Simple Question Answering with Subgraph Ranking and Joint-Scoring

Proceedings of NAACL-HLT 2019, pages 324–334 Minneapolis, Minnesota, June 2 - June 7, 2019. c⃝2019 Association for Computational Linguistics 324 Simple Question Answering with Subgraph Ranking and Joint-Scoring Wenbo Zhao∗ Carnegie Mellon University wzhao1@andrew.cmu.edu Tagyoung Chung, Anuj Goyal, and Angeliki Metallinou Amazon Alexa AI {tagyoung, anujgoya, ametalli}@amazon.com Abstract Knowledge graph based simple question an- swering (KBSQA) is a major area of research within question answering. Although only dealing with simple questions, i.e., questions that can be answered through a single knowl- edge base (KB) fact, this task is neither sim- ple nor close to being solved. Targeting on the two main steps, subgraph selection and fact selection, the research community has devel- oped sophisticated approaches. However, the importance of subgraph ranking and leverag- ing the subject–relation dependency of a KB fact have not been sufficiently explored. Moti- vated by this, we present a unified framework to describe and analyze existing approaches. Using this framework as a starting point, we focus on two aspects: improving subgraph se- lection through a novel ranking method and leveraging the subject–relation dependency by proposing a joint scoring CNN model with a novel loss function that enforces the well- order of scores. Our methods achieve a new state of the art (85.44% in accuracy) on the SimpleQuestions dataset. 1 Introduction Knowledge graph based simple question answer- ing (KBSQA) is an important area of research within question answering, which is one of the core areas of interest in natural language process- ing (Yao and Van Durme, 2014; Yih et al., 2015; Dong et al., 2015; Khashabi et al., 2016; Zhang et al., 2018; Hu et al., 2018). It can be used for many applications such as virtual home assis- tants, customer service, and chat-bots. A knowl- edge graph is a multi-entity and multi-relation di- rected graph containing the information needed to answer the questions. The graph can be repre- sented as collection of triples {(subject, relation, ∗Work conducted during an internship at Alexa AI, CA. object)}. Each triple is called a fact, where a di- rected relational arrow points from subject node to object node. A simple question means that the question can be answered by extracting a sin- gle fact from the knowledge graph, i.e., the ques- tion has a single subject and a single relation, hence a single answer. For example, the question “Which Harry Potter series did Rufus Scrimgeour appear in?” can be answered by a single fact (Ru- fus Scrimgeour, book.book-characters.appears-in- book, Harry Potter and the Deathly Hallows). Given the simplicity of the questions, one would think this task is trivial. Yet it is far from be- ing easy or close to being solved. The complex- ity lies in two aspects. One is the massive size of the knowledge graph, usually in the order of billions of facts. The other is the variability of the questions in natural language. Based on this anatomy of the problem, the solutions also con- sist of two steps: (1) selecting a relatively small subgraph from the knowledge …

Content truncated to 500 words