuracy of intent recognition. Furthermore, the task query vector generation process takes into account the context of the request, enabling a better understanding of the user's continuous intent in multi-turn dialogues and enhancing the system's dialogue management capabilities.

[0072] As an optional implementation, based on the comparison results, identifying the task category to which the query task belongs includes: in response to the comparison results indicating that the similarity between the task query vector and the target task vector is greater than a similarity threshold, identifying the target task category corresponding to the target task vector as the task category to which the query task belongs.

[0073] In this embodiment, after obtaining the comparison results, a similarity threshold is used as the judgment criterion to ensure the accuracy and timeliness of task category identification. When the comparison result indicates that the similarity between the task query vector and a target task vector exceeds a preset similarity threshold, it is considered that the query task matches the target task category corresponding to the target task vector, thereby identifying the target task category as the task category to which the query task belongs.

[0074] Optionally, the similarity threshold setting is derived from a large amount of training data and experimental adjustments, balancing recognition accuracy and recall, ensuring that true matches are not missed due to an excessively high threshold, nor are a large number of false positives generated due to an excessively low threshold.

[0075] Optionally, the similarity threshold is not fixed. As the intelligent agent application iterates and the enterprise knowledge base is updated, the similarity threshold can be periodically reassessed and adjusted to adapt to new business scenarios and user needs. This dynamic adjustment mechanism ensures the continuous accuracy and robustness of the recognition process.

[0076] In this step, by converting user queries into vectors and comparing them with target task vectors, the category of the task query request can be accurately identified in a short time, thereby initiating subsequent multi-round dialogues and business processing flows. This process is both efficient and accurate, significantly improving employee work efficiency and experience.

[0077] As an optional implementation, step S103 involves calling the agent corresponding to the task category, triggering a multi-turn dialogue mode, and obtaining query information. This includes: calling the agent corresponding to the task category from the agent database based on the identification information corresponding to the task category; inputting the task query request into the agent for processing to obtain the query keywords corresponding to the task query request; and triggering a multi-turn dialogue mode based on the query keywords to obtain query information.

[0078] In this embodiment, the identification information corresponding to the task category is used to uniquely identify the task category. For example, the identification information can be an identification code corresponding to the task category, without specific limitations. Each agent in the agent database corresponds one-to-one with a task category, and each agent also has its own identification information. The identification information corresponding to an agent is consistent with the identification information of the task category corresponding to the agent. Based on this, after identifying the task category corresponding to the task to be queried, the agent corresponding to that task category can be retrieved from the agent database based on the identification information corresponding to that task category.

[0079] Optionally, after retrieving the agent corresponding to the task category from the agent database, the agent can use natural language processing techniques, such as named entity recognition and semantic parsing, to extract key information from the task query request and form query keywords. These keywords may include date, location, amount, personnel name, or other business-related details.

[0080] Optionally, after obtaining the query keywords, a multi-turn dialogue mode can be triggered. The purpose of triggering this multi-turn dialogue mode is to further refine the query requirements and obtain all the necessary information required to execute the business task. Through a dynamic dialogue process, the agent can generate subsequent dialogue clues based on existing information and business rules, guiding the user to provide more information.

[0081] Optionally, during multi-turn dialogues, the agent automatically manages the context and tracks the dialogue history to ensure that information from each turn is properly processed and utilized. For example, if a user mentions their preferred mode of transportation in a travel booking conversation, the agent will avoid repeating the question in subsequent inquiries and instead provide more detailed information about that mode of transportation, such as departure time and seat preference.

[0082] Optionally, utilizing a multi-turn dialogue model ensures that the agent collects all necessary query information before executing the task, avoiding business processing failures or delays due to incomplete information. This model can also handle possible changes or supplementary requests from users midway through the process, making the entire dialogue flow more aligned with the user's actual needs and improving service flexibility and satisfaction.

[0083] In this step, by invoking the intelligent agent and triggering a multi-turn dialogue mode, the transformation from a fuzzy task query request to complete and accurate query information is achieved. This process fully leverages the natural language understanding and generation capabilities of the large model, combined with the enterprise's internal business knowledge and processes, laying a solid foundation for business queries and result generation.

