RikiNet: Reading Wikipedia Pages for Natural Question Answering

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 6762–6771 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 6762 RikiNet: Reading Wikipedia Pages for Natural Question Answering Dayiheng Liu♠† Yeyun Gong† Jie Fu♦ Yu Yan† Jiusheng Chen♥ Daxin Jiang‡ Jiancheng Lv♠ Nan Duan† ♠College of Computer Science, Sichuan University †Microsoft Research Asia ♦ Mila ‡Microsoft Search Technology Center Asia ♥Microsoft AI and Research losinuris@gmail.com Abstract Reading long documents to answer open- domain questions remains challenging in nat- ural language understanding. In this paper, we introduce a new model, called RikiNet, which reads Wikipedia pages for natural question an- swering. RikiNet contains a dynamic para- graph dual-attention reader and a multi-level cascaded answer predictor. The reader dynam- ically represents the document and question by utilizing a set of complementary attention mechanisms. The representations are then fed into the predictor to obtain the span of the short answer, the paragraph of the long answer, and the answer type in a cascaded manner. On the Natural Questions (NQ) dataset, a single RikiNet achieves 74.3 F1 and 57.9 F1 on long- answer and short-answer tasks. To our best knowledge, it is the first single model that out- performs the single human performance. Fur- thermore, an ensemble RikiNet obtains 76.1 F1 and 61.3 F1 on long-answer and short- answer tasks, achieving the best performance on the official NQ leaderboard1. 1 Introduction Machine reading comprehension (MRC) refers to the task of finding answers to given questions by reading and understanding some documents. It rep- resents a challenging benchmark task in natural language understanding (NLU). With the progress of large-scale pre-trained language models (Devlin et al., 2018), state-of-the-art MRC models (Ju et al., 2019; Yang et al., 2019; Lan et al., 2019; Zhang et al., 2019; Liu et al., 2019) have already surpassed human-level performance on certain commonly used MRC benchmark datasets, such as SQuAD 1.1 (Rajpurkar et al., 2016), SQuAD 2.0 (Rajpurkar et al., 2018), and CoQA (Reddy et al., 2019). 1Till our submission time, 29 Nov. 2019. We re- fer readers to https://ai.google.com/research/ NaturalQuestions/leaderboard for the latest re- sults. Recently, a new benchmark MRC dataset called Natural Questions2 (NQ) (Kwiatkowski et al., 2019) has presented a substantially greater chal- lenge for the existing MRC models. Specifically, there are two main challenges in NQ compared to the previous MRC datasets like SQuAD 2.0. Firstly, instead of providing one relatively short paragraph for each question-answer (QA) pair, NQ gives an entire Wikipedia page which is signifi- cantly longer compared to other datasets. Sec- ondly, NQ task not only requires the model to find an answer span (called short answer) to the question like previous MRC tasks but also asks the model to find a paragraph that contains the infor- mation required to answer the question (called long answer). In this paper, we focus on the NQ task and propose a new MRC model called RikiNet tai- lored to its associated challenges, which Reads the Wikipedia pages for natural question answering. For …

Content truncated to 500 words