CN-121981674-A - Real-time intelligent auxiliary method and system based on enterprise micro-seat

CN121981674ACN 121981674 ACN121981674 ACN 121981674ACN-121981674-A

Abstract

The application relates to a real-time intelligent auxiliary method and a system based on enterprise micro agents, which belong to the technical field of man-machine interaction. And analyzing the client intention and the dialogue theme through natural language processing, and combining the dialogue history and the client portrait to output a dynamic context feature vector. The vector is matched with a preset policy rule to generate an auxiliary instruction set. The system calls multi-source data according to the instruction type, generates formatted auxiliary content blocks after screening, and distributes the formatted auxiliary content blocks to native components such as sidebars, shortcut replies and the like through an enterprise WeChat interface. The system captures the seat operation to update the context at the same time, and realizes closed-loop interaction. When the instruction relates to flow collaboration, tasks are automatically created and the service system states are synchronized to form a continuously optimized auxiliary loop. The application improves the response efficiency, service specialty and work satisfaction of the seat.

Inventors

XI JUN
ZHANG JINGYI
CHEN GUIBO

Assignees

北京智保惠众数字科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260120

Claims (10)

1. The real-time intelligent auxiliary method based on the enterprise micro seat is characterized by comprising the following steps of: Collecting text messages, voice-to-text messages and associated client identity data in an enterprise WeChat dialogue stream in real time, and simultaneously monitoring a flow event triggered by a service system to generate a standardized context data packet; Analyzing client intention and dialogue theme through natural language processing model based on the standardized context data packet, marking current dialogue stage by combining dialogue history sequence, and associating client relation management system to complement client portrait, and outputting dynamic context feature vector; Inputting the dynamic context feature vector into a predefined auxiliary strategy rule base for matching, and generating an auxiliary instruction set containing instruction types and parameters; according to the instruction type of the auxiliary instruction set, a corresponding data source interface is called to acquire original information, redundant contents are filtered according to a preset scene screening rule, and a formatted auxiliary content block is generated; Distributing the formatted auxiliary content blocks to corresponding original interaction components through an enterprise WeChat open platform interface, and triggering a sidebar to dynamically refresh, quickly reply to recommend or apply message reminding; Capturing the operation behavior of the agent in the original interaction component, updating the dynamic context feature vector and triggering a new round of auxiliary instruction set to generate; When the auxiliary instruction set contains a flow cooperation instruction, a cooperation task is automatically created, the service system state is synchronously updated, and a service system state change event is used as input of a new round of data acquisition.
2. The method for real-time intelligent assistance based on enterprise micro agents according to claim 1, wherein the step of analyzing the client intention and the dialogue topic through a natural language processing model based on the standardized context data packet, marking the current session stage in combination with the session history sequence, associating the client relationship management system to complement the client portrait, and outputting the dynamic context feature vector comprises the following steps: extracting a current dialogue text and a historical dialogue sequence from the standardized context data packet; carrying out semantic analysis on the current dialogue text, and outputting a structured semantic label comprising a client intention label, a dialogue topic label and an emotion tendency value; marking the current session stage type through a session stage judging model based on the historical dialog sequence and the client intention labels in the structured semantic labels; calling an interface of a client relation management system to acquire historical work order records and risk level data according to the client identity data in the standardized context data packet, and generating an enhanced client portrait; integrating the structured semantic tags, the current session stage type and the enhanced customer portrayal, and encoding into a dynamic context feature vector.
3. The method of claim 2, wherein the step of inputting the dynamic context feature vector into a predefined auxiliary policy rule base for matching, generating an auxiliary instruction set comprising instruction types and parameters comprises: Receiving the dynamic context feature vector, and extracting a client intention label, a dialogue topic label, a current dialogue stage type and an enhanced client portrait from the dynamic context feature vector; Performing multi-dimensional matching on the extracted features and a predefined auxiliary policy rule base to obtain a multi-dimensional matching result, wherein matching logic associates a combination condition of a client intention label, a dialogue topic label and a dialogue stage type; determining an instruction type according to the multi-dimensional matching result, and generating instruction parameters based on the client risk level in the enhanced client portrait and the historical work order record; The encapsulated instruction type and instruction parameters generate an auxiliary instruction set.