[0084] As an optional implementation, an intelligent agent is used to process the query information to obtain task query results, including: using the intelligent agent to query the query information in a task database to obtain query results, wherein the task database is the database associated with the intelligent agent; and combining the query results to obtain task query results.

[0085] In this embodiment, after obtaining the query information, an intelligent agent can be used to process the query information, generate a query statement, and directly call the task database query service in the business system through the stable and low-latency interface provided by the cloud-native platform to perform the query.

[0086] Optionally, after obtaining the query results, the query results can be combined to generate task query results.

[0087] This step provides richer and more in-depth services than traditional search-based question-and-answer databases and rule engine systems, which not only improves the level of automation in enterprise operations but also optimizes the work experience of employees, injecting new vitality into the digital transformation of enterprises.

[0088] As an optional implementation, the task processing method further includes: updating the task database based on the updated task rules in response to an update of the task rules corresponding to the task category.

[0089] In this embodiment, the enterprise intelligent service assistant, through integration with the enterprise's internal change management system, can proactively monitor any changes to task rules. These changes may stem from rule revisions, process optimizations, or system upgrades. By subscribing to change events in the internal system, the assistant can obtain updates immediately.

[0090] Optionally, an automatic synchronization process can be initiated after an update to the task rules is detected. This process typically includes parsing, standardizing, and vectorizing the updated rules, as well as comparing and integrating them with the existing task database. Through RAG retrieval enhancement technology, new rules can be quickly integrated into the existing task database, ensuring timely updates and consistency.

[0091] Optionally, the system design includes an automated processing module for receiving, parsing, and executing update tasks. This module can automatically modify relevant entries in the task database based on the nature of the update (e.g., addition, modification, deletion), avoiding delays and errors that may occur with manual updates. Furthermore, the system will push update notifications to relevant agents, ensuring that agents can make decisions and process data based on the latest task database.

[0092] This update process demonstrates the high real-time performance and flexibility of the enterprise intelligent service assistant system. Unlike traditional systems that require manual intervention and involve lengthy update cycles, the enterprise intelligent service assistant achieves near-instantaneous rule synchronization, significantly reducing problems caused by rule lag. Simultaneously, this mechanism reduces the workload of operations and maintenance personnel, lowers maintenance costs, and improves the overall operational efficiency of the system.

[0093] For example, suppose a company's HR department updates an employee benefits rule, such as increasing employees' annual leave days from 10 to 15 days. This change is detected by subscribing to HR's update events. The newly added rule text is parsed to extract key information, such as the change in annual leave days. The parsed rule change is automatically synchronized to the task database, updating relevant fields such as "annual leave days." The HR agent will fine-tune itself based on this update, for example, by retraining its intent recognition model to ensure it can correctly handle query requests involving adjustments to annual leave days. After the update, the system conducts internal testing to verify that the agent's processing logic correctly reflects the new rule. Ultimately, users will receive accurate answers based on the latest regulations when interacting with the company's intelligent service assistant, without worrying about rule lag or untimely updates.

[0094] The technical solutions of the embodiments of this application will be illustrated below with reference to preferred embodiments.

[0095] Currently, enterprise-level intelligent service assistants have brought unprecedented opportunities to intelligent office systems. Large language models, due to their massive parameter scale and extensive training data, are capable of understanding complex natural language instructions, generating high-quality text, and integrating knowledge from multiple domains, providing enterprises with more intelligent and efficient solutions. Against this backdrop, enterprises are beginning to explore how to integrate large language models into daily office processes to achieve a series of intelligent services such as automated task processing, intelligent data analysis, and knowledge-based question answering.

[0096] While enterprise intelligent service assistants (such as rule engine systems, retrieval-based question-answering libraries, and traditional machine learning models) have achieved some success in related technologies, significant technical challenges remain when handling complex and diverse business scenarios. For example, existing enterprise intelligent service assistants rely on a large number of pre-labeled rules and question-answer pairs, which means a significant amount of manual input when launching new businesses or deploying models for the first time, extending system preparation time and increasing initial costs. Traditional technical solutions are often based on fixed rules and templates, making it difficult to adapt to various non-standard user expressions, resulting in limited coverage of user intent understanding, especially when facing novel or long-tail questions. These technical problems limit the effectiveness of enterprise intelligent service assistants and hinder their widespread application in enterprise environments.

