閱讀是人們從外界獲取知識的一種重要渠道，而通過問題可以進一步加強人們對文本的理解，提高對重點知識的吸收程度。近年來，問題生成逐漸進入了人們的視野，成為自然語言處理領域內的研究熱點之一。問題生成任務的是通過給定一個句子或一段文章，讓系統自動生成相關的問題，而已知答案的問題生成是大部分工作的重點。隨著深度學習的快速發展，問題生成也從原先基於規則與範本的方法轉變成基於神經網路進行生成。然而如今大部分的問題生成工作仍然存在一些問題，包括模型時常無法正確理解文本的語義，在生成過程中發生誤用，或是經常生成有著錯誤問題類型的問題，這些都極大影響了問題的生成品質。

針對這些問題，本文在已知答案的段落級問題生成的基礎上進行改進。該模型遵循編碼器-解碼器的結構，在編碼器中引入門控自注意力機制，幫助模型合理處理長文本資訊；在解碼器中使用最大指針網路來解決重複生成問題。基於以上模型架構，我們在詞嵌入層引入更豐富的詞彙特徵以及常識知識，來增強輸入文檔的表示，幫助模型更好地理解文本語義；還引入了答案-問題類型的編碼器，使模型更好利用答案資訊，幫助模型提高問題類型生成的準確率。

最後，我們使用客觀評估以及主觀人工評估來衡量模型的生成品質，並取得了優於基線模型的表現。

Reading is one of the most basic and significant ways for people to acquire knowledge, and questions can further enhance people's understanding of the text and improve their absorption of key knowledge.
In the past few years, question generation (QG) has gradually entered the vision of researchers and become one of the hot research topics in the field of natural language processing. Question generation is a task that let the system automatically generate relevant questions by giving a natural language text, and answer-aware question generation is the keynote of most of the QG work. As deep learning improves by leaps and bounds, the mainstream method of question generation has changed from a rule-based and template-based approach to the neural network. However, most of the question generation work still has some problems, including models that often fail to understand the semantic meaning of the text correctly, or often generate questions with wrong question types, which greatly affects the quality of question generation.

In view of the above problems, this thesis improves the question generation on an answer-aware paragraph-level question generation model.
This model follows the encoder-decoder architecture and introduces a gated self-attention mechanism in the encoder to help the model reasonably handle long text information. In the decoder, it utilizes a maxout pointer network to solve the repetition problem. Based on the above model architecture, we introduce richer lexical features and commonsense knowledge in the embedding layer to enhance the representation of input text and help the model better understand the semantics of text. An AQT (answer\&question type) encoder is also introduced to make better use of answer information and help the model improve the accuracy of question type generation.

Finally, we employed automated evaluation metrics as well as manual evaluation to measure the questions' quality and achieve better performance than the baseline model.

Abstract (Chinese) I
Abstract II
Acknowledgements (Chinese) IV
Contents V
List of Figures VIII
List of Tables IX
1 Introduction 1
1.1 Motivation 1
1.2 Objectives 3
1.3 Significance and Contribution of the Research 4
2 Background and Related Work 6
2.1 Part-of-speech Tagging 6
2.2 Named Entity Recognition 7
2.3 Word Embedding 7
2.4 Attention Mechanisms 8
2.5 Sequence-to-Sequence Framework 8
2.6 Literature Review on Question Generation 9
3 Methodology 12
3.1 Overview and architecture 12
3.2 The Commonsense Interpreter 13
3.3 Feature-Enriched Paragraph Encoder 14
3.3.1 Feature-Enriched Embedding Layer 14
3.3.2 Paragraph Encoder 16
3.4 Answer-Question Type Encoder 17
3.5 The Decoder as the question generator 19
4 Experiments 22
4.1 Dataset and Metrics 22
4.1.1 SQuAD 22
4.1.2 Commonsense Knowledge Base 22
4.1.3 Metrics 23
4.2 Experiments setup 25
4.3 Baseline 26
4.4 Objective Evaluation 26
4.4.1 Main Results 26
4.4.2 Question Type Accuracy 27
4.5 Subjective Evaluation 29
4.6 The Ablation Study 30
4.7 Case Study 32
4.8 Discussion 34
5 Conclusion and Future Work 36
Bibliography 38
Appendices 45
A Questionnaire Design 45
B Question Samples 48