4. The method for real-time intelligent auxiliary based on enterprise micro agents according to claim 1, wherein the step of calling a corresponding data source interface to obtain original information according to the instruction type of the auxiliary instruction set, filtering redundant content according to a preset scene filtering rule, and generating formatted auxiliary content blocks comprises the steps of: analyzing the instruction type and the instruction parameters in the auxiliary instruction set, and selecting a corresponding data source interface according to the instruction type; Constructing a data query request based on the instruction parameters, and initiating a call to a selected data source interface to acquire an original information set; loading a preset scene screening rule according to the dialogue topic label and the customer intention label in the dynamic context feature vector; Filtering the original information set by applying the scene screening rule, and removing redundant contents irrelevant to the current dialogue theme and the client intention; And formatting and rendering the filtered information according to a predefined template structure to generate an auxiliary content block of the adaptation enterprise WeChat interaction component.
5. The method of claim 1, wherein capturing operational behavior of an agent in the native interaction component, updating dynamic context feature vectors and triggering a new round of auxiliary instruction set generation comprises: monitoring operation events of agents in an enterprise WeChat native interaction component, and capturing operation types and associated data; According to the operation type, analyzing semantic influence, and updating a session stage mark and a service state mark in the dynamic context feature vector; and re-inputting the updated dynamic context feature vector into a predefined auxiliary strategy rule base, and triggering a new round of auxiliary instruction set generation.
6. The method for intelligent real-time assistance based on enterprise micro agents according to claim 5, wherein when the assistance instruction set contains a process collaboration instruction, the steps of automatically creating a collaboration task and synchronously updating the business system state, and taking the business system state change event as an input of a new round of data collection include: when the instruction type of the auxiliary instruction set is determined to be flow collaboration, analyzing a collaboration template identification and target node information in instruction parameters; Invoking a predefined collaboration rule engine according to the collaboration template identification to generate a collaboration task entity comprising a task allocation object, an execution time limit and an input data structure; Submitting the collaborative task entity through an interface of a business process management system, and monitoring a real-time state change event; when capturing a task state update event, extracting a task progress identifier in the event and outputting result data; and synchronizing the task progress identification to a service state field in the dynamic context feature vector, and packaging output result data into newly added input of a standardized context data packet.
7. A method of intelligent real-time assistance based on enterprise micro agents as claimed in any one of claims 1 to 6, further comprising, after the step of outputting the dynamic context feature vector: detecting conflict items of customer intention labels and historical service records in the dynamic context feature vector; traversing version change records of associated products in a knowledge base system, and matching the latest policy documents corresponding to the conflict items; extracting a change term key field in the latest policy document to generate a policy difference prompt block; And inserting the strategy difference prompt block into a visual warning area of the formatting auxiliary content block.
8. Real-time intelligent auxiliary system based on little seat of enterprise, its characterized in that, real-time intelligent auxiliary system includes: The data acquisition processing module is used for acquiring text messages, voice-to-text messages and associated client identity data in the enterprise WeChat dialogue stream in real time, and monitoring flow events triggered by the service system to generate a standardized context data packet; The feature modeling module is used for analyzing the client intention and the dialogue theme through a natural language processing model based on the standardized context data packet, marking the current session stage by combining the session history sequence, and associating the client relation management system to complement the client portrait to output a dynamic context feature vector; The intelligent decision module is used for inputting the dynamic context feature vector into a predefined auxiliary strategy rule base for matching, and generating an auxiliary instruction set containing instruction types and parameters; the auxiliary content rendering module is used for calling a corresponding data source interface to acquire original information according to the instruction type of the auxiliary instruction set, filtering redundant content according to a preset scene screening rule and generating a formatted auxiliary content block; The interactive distribution module is used for distributing the formatted auxiliary content blocks to corresponding original interactive components through an enterprise WeChat open platform interface, and triggering dynamic refreshing of a sidebar, quick reply recommendation or application message reminding; The interaction feedback module is used for capturing the operation behavior of the agent in the original interaction component, updating the dynamic context feature vector and triggering a new round of auxiliary instruction set generation; And the business collaborative management module is used for automatically creating a collaborative task and synchronously updating the state of a business system when the auxiliary instruction set contains a flow collaborative instruction, and taking a business system state change event as the input of a new round of data acquisition.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method of any one of claims 1 to 7 when the program is executed.
10. A computer-readable storage medium, characterized in that a computer program is stored that can be loaded by a processor and that performs the method according to any one of claims 1 to 7.