[0097] However, this application provides a method for automatically handling enterprise-level intelligent office tasks based on a large language model. Using a large language model as a foundation, combined with RAG and intelligent agent technologies, it constructs knowledge bases for various business domains within the enterprise, providing all employees with exclusive knowledge Q&A within each business domain; it restructures enterprise business processes to achieve intelligent business processing; it provides intelligent question counting and intelligent data analysis, utilizing natural language to achieve data insights and assist enterprise management in decision-making; and it provides knowledge retrieval and generation services for all employees, improving office efficiency and effectively addressing the shortcomings of traditional artificial intelligence technologies in handling enterprise-level intelligent office tasks, such as strong rule dependence, weak generalization ability, and high maintenance costs.

[0098] In this embodiment, the enterprise-level intelligent customer service and business automation platform aims to balance resource efficiency, data security, and business agility. Through the Bailian platform, the enterprise intelligent service assistant and business automation platform can perform large-scale language model training and inference (e.g., the Qwen series) in the cloud. During the large model training phase, it can rapidly allocate significant computing power, shortening the training cycle; during the inference phase, it allocates resources on demand, effectively handling peak traffic.

[0099] Optionally, by leveraging elastic computing resources and services, enterprise-grade intelligent service assistants and business automation platforms can adjust computing power supply according to real-time business needs. Whether facing the challenge of a surge in user inquiries in the short term or the agile deployment of new features, they can respond in a timely manner without worrying about resource bottlenecks, laying a solid foundation for continuous business growth.

[0100] Optionally, for functions that process core enterprise data or involve sensitive information, the automation platform has chosen to deploy cloud-native technologies in a private cloud environment. This allows for fast local data access, avoiding network latency, while strictly adhering to the enterprise's internal data security compliance requirements. Through isolation and access control mechanisms, it protects enterprise data from leakage and enhances user trust.

[0101] Optionally, by employing an advanced intent recognition model, the enterprise intelligent service assistant and business automation platform can accurately analyze the user's multi-dimensional needs and route them to the corresponding professional intelligent agents. These intelligent agents perform in-depth knowledge retrieval and logical reasoning for different business areas (such as human resources, finance, and regulations), forming decoupled business processing units that ensure the vertical domain specialization of knowledge and prevent interference between them.

[0102] Optionally, the construction and iteration of intelligent agents in each business domain can be carried out independently. This can make full use of the business knowledge of domain experts and respond quickly to market changes, such as rule adjustments and business process optimizations, without affecting the stability of other domains or the entire system.

[0103] Optionally, the enterprise intelligent service assistant and business automation platform, through open API interfaces and a rich portfolio of AI products, can seamlessly integrate the latest AI models and technologies, such as RAG retrieval enhancement, the establishment of vector knowledge bases, and large models fine-tuned for specific scenarios, bringing enterprises the most advanced and effective intelligent solutions.

[0104] Optionally, by integrating technologies such as Prompt engineering, traditional small models, RAG and vector knowledge bases, enterprise intelligent service assistants can automate the understanding and processing of complex business processes, significantly reducing manual intervention, lowering operating costs, improving business efficiency, and accelerating the pace of enterprise digital transformation.

[0105] Optionally, the hybrid cloud architecture adopted by the enterprise-level intelligent customer service and business automation platform not only meets the enterprise's demand for efficient use of computing resources, but also takes into account the dual needs of data security and business agility. By building a multi-agent collaborative framework, it realizes a high degree of personalization and professionalism of intelligent services, bringing significant business value and management efficiency improvement to enterprises.

[0106] The enterprise-level intelligent customer service and business automation platform in this application embodiment will be further described below.

