HeterMPC: A Heterogeneous Graph Neural Network for Response Generation in Multi-Party Conversations

Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 447 - 460 May 22-27, 2022 c⃝2022 Association for Computational Linguistics Fantastic Questions and Where to Find Them: FairytaleQA – An Authentic Dataset for Narrative Comprehension Ying Xu†∗ University of California Irvine Dakuo Wang† IBM Research Mo Yu† WeChat AI, Tencent Daniel Ritchie† University of California Irvine Bingsheng Yao† Rensselaer Polytechnic Institute Tongshuang Wu University of Washington Zheng Zhang University of Notre Dame Toby Jia-Jun Li University of Notre Dame Nora Bradford University of California Irvine Branda Sun University of California Irvine Tran Bao Hoang University of California Irvine Yisi Sang Syracuse University Yufang Hou IBM Research Ireland Xiaojuan Ma Hong Kong Univ. of Sci and Tech Diyi Yang Georgia Institute of Technology Nanyun Peng University of California Los Angeles Zhou Yu Columbia University Mark Warschauer University of California Irvine Abstract Question answering (QA) is a fundamental means to facilitate assessment and training of narrative comprehension skills for both ma- chines and young children, yet there is scarcity of high-quality QA datasets carefully designed to serve this purpose. In particular, existing datasets rarely distinguish fine-grained read- ing skills, such as the understanding of vary- ing narrative elements. Drawing on the read- ing education research, we introduce Fairy- taleQA1, a dataset focusing on narrative com- prehension of kindergarten to eighth-grade stu- dents. Generated by educational experts based on an evidence-based theoretical framework, FairytaleQA consists of 10,580 explicit and implicit questions derived from 278 children- friendly stories, covering seven types of narra- tive elements or relations. Our dataset is valu- able in two folds: First, we ran existing QA models on our dataset and confirmed that this annotation helps assess models’ fine-grained learning skills. Second, the dataset supports question generation (QG) task in the education domain. Through benchmarking with QG mod- els, we show that the QG model trained on FairytaleQA is capable of asking high-quality and more diverse questions. ∗†Equal contributions ying.xu@uci.edu, dakuo.wang@ibm.com, moyumyu@tencent.com, drritchi@uci.edu, yaob@rpi.edu. Work done while Mo was at IBM. ∗ Corresponding Author. 1Our dataset is available at https://github.com/ uci-soe/FairytaleQAData. Story Title: Magic Apples Story Text: [Sect 1] Once upon a time there was a lad who was better off than all the others. He was never short of money, for he had a purse which was never empty. ... ... [Sect 6] When the king’s daughter had eaten of the apples, she had a pair of horns. And then there was such a wailing in the castle that it was pitiful to hear. ... But one day a foreign doctor from afar came to court. He was not from their country, he said, and had made the journey purposely just to try his luck here. But he must see the king’s daughter alone, said he, and permission was granted him. ... [Sect 8] ... • Q1:Who will the foreign doctor turn out to be? •[explicit][prediction][sect 5, sect 6] A: The Lad. • Q2:How did the princess feel when she had …

Content truncated to 500 words