Leveraging Structured Information from a Passage to Generate Questions

作     者：Jian Xu Yu Sun Jianhou Gan Mingtao Zhou Di Wu Jian Xu;Yu Sun;Jianhou Gan;Mingtao Zhou;Di Wu

作者机构：Key Laboratory of Educational Informatization for NationalitiesYunnan Normal UniversityKunming 650500China School of Information Science and TechnologyYunnan Normal UniversityKunming 650500China Yunnan Key Laboratory of Smart EducationYunnan Normal UniversityKunming 650500China School of Information EngineeringQujing Normal UniversityQujing 655011China 

出 版 物：《Tsinghua Science and Technology》 (清华大学学报（自然科学版（英文版）)

年 卷 期：2023年第28卷第3期

页      面：464-474页

核心收录：

学科分类：0831[工学-生物医学工程（可授工学、理学、医学学位）] 0711[理学-系统科学] 12[管理学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 07[理学] 08[工学] 081201[工学-计算机系统结构] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：This work was supported by the National Natural Science Foundation of China(No.62166050) Yunnan Fundamental Research Projects(No.202201AS070021) Yunnan Innovation Team of Education Informatization for Nationalities,Scientific Technology Innovation Team of Educational Big Data Application Technology in University of Yunnan Province,and Yunnan Normal University Graduate Research and innovation fund in 2020(No.ysdyjs2020006) 

主　　题：automatic Question Generation(QG) RACE4QG dataset Answer-Oriented GAT(AO-GAT) attention mechanism structured information 

摘      要：Question Generation(QG)is the task of utilizing Artificial Intelligence(AI)technology to generate questions that can be answered by a span of text within a given *** research on QG in the educational field struggles with two challenges:the mainstream QG models based on seq-to-seq fail to utilize the structured information from the passage;the other is the lack of specialized educational QG *** address the challenges,a specialized QG dataset,reading comprehension dataset from examinations for QG(named RACE4QG),is reconstructed by applying a new answer tagging approach and a data-filtering strategy to the RACE ***,an end-to-end QG model,which can exploit the intra-and inter-sentence information to generate better questions,is *** our model,the encoder utilizes a Gated Recurrent Units(GRU)network,which takes the concatenation of word embedding,answer tagging,and Graph Attention neTworks(GAT)embedding as *** hidden states of the GRU are operated with a gated self-attention to obtain the final passage-answer representation,which will be fed to the *** show that our model outperforms baselines on automatic metrics and human ***,the model improves the baseline by 0.44,1.32,and 1.34 on BLEU-4,ROUGE-L,and METEOR metrics,respectively,indicating the effectivity and reliability of our *** gap with human expectations also reflects the research potential.