[0107] The enterprise-level intelligent customer service and business automation platform comprises four parts: interactive terminals, intelligent agent applications, enterprise applications, and business systems. The interactive terminals handle user interaction and support both voice and text. The web terminals provide system management and operational capabilities, including permission management configuration, intent recognition management configuration, system operation data analysis, and log analysis. Intelligent agent applications include those created using the Assistant API and those directly created on the Bailian platform, categorized by function as intent recognition agents and agents for various business domains. Enterprise applications, designed with a microservice architecture, are deployed within the system and include domain access permission management, multi-turn dialogue management, agent routing, plugin service routing, middleware data query, and log management. The business systems provide business processing interfaces, supporting automated business processing through the enterprise intelligent service assistant.

[0108] Optionally, the main functions provided by the enterprise intelligent service assistant and business automation platform based on AI big data model technology in this application embodiment can be divided into the following categories: knowledge Q&A, automatic business processing, indicator analysis, document generation, and visual processing.

[0109] For knowledge-based question-and-answer questions, a comprehensive knowledge vector library has been pre-built. This library uses techniques such as document parsing, knowledge vectorization, and data slicing to transform the professional knowledge of various enterprise fields into machine-understandable vector forms, achieving efficient knowledge storage and retrieval. Combined with RAG retrieval enhancement technology, the enterprise intelligent service assistant can quickly locate and provide the most relevant information from massive amounts of knowledge based on the specific needs of users' daily work, significantly improving the user experience of rule consultation and knowledge retrieval, and increasing efficiency.

[0110] For automated business processing, leveraging the COT (Content-Oriented Thinking Chain) technology of the Prompt project, the enterprise intelligent service assistant, through deep understanding of business processes and users' natural language expressions, can automatically analyze and extract the parameters and conditions required for business processing. This, in turn, automatically generates corresponding business system API call information, achieving automated business processing. This innovation breaks away from the traditional form-filling and rule-driven model, significantly lowering the threshold for business processing, shortening the processing cycle, and achieving seamless integration and automated execution of business processes.

[0111] For indicator analysis, the Enterprise Intelligent Service Assistant possesses powerful NL2SQL capabilities, capable of translating natural language into precise SQL queries and directly retrieving high-quality data from the enterprise data platform for analysis. Based on a pre-trained model, through specific fine-tuning and training, the Enterprise Intelligent Service Assistant can automatically generate descriptive analysis results, providing management with instant data insights. This reduces the traditional report development cycle from days to seconds, truly realizing a highly efficient data query and analysis model of "question-answer-instant insight."

[0112] For document generation, the enterprise intelligent service assistant can leverage the summarizing capabilities of large models to generate data insight analysis reports based on data analysis characteristics and business-related knowledge, facilitating management's business decision-making analysis.

[0113] For visual processing, the Qwen-VL large model of the enterprise intelligent service assistant integrates advanced visual processing technology, enabling deep analysis of images and extraction of key information. This function is suitable for scenarios such as document review and data entry. Through automated visual analysis, it significantly reduces the time and error rate of manual inspection, improving the level of office automation and business efficiency.

[0114] Figure 2 This is a schematic diagram of a task processing flow according to an embodiment of this application, such as... Figure 2 As shown, the dialogue interface is the front-end endpoint for interaction between the enterprise intelligent service assistant and the user. It can be provided to users through micro-applications, web terminals, etc., and supports multiple interaction methods such as voice and text. Users can ask questions or make requests on this interface, whether it is for rule consultation, business processing, or data analysis needs. The enterprise intelligent service assistant can receive and parse the user's natural language input at this step, laying the foundation for subsequent processes.

[0115] The intent recognition agent is a crucial step in the process. Utilizing Prompt engineering and the Qwen-plus model, combined with vector threshold matching technology, it performs deep understanding of the user's natural language input to accurately determine the user's true intent and needs. This agent can handle complex and varied user expressions, quickly routing the user's intent to the appropriate business domain, ensuring the relevance and efficiency of subsequent processing steps.

[0116] Domain agents are independent intelligent processing units designed for specific business domains. Each agent possesses its own knowledge base and processing logic, enabling it to directly handle domain-related business upon receiving a request from an intent-recognition agent. The design of domain agents not only decouples business domains but also allows each agent to be independently optimized and upgraded according to its own needs, enhancing the flexibility and scalability of the entire system.

