Question Generation (QG)

Question Generation (QG) aims to generate natural language questions based on given contents (knowledge base triples, tables, sentences, or images), where the generated questions need to be able to be answered by the contents. The motivation of QG task is two-fold: (i) transforming customized contents into Q-A pairs, which can be easily used to build customized QA or dialogue systems; (ii) generating large scale Q-A pairs with acceptable quality, which can either be used as additional QA model training data, or improve the efficiency of human annotation on QA dataset construction.

Recently, we are working on three QG tasks, including Structured data-based QG, which geneartes questions from knowledge base sub-graphs or semi-structured tables, Text-based QG, which generates questions from natural language texts, and Image-based QG, which generates questions from images. Some of our work are listed below:

• Yikang Li, Nan Duan, Bolei Zhou, Xiao Chu, Wanli Ouyang, Xiaogang Wang, “Visual Question Generation as Dual Task of Visual Question Answering”, CVPR, 2018. (Image-based QG)
• Duyu Tang, Nan Duan, Zhao Yan, Zhirui Zhang, Yibo Sun, Shujie Liu, Yuanhua Lv, Ming Zhou, “Learning to Collaborate for Question Answering and Asking”, NAACL, 2018. (Structured data-based QG)
• Junwei Bao, Duyu Tang, Nan Duan, Zhao Yan, Yuanhua Lv, Ming Zhou, Tiejun Zhao, “Table-to-Text: Describing Table Region with Natural Language”, AAAI, 2018. (Structured data-based QG)
• Nan Duan, Duyu Tang, Peng Chen, Ming Zhou, “Question Generation for Question Answering”, EMNLP, 2017. (Text-based QG)
• Duyu Tang, Nan Duan, Tao Qin, Zhao Yan, Ming Zhou, “Question Answering and Question Generation as Dual Tasks”, arXiv, 2017. (Text-based QG)