Constructing Datasets for Multi-hop Reading Comprehension Across Documents

TL;DR

This paper introduces a new multi-hop reading comprehension task across multiple documents, along with datasets and evaluation of models, highlighting current limitations and potential for future improvements.

Contribution

It presents a novel dataset creation methodology for multi-document comprehension and evaluates existing models, revealing their challenges in multi-hop inference.

Findings

Models can integrate information across documents.

Models struggle to select relevant evidence.

Performance gap remains between models and humans.

Abstract

Most Reading Comprehension methods limit themselves to queries which can be answered using a single sentence, paragraph, or document. Enabling models to combine disjoint pieces of textual evidence would extend the scope of machine comprehension methods, but currently there exist no resources to train and test this capability. We propose a novel task to encourage the development of models for text understanding across multiple documents and to investigate the limits of existing methods. In our task, a model learns to seek and combine evidence - effectively performing multi-hop (alias multi-step) inference. We devise a methodology to produce datasets for this task, given a collection of query-answer pairs and thematically linked documents. Two datasets from different domains are induced, and we identify potential pitfalls and devise circumvention strategies. We evaluate two previously proposed competitive models and find that one can integrate information across documents. However, both models struggle to select relevant information, as providing documents guaranteed to be relevant greatly improves their performance. While the models outperform several strong baselines, their best accuracy reaches 42.9% compared to human performance at 74.0% - leaving ample room for improvement.