[0117] The domain knowledge base contains detailed information and rules for various business domains within an enterprise. Built using technologies such as document parsing and knowledge vectorization, it serves as the basis for domain-specific intelligent agents to make decisions and perform processing. The enterprise intelligent service assistant, through RAG-enhanced retrieval, can quickly and accurately retrieve the most relevant information to user requests from the knowledge base, achieving efficient utilization of knowledge.

[0118] The business system serves as the backend for communication between the enterprise's intelligent service assistant and internal business processes. It provides various business processing interfaces, ranging from simple query services to complex approval workflows. Domain-specific intelligent agents automate business processes by calling these interfaces, eliminating the need for manual intervention. This step ensures the security of business data and the automation of business processes, significantly improving employee efficiency and business processing speed.

[0119] For example, at the start of a conversation, the user asks a question or make a request in natural language through the dialogue interface. The enterprise intelligent service assistant's intent recognition agent receives the user input and quickly parses the user's intent using the Prompt project and the Qwen-plus model to determine the processing domain. The domain-routing intent is sent to the corresponding domain agent, which is responsible for handling the request for that specific business domain. Based on the user's needs, the domain agent retrieves relevant knowledge from the domain knowledge base or automatically executes business processes, such as querying data, booking services, and generating reports, by calling interfaces of the business system. The processing results are fed back to the user through the domain agent, which may include direct answers, business processing results, or data insight reports. The entire process aims to provide a fast, accurate, and personalized service experience.

[0120] Figure 3 This is a schematic diagram of an intent recognition framework according to an embodiment of this application, such as... Figure 3 As shown, the intent recognition model Prompt consists of several parts: roles, tasks, outputs, domains, constraints, examples, and thought chains. Roles define the model's identity; tasks clarify what the model needs to accomplish; outputs describe the content and format of the model's output; domains introduce the business domain, with each domain containing a domain name, domain code, and domain description; constraints specify the limitations on the model's understanding and output, such as "only the domain code of the most matching business domain can be returned; other content cannot be returned; if there is no matching business domain, then 'XXXX' content will be returned"; examples provide input and output examples to ensure the model accurately executes the task; and thought chains provide hints and explanations when the model's responses are poor in certain domains.

[0121] Figure 4 This is a schematic diagram of a business processing flow according to an embodiment of this application, such as... Figure 4As shown, different processing methods are adopted for simple business (no complex information to be submitted, no attachments to be uploaded, no approval workflow, etc.) and complex business (requiring the submission of more information, the uploading of attachments, and business approval, etc.), taking into account factors such as the number of interaction rounds, processing time, and accuracy of successful processing.

[0122] Optionally, simple business operations typically refer to those with straightforward processing logic, requiring no additional data uploads, and without complex approval processes, such as routine travel bookings and expense reimbursement applications. The processing flow for these types of business operations typically includes: receiving user queries, identifying the intent behind the user query, generating inquiry information, questioning the user, receiving supplementary information from the user, and displaying the business processing result to the user.

[0123] Optionally, when handling simple business transactions, users can submit business requests through the interactive terminal of the enterprise intelligent service assistant, describing their needs using natural language. The enterprise intelligent service assistant then leverages its powerful natural language understanding capabilities, using the Prompt framework and the Qwen-plus model to quickly identify user intent and determine the business domain (e.g., travel booking). Utilizing Natural Language Understanding (NLU) and semantic understanding technologies, key business information is extracted from the user input, and then corresponding business system API call instructions are automatically generated based on this information. The enterprise intelligent service assistant directly calls the business system's API to execute the business, such as querying and booking high-speed rail tickets. Afterwards, the business execution results are fed back to the user in real time, such as displaying successfully booked train information, while ensuring synchronized data updates.

[0124] Optionally, complex business processes typically involve multiple steps, require uploading attachments, and involve approval processes. The processing flow for these types of business processes usually includes: receiving a user query, identifying the user's intent in the query, returning an embedded page, having the user supplement information and submit it, and then performing a query based on the supplemented information.

