MICRON: Multigranular Interaction for Contextualizing RepresentatiON in Non-factoid Question Answering

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 5890–5895, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 5890 MICRON: Multigranular Interaction for Contextualizing RepresentatiON in Non-factoid Question Answering Hojae Han∗, Seungtaek Choi∗, Haeju Park, Seung-won Hwang Department of Computer Science, Yonsei University, Seoul, Korea {stovecat, hist0613, phj0225, seungwonh}@yonsei.ac.kr Abstract This paper studies the problem of non-factoid question answering, where the answer may span over multiple sentences. Existing solu- tions can be categorized into representation- and interaction-focused approaches. We com- bine their complementary strength, by a hy- brid approach allowing multi-granular interac- tions, but represented at word level, enabling an easy integration with strong word-level sig- nals. Specifically, we propose MICRON: Multigranular Interaction for Contextualiz- ing RepresentatiON, a novel approach which derives contextualized uni-gram representation from n-grams. Our contributions are as fol- lows: First, we enable multi-granular matches between question and answer n-grams. Sec- ond, by contextualizing word representation with surrounding n-grams, MICRON can natu- rally utilize word-based signals for query term weighting, known to be effective in information retrieval. We validate MICRON in two pub- lic non-factoid question answering datasets: WikiPassageQA and InsuranceQA, showing our model achieves the state of the art among baselines with reported performances on both datasets. 1 Introduction Non-factoid questions, unlike factoid questions an- swered by short facts like a word or a phrase, may get answered by a long answer spanning across multiple sentences. Following the definition in (Guo et al., 2019), neural approaches for this task can be roughly categorized into representation- and interaction-focused approaches. First, representation-focused approaches (Rücklé and Gurevych, 2017; Shao et al., 2019) encode query and answer into vectors of the same size, and match the two by computing vector simi- larity. Models in this category have advantages of ∗The authors contribute equally to this paper. efficiency, as representations can be pre-computed and indexed for efficient retrieval. However, structural information, such as some question word matching another in answer, is missing in this representation. In addition, structural information can be diluted when squeezing a long text into a single vector. This weakness is often complemented by auxiliary information such as attention (Tan et al., 2016; Santos et al., 2016; Wang and Jiang, 2016). Figure 1a illustrates a representative architecture in this category (Shao et al., 2019). Second, interaction-focused approaches aim to preserve structural information above. A naive structural information is a matrix storing pairwise word interaction, or 1:1. However, due to a typical length difference between a question and a long answer in our problem setting, most answer words are left unmatched, except a few uni-gram in the answer. Later work relaxes 1:1 constraint, to 1:N and M:N, by allowing a match to n-gram (1:N) or a match between query and answer bi-grams (2:2). A state-of-the-art in this category (Rücklé et al., 2019), shown in Figure 1b, uses bi-gram Convolutional Neural Network (CNN) to represent query/answer bi-grams and their interactions. Sim- ilar architecture …

Content truncated to 500 words