一种可控难度的常识推理考题生成方法
By acquiring text entities and reasoning relationships, and combining common sense knowledge graphs and large language models, we can generate common sense reasoning questions with varying difficulty levels. This solves the problem of insufficient machine modeling ability for hidden common sense clues, and enables precise control of question difficulty and assessment of comprehensive thinking ability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SUN YAT SEN UNIV
- Filing Date
- 2024-04-15
- Publication Date
- 2026-07-17
AI Technical Summary
In existing technologies, deterministic machines have a weak ability to model uncertain reasoning processes that hide common sense clues, resulting in unsatisfactory test questions and difficulty in controlling the complexity of test questions to meet the needs of students of different levels.
By extracting entities and reasoning relationships from the text, supplementing external common sense with common sense knowledge graphs and large language models, generating sub-problems using soft templates, and generating common sense reasoning test questions of multiple difficulty levels through combination and sorting, the difficulty and rationality of the test questions are verified using the BART model and various voting functions.
It achieves precise control over the difficulty of test questions, generates questions that require multiple steps of logical reasoning to answer, comprehensively examines the test taker's integrated thinking ability, and improves the quality and controllability of test questions.
Smart Images
Figure CN118246559B_ABST