[0125] Optionally, the processing flow for this type of business typically includes the following steps: Initial interaction and information collection: Similar to handling simple business, the enterprise intelligent service assistant first receives user input, performs initial intent recognition and information extraction; Multi-turn dialogue and data upload: For complex businesses that require uploading additional data, the enterprise intelligent service assistant will conduct multi-turn dialogues to guide the user to upload the required files (e.g., scanned copies of expense reports), while maintaining the continuity of the dialogue context; Business deconstruction and API command generation: The enterprise intelligent service assistant, combined with the COT thinking chain technology of the Prompt project, decomposes complex business processes and automatically generates a series of API commands for use in multiple stages of the business system. Invocation; Approval flow integration and automatic triggering: If the business process includes an approval step, the enterprise intelligent service assistant will automatically trigger the enterprise's internal approval mechanism, submitting the business request to the relevant person in charge, while maintaining real-time tracking of the approval status; Business execution and result feedback: After obtaining all necessary information and approval, the enterprise intelligent service assistant will execute the business process, such as completing a high-amount reimbursement application, and provide feedback on the final result to the user and relevant managers, while updating the business status in the enterprise data platform; Closed-loop verification and log recording: The enterprise intelligent service assistant will verify the business execution result to ensure that all steps are executed correctly, and record the interaction log of the entire business process for subsequent tracking and auditing.

[0126] Optionally, different processing methods can be adopted for different business operations, which can improve the efficiency of task processing and user interaction experience.

[0127] Figure 5 This is a flowchart of a method for automating business travel processing according to an embodiment of this application, such as... Figure 5 As shown, the method includes the following steps.

[0128] Step S501, Intent Identification and Routing.

[0129] In this embodiment, users can make travel booking requests through the client using natural language (e.g., voice input), such as: "Please book me a trip to Shanghai next Monday, high-speed rail preferred."

[0130] Optionally, the enterprise intelligent service assistant has a pre-defined set of prompt templates to guide the model in understanding and executing specific tasks. In this scenario, the prompt is designed to identify travel booking needs, including role definition (travel business domain analysis assistant), task description (identifying travel bookings), output requirements (domain code), and domain details (such as air tickets, hotels, high-speed rail services), and specifies the uniqueness and accuracy of the model's output.

[0131] Optionally, by combining the Qwen-plus large model and vector threshold matching technology, the enterprise intelligent service assistant can quickly and accurately identify that the above user request belongs to the travel booking domain and output the corresponding domain code. This code is used as the routing basis to send the processing request to the intelligent agent specifically responsible for travel business.

[0132] Step S502: Multi-round dialogue and data extraction.

[0133] In this embodiment, after receiving the domain routing information, the travel agent begins a multi-turn dialogue with the user, using RAG technology to retrieve the company's travel regulations to ensure that the booking complies with the regulations (e.g., "management-level high-speed rail second-class seat standard"). Simultaneously, NLU technology is used to parse details of the user's needs, such as departure time, destination, and mode of transportation preferences.

[0134] Optionally, the intelligent system can dynamically generate further inquiries based on existing information to refine the booking details, such as "Do I need to book a return ticket?" and "How many days are you expected to stay?" These questions help gather complete information and ensure a smooth booking process.

[0135] Step S504, Business Execution and System Integration.

[0136] In this embodiment, the intelligent agent, based on the complete information provided by the user, calls an internal enterprise API, such as the "High-Speed ​​Rail Ticketing System API," to query and locate trains that meet the criteria. If the booking cost exceeds the budget or standard, the intelligent agent will also automatically trigger the "Enterprise Approval Workflow Plugin" to report the situation to the supervisor for approval.

[0137] Optionally, after a successful booking, the enterprise's intelligent service assistant will push the booking details (including train information and seat number) and electronic ticket to the user's terminal, while updating the workbench to ensure information consistency and real-time performance.

[0138] In steps S501 to S503 above, the enterprise intelligent service assistant significantly improves many pain points in the traditional business travel booking process through its intelligent and efficient processing flow, realizing a one-stop service from user intent recognition, multi-round dialogue interaction to business execution.

[0139] Figure 6 This is a flowchart of an intelligent question-and-decision method according to an embodiment of this application, such as... Figure 6 As shown, the method includes the following steps:

[0140] Step S601: Conversion from natural language to structured query language.