Description

Real-time intelligent auxiliary method and system based on enterprise micro-seat Technical Field The application relates to the technical field of man-machine interaction, in particular to a real-time intelligent auxiliary method and system based on enterprise micro seats. Background With the deep advancement of enterprise digital transformation, an instant communication platform such as enterprise WeChat has gradually become a core channel for service interaction between enterprises and external clients, and plays a key role in high-value and high-complexity consultation scenes such as finance, government affairs and the like. Under the background, the auxiliary technology for serving the agent also undergoes evolution from the conventional knowledge base inquiry to the screen flicking based on the basic information of the client, and aims to improve the response efficiency and the service accuracy of the agent. These prior art techniques typically rely on agents actively querying a static knowledge base or retrieving customer base material from a customer relationship management system for display at session initiation, constituting the mainstream implementation of current enterprise customer service support systems. However, with the increasing complexity of service scenarios and the increasingly demanding efficiency requirements for sitting mat work, the limitations of the existing assistance techniques are also increasingly highlighted. Due to the lack of deep understanding capability of the real-time conversation flow content, the system is difficult to capture the context information dynamically changed in the conversation process, so that the auxiliary information is updated later and cannot be synchronized with the specific topics currently concerned by the clients, the agents often need to interrupt the conversation flow and manually switch among different system interfaces to search for related information, and the process not only breaks the continuity of the service, but also increases the operation burden and the cognitive load. Meanwhile, because information from different sources such as a knowledge base, a business process system and the like is in a splitting state, the information is provided for agents, which is generalized and lacks a targeted information set, and collaborative processing and accurate guiding of complex business processes are difficult to support. Moreover, the integration level of the auxiliary function and the enterprise WeChat primary working environment is insufficient, so that the seat cannot obtain seamless immersive auxiliary experience in a familiar chat window, and the actual application effect of the technology and the working experience of the seat are affected. Disclosure of Invention In order to solve the technical problems, the application provides a real-time intelligent auxiliary method and a system based on enterprise micro seats. In a first aspect, the application provides a real-time intelligent auxiliary method based on enterprise micro seats, which adopts the following technical scheme: a real-time intelligent auxiliary method based on enterprise micro-agents comprises the following steps: Collecting text messages, voice-to-text messages and associated client identity data in an enterprise WeChat dialogue stream in real time, and simultaneously monitoring a flow event triggered by a service system to generate a standardized context data packet; Analyzing client intention and dialogue theme through natural language processing model based on the standardized context data packet, marking current dialogue stage by combining dialogue history sequence, and associating client relation management system to complement client portrait, and outputting dynamic context feature vector; Inputting the dynamic context feature vector into a predefined auxiliary strategy rule base for matching, and generating an auxiliary instruction set containing instruction types and parameters; according to the instruction type of the auxiliary instruction set, a corresponding data source interface is called to acquire original information, redundant contents are filtered according to a preset scene screening rule, and a formatted auxiliary content block is generated; Distributing the formatted auxiliary content blocks to corresponding original interaction components through an enterprise WeChat open platform interface, and triggering a sidebar to dynamically refresh, quickly reply to recommend or apply message reminding; Capturing the operation behavior of the agent in the original interaction component, updating the dynamic context feature vector and triggering a new round of auxiliary instruction set to generate; When the auxiliary instruction set contains a flow cooperation instruction, a cooperation task is automatically created, the service system state is synchronously updated, and a service system state change event is used as input of a new round of data acquisition. By a