Open domain question answering method based on knowledge graph and related device

By extracting entity attribute candidate sets, mining and combining paths from knowledge graphs, and combining similarity scores from classification models to recommend answers, the problems of low recall and inaccurate path queries in existing technologies are solved, resulting in more accurate question-answering results.

CN115470328BActive Publication Date: 2026-03-03NAOPU CLOUD (SUZHOU) ARTIFICIAL INTELLIGENCE TECHNOLOGY CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202210964248.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-08-11
Publication Date
2026-03-03
Estimated Expiration
2042-08-11

AI Technical Summary

Technical Problem

Existing knowledge graph-based question answering technologies suffer from low recall rates when extracting entities and attributes, and lack of constraints when querying paths, resulting in inaccurate question answering results.

Method used

Entity attribute candidate sets are extracted using multiple matching methods, path mining and combination are performed, restricted combination paths are selected, and similarity scores are calculated using a pre-trained classification model to recommend answers.

Benefits of technology

It improved the recall rate of entity attributes, reduced the recall of invalid paths, and improved the accuracy of question-answering results and user experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115470328B_ABST
    Figure CN115470328B_ABST
Patent Text Reader

Abstract

This application provides an open-domain question answering method and related equipment based on knowledge graphs. The method extracts entity and attribute candidate sets from the question description using multiple matching methods, effectively improving the recall rate of entity attribute extraction and avoiding missing hidden information in the question description. Path mining is performed based on the entity attribute candidate sets to obtain initial paths. These initial paths are then combined and filtered to obtain restricted combination paths with constraints, effectively reducing the recall of invalid paths. Precise path modeling is performed for question descriptions with multiple constraints, thereby reducing the computational cost of subsequent path-related tasks. A classification model outputs a similarity score between the restricted combination paths and the question description. The knowledge graph subgraph corresponding to the restricted combination path with the highest similarity score is used as the recommended answer data, improving the user experience.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Case information semantic retrieval method and device based on knowledge graph

    CN111475623A

  • Complex question answering method and device based on knowledge graph and storage medium

    CN113468311A