[0141] In this embodiment, when the enterprise intelligent service assistant receives a task query request in natural language, it can utilize built-in large model technology to instantly convert complex natural language instructions into precise SQL query statements. This conversion process greatly simplifies the complexity of data queries, enabling management to obtain the necessary information from the enterprise data platform without requiring database query skills.

[0142] Step S602: Data platform query and visualization.

[0143] In this embodiment, after the query statement is generated, the enterprise intelligent service assistant directly calls the database API of the enterprise data platform to execute the SQL query statement and extract the sales data of a certain vehicle model in a certain region for the third quarter from the sales data table. After the query results are returned, the visual processing capabilities of the Qwen-VL large model are used to transform the data into easy-to-understand line charts and heatmaps, intuitively displaying the monthly fluctuations and regional distribution of sales.

[0144] Step S603: Root cause analysis and suggestion generation.

[0145] In this embodiment, based on data visualization, the enterprise intelligent service assistant further combines historical sales data, promotional activity records, and external information such as competitor price changes, and uses its in-depth analysis capabilities to generate a report explaining the root cause of sales fluctuations.

[0146] Optionally, the report might state: "August sales declined due to a 10% price reduction by competitor A; we recommend launching a limited-time regional offer in the fourth quarter." This data-driven insight provides management with clear direction and strategic recommendations for problem-solving.

[0147] In steps S601 to S606 above, the NL2SQL conversion function of the enterprise intelligent service assistant, combined with high-speed database query capabilities, greatly improves decision-making efficiency.

[0148] According to an embodiment of this application, a task processing apparatus is also provided. It should be noted that this task processing apparatus can be used to execute the task processing method described in the embodiments.

[0149] Figure 7 This is a schematic diagram of a task processing apparatus according to an embodiment of this application. Figure 7 As shown, the task processing device 700 may include: an acquisition unit 701, an identification unit 702, a triggering unit 703, a query unit 704, and a sending unit 705.

[0150] The acquisition unit 701 is used to acquire the client's task query request, wherein the task query request is used to represent the client's query task.

[0151] The identification unit 702 is used to identify the task category to which the task to be queried belongs based on the task query request.

[0152] Triggering unit 703 is used to call the agent corresponding to the task category, trigger a multi-turn dialogue mode, and obtain query information of the task to be queried. The query information is used to represent the query requirements of the task to be queried.

[0153] The query unit 704 is used to query the query information using an intelligent agent to obtain the task query result.

[0154] Sending unit 705 is used to send the task query results to the client.

[0155] Optionally, the identification unit 702 is further configured to: convert the task query request into a vector representation to obtain the task query vector corresponding to the task query request, wherein the task query vector is used to characterize the semantic features corresponding to the task query request; compare the similarity between the task query vector and at least one task vector in a preset vector database to obtain a comparison result, wherein the task vector corresponds one-to-one with a preset task category; and identify the task category to which the task to be queried belongs based on the comparison result.

[0156] Optionally, the identification unit 702 is further configured to: in response to the comparison result indicating that the similarity between the task query vector and the target task vector is greater than a similarity threshold, identify the target task category corresponding to the target task vector as the task category to which the query task belongs.

[0157] Optionally, the triggering unit 703 is further configured to: call the intelligent agent corresponding to the task category from the intelligent agent database based on the identification information corresponding to the task category; input the task query request into the intelligent agent for processing to obtain the query keywords corresponding to the task query request; and trigger a multi-turn dialogue mode based on the query keywords to obtain query information.

[0158] Optionally, the query unit 704 is further configured to: use the intelligent agent to query the query information in the task database to obtain query results, wherein the task database is the database associated with the intelligent agent; and combine the query results to obtain task query results.

[0159] Optionally, the device 700 is also used to: update the task database based on the updated task rules in response to an update of the task rules corresponding to the task category.

[0160] Optionally, the task query request can be in voice or text format.

[0161] In the task processing device described above in this application, based on the client's task query request, the task category to which the task to be queried belongs is identified, and then the intelligent agent corresponding to the task category is invoked to trigger a multi-turn dialogue mode to supplement and improve the query information. Even in the case of incomplete initial information, the query task can be successfully completed, which greatly optimizes the user experience, improves the task processing efficiency, and thus solves the technical problem of poor task processing effect.

