Knowledge injection method and device for large language model, equipment and storage medium
By building a domain knowledge base and integrating query statements and knowledge expression vectors, the problem of difficult knowledge injection in a specific domain of large language models is solved, and the reasoning effect and adaptability are improved.
CN119691095APending Publication Date: 2025-03-25TONGFANG KNOWLEDGE DIGITAL PUBLISHING TECH CO LTD
0 Cites 0 Cited by
Patent Information
- Application Number
- CN202411714693.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-27
- Publication Date
- 2025-03-25
AI Technical Summary
Technical Problem
Large language models are difficult to involve knowledge in specific fields during the knowledge injection process, resulting in poor reasoning.
Method used
By obtaining domain knowledge data, building a domain knowledge base, recording the knowledge expression vectors of different knowledge points, and when receiving the query statement, the target knowledge expression vector is retrieved according to the query intention, fused it with the query statement and entered into a large language model for inference.
Benefits of technology
The knowledge injection of large language models in specific fields is realized, the inference effect and adaptability to query statements is improved, and the shortcomings of using public training corpus are avoided.
✦ Generated by Eureka AI based on patent content.
Abstract
The invention discloses a knowledge injection method and device for a large language model, equipment and a storage medium, relates to the technical field of natural language processing, and can improve the adaptability of the large language model to query statements while ensuring the reasoning effect of the large language model. The method comprises the steps that domain knowledge data is obtained, the domain knowledge data comprises knowledge points which are obtained through format conversion of document resources in a domain and have a preset format, a domain knowledge base is constructed according to the domain knowledge data, and the domain knowledge base is used for recording knowledge expression vectors corresponding to different knowledge points in the domain, the knowledge expression vector is obtained by vectorizing knowledge points through an algorithm model, when a query statement is received, according to a query intention corresponding to the query statement, retrieving a target knowledge expression vector associated with the query intention in the domain knowledge base, and fusing the target knowledge expression vector and the query statement, and inputting the fused target knowledge expression vector and the query statement into the large language model for reasoning so as to inject domain knowledge into the large language model.
Need to check novelty before this filing date? Find Prior Art