[1] Vasile Rus, Zhiqiang Cai, and Art Graesser. Question generation: Example of a multi-year evaluation campaign. Online Proceedings of 1st Question Generation Workshop, 2008.
[2] Megha Mishra, Vishnu Kumar Mishra, and HR Sharma. Question classification using semantic, syntactic and lexical features. International Journal of Web & Semantic Technology, 4(3):39, 2013.
[3] Pranav Rajpurkar, Jian Zhang, Konstantin Lopyrev, and Percy Liang. Squad: 100,000+ questions for machine comprehension of text. The 2016 Conference on Empirical Methods in Natural Language Processing, pages 2383–2392, 2016.
[4] Anselm Rothe, Brenden M. Lake, and Todd M. Gureckis. Question asking as program generation. Neural Information Processing Systems (NIPS), page 1046–1055, 2017.
[5] Michael Heilman and Noah A Smith. Good question! statistical ranking for question generation. In Human Language Technologies: The 2010 Annual Conference of the North American Chapter of the Association for Computational Linguistics, pages 609–617, 2010.
[6] Yansen Wang, Chenyi Liu, Minlie Huang, and Liqiang Nie. Learning to ask questions in open-domain conversational systems with typed decoders. In Proceedings of the 56th Annual Meting of the Association for Computational Linguistics, pages 2193–2203, 2018.
[7] Diana Lea and Jennifer Bradbery. Oxford Advanced Learner’s Dictionary 10th Edition. Oxford Advanced Learner’s Dictionary. Oxford University Press, 2020.
[8] Rahul Sharnagat. Named entity recognition: A literature survey. Center For Indian Language Technology, pages 1–27, 2014.
[9] Ralph Grishman and Beth M Sundheim. Message understanding conference-6: A brief history. In COLING 1996 Volume 1: The 16th International Conference on Computational Linguistics, 1996.
[10] Christopher D Manning, Mihai Surdeanu, John Bauer, Jenny Rose Finkel, Steven Bethard, and David McClosky. The stanford corenlp natural language processing toolkit. In Proceedings of 52nd annual meeting of the association for computational linguistics: system demonstrations, pages 55–60, 2014.
[11] Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781, 2013.
[12] Matthew E. Peters, Mark Neumann, Mohit Iyyer, Matt Gardner, Christopher Clark, Kenton Lee, and Luke Zettlemoyer. Deep contextualized word representations. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers), pages 2227–2237, New Orleans, Louisiana, June 2018. Association for Computational Linguistics.
[13] Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901, 2020.
[14] Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805, 2018.
[15] Jeffrey Pennington, Richard Socher, and Christopher D Manning. Glove: Global vectors for word representation. In Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP), pages 1532–1543, 2014.
[16] Sneha Chaudhari, Varun Mithal, Gungor Polatkan, and Rohan Ramanath. An attentive survey of attention models. ACM Transactions on Intelligent Systems and Technology (TIST), 12(5):1–32, 2021.
[17] Ilya Sutskever, Oriol Vinyals, and Quoc V Le. Sequence to sequence learning with neural networks. Advances in neural information processing systems, 27, 2014.
[18] Ruslan Mitkov et al. Computer-aided generation of multiple-choice tests. In Proceedings of the HLT-NAACL 03 workshop on Building educational applications using natural language processing, pages 17–22, 2003.
[19] George A Miller. WordNet: An electronic lexical database. MIT press, 1998.
[20] Qingyu Zhou, Nan Yang, Furu Wei, Chuanqi Tan, Hangbo Bao, and Ming Zhou. Neural question generation from text: A preliminary study. In National CCF Conference on Natural Language Processing and Chinese Computing,
pages 662–671. Springer, 2017.
[21] Xinya Du, Junru Shao, and Claire Cardie. Learning to ask: Neural question generation for reading comprehension. arXiv preprint arXiv:1705.00106, 2017.
