Question Answering as Global Reasoning over Semantic Abstractions

Abstract

We propose a novel method for exploiting the semantic structure of text to answer multiple-choice questions. The approach is especially suitable for domains that require reasoning over a diverse set of linguistic constructs but have limited training data. To address these challenges, we present the first system, to the best of our knowledge, that reasons over a wide range of semantic abstractions of the text, which are derived using off-the-shelf, general-purpose, pre-trained natural language modules such as semantic role labelers, coreference resolvers, and dependency parsers. Representing multiple abstractions as a family of graphs, we translate question answering (QA) into a search for an optimal subgraph that satisfies certain global and local properties. This formulation generalizes several prior structured QA systems. Our system, SEMANTICILP, demonstrates strong performance on two domains simultaneously. In particular, on a collection of challenging science QA datasets, it outperforms various state-of-the-art approaches, including neural models, broad coverage information retrieval, and specialized techniques using structured knowledge bases, by 2%-6%.

arXiv commentAppeared in AAAI'18

Cited by in corpus (19)

• A Survey on Explainability in Machine Reading Comprehension
• Improving Question Answering with External Knowledge
• From 'F' to 'A' on the N.Y. Regents Science Exams: An Overview of the Aristo Project
• Bridging Knowledge Graphs to Generate Scene Graphs
• Discourse-Aware Semantic Self-Attention for Narrative Reading Comprehension
• On the Possibilities and Limitations of Multi-hop Reasoning Under Linguistic Imperfections
• Joint Constrained Learning for Event-Event Relation Extraction
• Case-Based Abductive Natural Language Inference
• Temporal Reasoning on Implicit Events from Distant Supervision
• Encoding Explanatory Knowledge for Zero-shot Science Question Answering
• Keyword-Attentive Deep Semantic Matching
• Knowledge Based Machine Reading Comprehension
• Robust Question Answering Through Sub-part Alignment
• Semantic Role Labeling as Syntactic Dependency Parsing
• Playing by the Book: An Interactive Game Approach for Action Graph Extraction from Text
• QuASE: Question-Answer Driven Sentence Encoding
• Attentive Semantic Role Labeling with Boundary Indicator
• Challenge Closed-book Science Exam: A Meta-learning Based Question Answering System
• Multi-Relational Question Answering from Narratives: Machine Reading and Reasoning in Simulated Worlds