西北师范大学学报（自然科学版）2025，Vol.61Issue(4)：82-90,9.DOI:10.16783/j.cnki.nwnuz.2025.04.010
知识图谱嵌入的机械故障问题生成
Mechanical failure problem generation based on knowledge graph embedding
摘要
Abstract
Incorporating knowledge graphs into pre-trained models significantly enhances question generation for mechanical failure domains,a task critical for improving intelligence in question answering systems and information search.Traditional methods,geared towards general fields,often falter in specialized areas due to insufficient professional training data,leading to misunderstandings.This paper introduces a method to embed a knowledge graph into the pre-trained T5 model,adapting it for vertical domain question generation by refining input,output,and training strategies.The approach enhances the model's comprehension of mechanical failure knowledge,leading to substantial improvements in BLEU-4 and ROUGE-L scores and generating more diverse and interpretable questions.The effectiveness of knowledge graph embedding in bolstering question generation quality is thereby demonstrated.关键词
机械故障/问题生成/知识图谱/预训练模型Key words
mechanical failure/question generation/knowledge graph/pre-trained model分类
信息技术与安全科学引用本文复制引用
陈红红,王志涛,骆军军,董晓辉..知识图谱嵌入的机械故障问题生成[J].西北师范大学学报（自然科学版）,2025,61(4):82-90,9.基金项目
国家自然科学基金资助项目(62167007) （62167007）
教育部人文社会科学规划基金(24YJA880007) （24YJA880007）