[22] Yanghoon Kim, Hwanhee Lee, Joongbo Shin, and Kyomin Jung. Improving neural question generation using answer separation. In Proceedings of the AAAI conference on artificial intelligence, volume 33, pages 6602–6609, 2019.
[23] Xiyao Ma, Qile Zhu, Yanlin Zhou, and Xiaolin Li. Improving question generation with sentence-level semantic matching and answer position inferring. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 34, pages 8464–8471, 2020.
[24] Yuichi Sasazawa, Sho Takase, and Naoaki Okazaki. Neural question generation using interrogative phrases. In Proceedings of the 12th International Conference on Natural Language Generation, pages 106–111, 2019.
[25] Wenjie Zhou, Minghua Zhang, and Yunfang Wu. Question-type driven question generation. arXiv preprint arXiv:1909.00140, 2019.
[26] Yao Zhao, Xiaochuan Ni, Yuanyuan Ding, and Qifa Ke. Paragraph-level neural question generation with maxout pointer and gated self-attention networks. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 3901–3910, 2018.
[27] Yu Chen, Lingfei Wu, and Mohammed J Zaki. Reinforcement learning based graph-to-sequence model for natural question generation. arXiv preprint arXiv:1908.04942, 2019.
[28] Deepak Gupta, Kaheer Suleman, Mahmoud Adada, Andrew McNamara, and
Justin Harris. Improving neural question generation using world knowledge. arXiv preprint arXiv:1909.03716, 2019.
[29] Xin Jia, Hao Wang, Dawei Yin, and Yunfang Wu. Enhancing question generation with commonsense knowledge. In China National Conference on Chinese Computational Linguistics, pages 145–160. Springer, 2021.
[30] Robyn Speer, Joshua Chin, and Catherine Havasi. Conceptnet 5.5: An open multilingual graph of general knowledge. In Thirty-first AAAI conference on artificial intelligence, 2017.
[31] Xin Jia, Wenjie Zhou, Xu Sun, and Yunfang Wu. How to ask good questions? try to leverage paraphrases. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 6130–6140, 2020.
[32] Hao Zhou, Tom Young, Minlie Huang, Haizhou Zhao, Jingfang Xu, and Xiaoyan Zhu. Commonsense knowledge aware conversation generation with
graph attention. In IJCAI, pages 4623–4629, 2018.
[33] Antoine Bordes, Nicolas Usunier, Alberto Garcia-Duran, Jason Weston, and Oksana Yakhnenko. Translating embeddings for modeling multi-relational
data. Advances in neural information processing systems, 26, 2013.
[34] Sepp Hochreiter and J ̈urgen Schmidhuber. Long short-term memory. Neural computation, 9(8):1735–1780, 1997.
[35] Qingbao Huang, Mingyi Fu, Linzhang Mo, Yi Cai, Jingyun Xu, Pijian Li, Qing Li, and Ho-fung Leung. Entity guided question generation with contextual structure and sequence information capturing. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, pages 13064–13072, 2021.
[36] Minh-Thang Luong, Hieu Pham, and Christopher D Manning. Effective
approaches to attention-based neural machine translation. arXiv preprint arXiv:1508.04025, 2015.
[37] Jiatao Gu, Zhengdong Lu, Hang Li, and Victor OK Li. Incorporat-
ing copying mechanism in sequence-to-sequence learning. arXiv preprint
arXiv:1603.06393, 2016.
[38] Luu Anh Tuan, Darsh Shah, and Regina Barzilay. Capturing greater context for question generation. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 34, pages 9065–9072, 2020.
[39] Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th annual meeting of the Association for Computational Linguistics, pages 311–318, 2002.
[40] Michael Denkowski and Alon Lavie. Meteor universal: Language specific translation evaluation for any target language. In Proceedings of the ninth workshop on statistical machine translation, pages 376–380, 2014.
[41] Chin-Yew Lin. Rouge: A package for automatic evaluation of summaries. In Text summarization branches out, pages 74–81, 2004.
[42] Xu Chen and Jungang Xu. An answer driven model for paragraph-level question generation. In 2021 International Joint Conference on Neural Networks(IJCNN), pages 1–7. IEEE, 2021