Common sense knowledge scoring and ranking method and apparatus

CN114861913BActive Publication Date: 2026-06-23TSINGHUA UNIVERSITY
View PDF 1 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
TSINGHUA UNIVERSITY
Filing Date
2022-03-23
Publication Date
2026-06-23

AI Technical Summary

Technical Problem

The low accuracy of common sense knowledge scoring and ranking in existing technologies is mainly due to the fact that only the contextual information of the triples themselves is used, resulting in a single source of verification information and a lack of non-co-occurring semantic knowledge.

Method used

By combining ConceptNet and HowNet knowledge bases, we obtain the semantic primitive tree group of the common sense triple object to be scored, and use a scoring model for comprehensive scoring, including a first parsing module, a second parsing module, and a voting module. We comprehensively analyze the information of the triple itself and the evidence from the external knowledge base to improve the accuracy of scoring.

Benefits of technology

By combining information from triples and semantic primitives, the accuracy of scoring common sense knowledge was improved, thereby enhancing the accuracy of ranking common sense knowledge.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114861913B_ABST
    Figure CN114861913B_ABST
Patent Text Reader

Abstract

The application provides a common sense knowledge scoring and sorting method and device, wherein the common sense knowledge scoring method comprises the following steps: obtaining a common sense triple object to be scored and a semantic primitive tree group corresponding to the common sense triple object to be scored; inputting the common sense triple object to be scored and the semantic primitive tree group into a scoring model to obtain a voting result output by the scoring model; wherein the scoring model is obtained by training based on a sample group and corresponding voting labels; the sample group comprises common sense triple samples and semantic primitive tree group samples corresponding to the common sense triple samples; the voting labels are determined in advance according to the sample group and correspond to the sample group one by one. The common sense knowledge scoring and sorting method and device provided by the application realize comprehensive scoring of the common sense triple object to be scored by combining the triple itself information and the semantic primitive information of the external knowledge base, and improve the accuracy of common sense knowledge scoring and sorting.
Need to check novelty before this filing date? Find Prior Art