[0162] Embodiments of this application also provide an electronic device, including: a memory storing an executable program; and a processor for running the program, wherein the program executes the task processing methods of various embodiments of this application during runtime.

[0163] Embodiments of this application also provide a computer-readable storage medium, which includes a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to execute the task processing methods of various embodiments of this application.

[0164] Embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements the task processing methods of various embodiments of this application.

[0165] Embodiments of this application also provide a computer program product, including a non-volatile computer-readable storage medium for storing a computer program, which, when executed by a processor, implements the task processing methods of various embodiments of this application.

[0166] Embodiments of this application also provide a computer program that, when executed by a processor, implements the task processing methods described in the various embodiments of this application.

[0167] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0168] In the above embodiments of this application, 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.

[0169] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0170] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0171] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0172] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0173] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A task processing method characterized by, The method comprises: obtaining a task query request of a client, wherein the task query request is used to represent a to-be-queried task of the client; based on the task query request, identifying a task category to which the to-be-queried task belongs; calling an intelligent agent corresponding to the task category, triggering a multi-round dialogue mode, and obtaining query information of the to-be-queried task, wherein the query information is used to represent a query requirement of the to-be-queried task; querying the query information by using the intelligent agent to obtain a task query result; sending the task query result to the client.

2. The method of claim 1, wherein, Based on the task query request, identifying the task category to which the to-be-queried task belongs, comprising: convert the task query request into a vector representation to obtain a task query vector corresponding to the task query request, wherein the task query vector is used to represent the semantic features corresponding to the task query request; performing a similarity comparison between the task query vector and at least one task vector in a preset vector database to obtain a comparison result, wherein the task vector corresponds to a preset task category one by one; based on the comparison result, identifying the task category to which the to-be-queried task belongs.

3. The method of claim 2, wherein, Based on the comparison result, identifying the task category to which the to-be-queried task belongs, comprising: in response to the comparison result indicating that the similarity between the task query vector and a target task vector is greater than a similarity threshold, identifying a target task category corresponding to the target task vector as the task category to which the to-be-queried task belongs.

4. The method of claim 1, wherein, Calling the intelligent agent corresponding to the task category, triggering the multi-round dialogue mode, and obtaining the query information, comprising: based on the identification information corresponding to the task category, calling the intelligent agent corresponding to the task category from an intelligent agent database; inputting the task query request into the intelligent agent for processing to obtain a query keyword corresponding to the task query request; based on the query keyword, triggering the multi-round dialogue mode to obtain the query information.

5. The method of claim 1, wherein, Using the intelligent agent, querying the query information to obtain a task query result, comprising: using the intelligent agent, querying the query information in a task database to obtain the query result, wherein the task database is a database associated with the intelligent agent; combining the query result to obtain the task query result.

6. The method of claim 5, wherein, The method further comprises: in response to the task rules corresponding to the task category being updated, updating the task database based on the updated task rules.

7. The method according to any one of claims 1 to 6, characterized in that, The task query request is in the form of voice or text.

8. A task processing apparatus characterized by comprising: The method comprises: an obtaining unit, configured to obtain a task query request of a client, wherein the task query request is used to represent a to-be-queried task of the client; an identifying unit, configured to identify a task category to which the to-be-queried task belongs based on the task query request; a triggering unit, configured to call an intelligent agent corresponding to the task category, trigger a multi-round dialogue mode, and obtain query information of the to-be-queried task, wherein the query information is used to represent a query requirement of the to-be-queried task; The query unit is configured to query the query information by using the intelligent agent to obtain a task query result. The sending unit is configured to send the task query result to the client.

9. An electronic device, comprising: The computer program product comprises: The memory stores an executable program. The processor is configured to run the program, and the program performs the method in any one of claims 1 to 7 when running.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium comprises a stored executable program, and the executable program controls the device where the storage medium is located to perform the method in any one of claims 1 to 7 when running.

11. A computer program product, characterised in that, The computer program product comprises a computer program, and the computer program is executed by the processor to implement the method in any one of claims 1 to 7.

Images