Neural Module Networks for Reasoning over Text

Abstract

Answering compositional questions that require multiple steps of reasoning against text is challenging, especially when they involve discrete, symbolic operations. Neural module networks (NMNs) learn to parse such questions as executable programs composed of learnable modules, performing well on synthetic visual QA domains. However, we find that it is challenging to learn these models for non-synthetic questions on open-domain text, where a model needs to deal with the diversity of natural language and perform a broader range of reasoning. We extend NMNs by: (a) introducing modules that reason over a paragraph of text, performing symbolic reasoning (such as arithmetic, sorting, counting) over numbers and dates in a probabilistic and differentiable manner; and (b) proposing an unsupervised auxiliary loss to help extract arguments associated with the events in text. Additionally, we show that a limited amount of heuristically-obtained question program and intermediate module output supervision provides sufficient inductive bias for accurate learning. Our proposed model significantly outperforms state-of-the-art models on a subset of the DROP dataset that poses a variety of reasoning challenges that are covered by our modules.

Published in ICLR 2020 (International Conference on Learning Representations, 2020)

References in corpus (7)

• Learning to Map Sentences to Logical Form: Structured Classification with Probabilistic Categorial Grammars
• HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering
• Learning to Reason: End-to-End Module Networks for Visual Question Answering
• DROP: A Reading Comprehension Benchmark Requiring Discrete Reasoning Over Paragraphs
• Learning a Natural Language Interface with Neural Programmer
• Giving BERT a Calculator: Finding Operations and Arguments with Reading Comprehension
• A Multi-Type Multi-Span Network for Reading Comprehension that Requires Discrete Reasoning

Cited by in corpus (22)

• Post-hoc Interpretability for Neural NLP: A Survey
• A Survey on Explainability in Machine Reading Comprehension
• Zero-Shot Learning with Common Sense Knowledge Graphs
• Multi-Step Inference for Reasoning Over Paragraphs
• Toward Code Generation: A Survey and Lessons from Semantic Parsing
• Weakly Supervised Neuro-Symbolic Module Networks for Numerical Reasoning
• Dense Relational Image Captioning via Multi-task Triple-Stream Networks
• Temporal Reasoning on Implicit Events from Distant Supervision
• A Survey on Neural-symbolic Learning Systems
• Teaching Machine Comprehension with Compositional Explanations
• Question Directed Graph Attention Network for Numerical Reasoning over Text
• A Reinforcement Learning Environment for Mathematical Reasoning via Program Synthesis
• Program Enhanced Fact Verification with Verbalization and Graph Attention Network
• Mitigating False-Negative Contexts in Multi-document Question Answering with Retrieval Marginalization
• Web Question Answering with Neurosymbolic Program Synthesis
• VGNMN: Video-grounded Neural Module Network to Video-Grounded Language Tasks
• Improving Numerical Reasoning Skills in the Modular Approach for Complex Question Answering on Text
• EviDR: Evidence-Emphasized Discrete Reasoning for Reasoning Machine Reading Comprehension
• Teaching Autoregressive Language Models Complex Tasks By Demonstration
• Paired Examples as Indirect Supervision in Latent Decision Models
• Truth-Conditional Captioning of Time Series Data
• Exploring End-to-End Differentiable Natural Logic Modeling