CN-122021697-A - Skill construction method and device, electronic equipment and storage medium

CN122021697ACN 122021697 ACN122021697 ACN 122021697ACN-122021697-A

Abstract

The embodiment of the application discloses a skill construction method, a skill construction device, electronic equipment and a storage medium, relates to the technical field of artificial intelligence, and is designed for providing a new skill source mode, reducing skill construction cost and expanding skill coverage. The skill construction method comprises the steps of obtaining interaction track information of a user and an intelligent agent, obtaining skill construction data based on the interaction track information, describing a problem to be solved by a target skill to be constructed and a solution of the problem, and constructing the target skill based on the skill construction data to generate a target skill file package. The embodiment of the application can be used for the application scene of the intelligent agent.

Inventors

Zou Guanyun
LIU ZEMING

Assignees

物自体(上海)科技有限公司

Dates

Publication Date: 20260512
Application Date: 20251231

Claims (10)

1. A method of skill construction, the method comprising: Acquiring interaction track information of a user and an intelligent agent; Acquiring skill construction data based on the interaction track information, wherein the skill construction data describes a problem to be solved by a target skill to be constructed and a solution of the problem; Based on the skill construction data, constructing the target skill to generate a target skill file package.
2. The skill construction method according to claim 1, wherein the acquiring skill construction data based on the interaction trajectory information comprises: and constructing triplet data comprising a question segment, an answer segment and a logical reasoning chain segment based on the interaction track information, and taking the triplet data as the skill construction data.
3. The skill construction method according to claim 2, wherein after the obtaining of the interaction trajectory information of the user and the agent, the method further comprises: Screening high-quality interaction track information meeting preset quality requirements from the acquired interaction track information; based on the interaction track information, constructing triplet data comprising a question segment, an answer segment and a logical reasoning chain segment, wherein the triplet data as the skill construction data comprises: And constructing triplet data comprising question segments, answer segments and logical reasoning chain segments based on the screened high-quality interaction track information, and taking the triplet data as the skill construction data.
4. A skill construction method according to claim 3, wherein the interaction trajectory information comprises interaction sub-trajectory information of at least one complete task execution process; based on the screened high-quality interaction track information, constructing triplet data comprising a question segment, an answer segment and a logical reasoning chain segment, wherein the triplet data as the skill construction data comprises: Extracting and summarizing a target task of a user in a track information initial section for each interaction sub-track information in each high-quality interaction track information, and determining the target task as a problem section; For each piece of interaction sub-track information, recording a final answer adopted by a track information end section user, and determining the final answer as an answer section; And analyzing the interaction information of the track middle section of each interaction sub-track, and converting the interaction information into natural language description of the reasoning process to form a logic reasoning chain segment.
5. The method of claim 1, wherein constructing the target skills based on the skill construction data to generate a target skill package comprises: generating a target skill file package based on the problem solved by the target skill, the solution of the problem and a preset skill template, wherein the target skill file package comprises: An entry file comprising a function description, a thought chain description, and a routing rule; A reference file; Wherein the function description indicates a problem to be solved by the target skill, the thought chain describes an inference process for describing a solution of the problem, the reference file includes a file to be invoked in the solution of the problem described in the thought chain description, and the routing rule is used for indicating a invoking condition of the reference file.
6. A method of skill construction according to claim 5, wherein the mental chain description comprises: a checklist describing contents to be checked by checkpoints in the solution or input information to be acquired in the solution; a sequence of steps describing the execution steps in the solution; Branching conditions describing logic judgment conditions of different execution paths in the solution; Feedback loop information describing processing principles for the execution result of the execution step or external feedback.
7. A method of skill construction according to claim 1, wherein after the construction of the target skill, the method further comprises: releasing the target skills so that the intelligent agent can solve the user problem by using the target skills; Acquiring newly-added interaction track information after the target skills are released; And evaluating the target skills based on the newly added interaction track information to determine whether the target skills need to be repaired.
8. A skill building apparatus, the apparatus comprising: The first acquisition unit is used for acquiring interaction track information of the user and the intelligent agent; A second obtaining unit configured to obtain skill construction data describing a problem to be solved by a target skill to be constructed and a solution of the problem, based on the interaction trajectory information; And the construction unit is used for constructing the target skills based on the skill construction data to generate a target skill file packet.
9. An electronic device comprising a housing, a processor, a memory, a circuit board and a power supply circuit, wherein the circuit board is arranged inside a space enclosed by the housing, the processor and the memory are arranged on the circuit board, the power supply circuit is used for supplying power to each circuit or device of the electronic device, the memory is used for storing executable program codes, and the processor is used for executing a program corresponding to the executable program codes by reading the executable program codes stored in the memory and executing the skill construction method of any one of the claims 1 to 7.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores one or more programs, the one or more programs being executable by one or more processors to implement the skill building method of any of the preceding claims 1 to 7.

