Abstract：Multi-hop Question Generation is the task of reasoning over disjoint pieces of information and then generating complex questions. For the given Q&A pair, the context contains a large number of redundant and irrelevant sentences, and most previous methods require annotated corpus to select supporting facts as input to generate corresponding questions. To address this problem, this paper proposes a Graph-based Evidence Selection network (GES) for deep question generation over documents. The proposed model selects informative sentences from disjointed paragraphs, which serves as an inductive bias to refine question generation. We also employ a straight-through estimator to train the model in an end-to-end manner. Experimental results on the HotpotQA dataset demonstrate that our proposed solution outperforms state-of-the-art methods by a significant margin.
[1] Yuan X, Wang T, Gulcehre C, et al. Machine comprehension by text-to-text neural question generation[J]. arXiv preprint arXiv:1705.02012, 2017. [2] Heilman M, Smith N A. Good question! statistical ranking for question generation[C]//Proceedings of the Human Language Technologies: the Annual Conference of the North American Chapter of the Association for Computational Linguistics, 2010: 609-617. [3] Pan B, Li H, Yao Z, et al. Reinforced dynamic reasoning for conversational question generation[J]. arXiv preprint arXiv:1907.12667, 2019. [4] Du X, Cardie C. Identifying where to focus in reading comprehension for neural question generation[C]//Proceedings of the Conference on Empirical Methods in Natural Language Processing, 2017: 2067-2073. [5] Rajpurkar P, Zhang J, Lopyrev K, et al. Squad: 100,000+ questions for machine comprehension of text[J]. arXiv preprint arXiv:1606.05250, 2016. [6] Pan L, Xie Y, Feng Y, et al. Semantic graphs for generating deep questions[J]. arXiv preprint arXiv:2004.12704, 2020. [7] Xie Y, Pan L, Wang D, et al. Exploring question-specific rewards for generating deep questions[J]. arXiv preprint arXiv:2011.01102, 2020. [8] Yang Z, Qi P, Zhang S, et al. Hotpotqa: a dataset for diverse, explainable multi-hop question answering[J]. arXiv preprint arXiv:1809.09600, 2018. [9] Zhou Q, Yang N, Wei F, et al. Neural question generation from text: a preliminary study[C]//Proceedings of the National CCF Conference on Natural Language Processing and Chinese Computing. Springer, Cham, 2017: 662-671. [10] Song L, Wang Z, Hamza W, et al. Leveraging context information for natural question generation[C]//Proceedings of the Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2018: 569-574. [11] Tuan L A, Shah D, Barzilay R. Capturing greater context for question generation[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34(05): 9065-9072. [12] Chen Y, Wu L, Zaki M J. Reinforcement learning based graph-to-sequence model for natural question generation[J]. arXiv preprint arXiv:1908.04942, 2019. [13] Li W, Xiao X, Liu J, et al. Leveraging graph to improve abstractive multi-document summarization[J]. arXiv preprint arXiv:2005.10043, 2020. [14] Wang D, Liu P, Zheng Y, et al. Heterogeneous graph neural networks for extractive document summarization[J]. arXiv preprint arXiv:2004.12393, 2020. [15] Qiu L, Xiao Y, Qu Y, et al. Dynamically fused graph network for multi-hop reasoning[C]//Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 2019: 6140-6150. [16] Su D, Xu Y, Dai W, et al. Multi-hop Question Generation with Graph Convolutional Network[J]. arXiv preprint arXiv:2010.09240, 2020. [17] Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems, 2017: 5998-6008. [18] Devlin J, Chang M W, Lee K, et al. BERT: Pre-training of deep bidirectional transformers for language understanding[J]. arXiv preprint arXiv:1810.04805, 2018. [19] Dong L, Yang N, Wang W, et al. Unified language model pre-training for natural language understanding and generation[J]. arXiv preprint arXiv:1905.03197, 2019. [20] Velic^kovic′ P, Cucurull G, Casanova A, et al. Graph attention networks[J]. arXiv preprint arXiv:1710.10903, 2017. [21] Sutton R S, McAllester D A, Singh S P, et al. Policy gradient methods for reinforcement learning with function approximation[C]//Proceedings of the 12th Internation Conference on Neural Information Processing Systems, 2000: 1057-1063. [22] Bengio Y, Léonard N, Courville A. Estimating or propagating gradients through stochastic neurons for conditional computation[J]. arXiv preprint arXiv:1308.3432, 2013. [23] Kim Y, Lee H, Shin J, et al. Improving neural question generation using answer separation[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2019, 33(01): 6602-6609. [24] Zhao Y, Ni X, Ding Y, et al. Paragraph-level neural question generation with maxout pointer and gated self-attention networks[C]//Proceedings of the Conference on Empirical Methods in Natural Language Processing, 2018: 3901-3910. [25] Liu B, Zhao M, Niu D, et al. Learning to generate questions by learning what not to generate[C]//Proceedings of the World Wide Web Conference, 2019: 1106-1118.
2022, Vol. 36 
