Interactive Machine Comprehension with Dynamic Knowledge Graphs

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 6734–6750 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 6734 Interactive Machine Comprehension with Dynamic Knowledge Graphs Xingdi Yuan Microsoft Research, Montréal eric.yuan@microsoft.com Abstract Interactive machine reading comprehension (iMRC) is machine comprehension tasks where knowledge sources are partially observ- able. An agent must interact with an environ- ment sequentially to gather necessary knowl- edge in order to answer a question. We hy- pothesize that graph representations are good inductive biases, which can serve as an agent’s memory mechanism in iMRC tasks. We ex- plore four different categories of graphs that can capture text information at various levels. We describe methods that dynamically build and update these graphs during information gathering, as well as neural models to encode graph representations in RL agents. Extensive experiments on iSQuAD suggest that graph representations can result in significant perfor- mance improvements for RL agents.1 1 Introduction Machine reading comprehension (MRC) has gath- ered wide interest from the NLP community in recent years. It serves as a way to benchmark a system’s ability to understand and reason over nat- ural language. Typically, given a knowledge source such as a document, a model is required to read through the knowledge source to answer a question about some information contained therein. In the extractive QA paradigm, in particular, answers are typically sub-strings of the knowledge source (Ra- jpurkar et al., 2016; Trischler et al., 2016a; Yang et al., 2018a). Models are thus required to select a span from the knowledge source as their prediction. A recent line of work known as interactive ma- chine reading comprehension (iMRC) features in- teractive language learning and knowledge acqui- sition (Yuan et al., 2020; Ferguson et al., 2020). It shifts the focus of MRC research towards a more realistic setting where the knowledge sources (en- vironments) are partially observable. Under this 1We release code and data at https://github.com/ xingdi-eric-yuan/imrc_graph_public setting, agents must iteratively interact with the environment to discover necessary information in order to answer the questions. The sequence of interactions between an agent and the environment may resemble the agent’s reasoning path, rendering a higher level of interpretability in the agent’s be- haviour. The trajectories of interactions can also be seen as procedural knowledge, which potentially brings agents extra generalizability (humans do not necessarily know the answer to a question imme- diately, but they know the procedure to search it). Compared to many static MRC datasets, where the entire knowledge source (e.g., a paragraph in SQuAD) is presented to the model immediately, the iMRC setting may alleviate the risk of learning shallow pattern matching (Sugawara et al., 2018; Sen and Saffari, 2020). On a parallel track, there have been a plethora of studies that leverage graphs in MRC. Multiple linguistic features have been explored to help con- struct graphs, such as coreference (Dhingra et al., 2018; Song et al., 2018), entities (De Cao et al., 2019; Qiu et al., 2019; Tu et al., 2020) and se- mantic …

Content truncated to 500 words