Description

Skill construction method and device, electronic equipment and storage medium Technical Field The present application relates to the field of artificial intelligence technologies, and in particular, to a skill construction method, apparatus, electronic device, and storage medium. Background With the rapid development of large language models (Large Language Model, LLM), more and more individuals and enterprises accomplish a variety of complex tasks such as writing of documents, data analysis, advertisement diagnosis, and content marketing planning through conversational agents (agents). To reduce the degree of coupling, and improve the reusability and maintainability of an agent system, the industry has gradually explored a development paradigm with "Skill" as a core abstraction, a standardized and modularized agent capability packaging unit, which teaches an agent to execute a specific workflow by packaging instructions, scripts, resource organizations, etc. required to complete a specific task into a standardized folder. This development paradigm, with skills as the core, divides the capabilities of an agent into a plurality of combinable, operational capability units. At present, skills are highly dependent on manual design and maintenance of experts, so that the cost of skill construction is high, and in addition, the skill construction is usually based on cognition and experience of the experts, so that the coverage of the skills is limited, and further, the task processing capacity of an intelligent agent is limited. Disclosure of Invention In view of this, the embodiment of the application provides a skill construction method, a device, an electronic device and a storage medium, and provides a new skill source mode, so that the automation degree of the construction process is effectively improved, the construction cost is reduced, the coverage of the skill is obviously expanded, and the task processing capability of an intelligent agent can be effectively enhanced. In a first aspect, an embodiment of the present application provides a method for constructing a skill, where the method includes obtaining interaction track information of a user and an agent, obtaining skill construction data based on the interaction track information, where the skill construction data describes a problem to be solved by a target skill to be constructed and a solution to the problem, and constructing the target skill based on the skill construction data to generate a target skill file package. According to one embodiment of the application, acquiring skill construction data based on the interaction trajectory information comprises constructing triplet data comprising question segments, answer segments and logical inference segments as the skill construction data based on the interaction trajectory information. According to one embodiment of the application, after the interaction track information of the user and the intelligent agent is obtained, the method further comprises screening high-quality interaction track information meeting the preset quality requirement from the obtained interaction track information, and constructing triplet data comprising a question section, an answer section and a logic reasoning chain segment based on the interaction track information, wherein the constructing of the skill construction data comprises constructing the triplet data comprising the question section, the answer section and the logic reasoning chain segment based on the screened high-quality interaction track information, and the constructing of the triplet data comprising the question section, the answer section and the logic reasoning chain segment is performed as the skill construction data. According to one embodiment of the application, screening out the interaction track information meeting the preset quality requirement from the acquired interaction track information comprises extracting task integrity features, user evaluation features and business index features corresponding to the interaction track information for each piece of interaction track information, generating multi-dimensional feature vectors based on the task integrity features, the user evaluation features and the business index features, and judging the high-quality interaction track information based on the multi-dimensional feature vectors through a pre-trained classification model. According to one embodiment of the application, the interactive track information comprises interactive sub-track information of at least one complete task execution process, the construction of the triplet data comprising question segments, answer segments and logical reasoning segments based on the screened high-quality interactive track information comprises the steps of extracting and summarizing a target task of a user in a track information initial segment for each interactive sub-track information in each high-quality interactive track information, determining the t