FusionNet: Fusing via Fully-Aware Attention with Application to Machine Comprehension

TL;DR

FusionNet introduces a fully-aware multi-level attention mechanism that enhances neural models for machine comprehension, achieving state-of-the-art results on SQuAD and adversarial datasets by effectively capturing hierarchical attention information.

Contribution

It proposes a novel 'history of word' concept, an improved attention scoring function, and a fully-aware multi-level attention mechanism for better text understanding.

Findings

Achieved first position on SQuAD leaderboard for single and ensemble models.

Set new state-of-the-art results on adversarial SQuAD datasets AddSent and AddOneSent.

Significantly improved F1 scores on challenging datasets.

Abstract

This paper introduces a new neural structure called FusionNet, which extends existing attention approaches from three perspectives. First, it puts forward a novel concept of "history of word" to characterize attention information from the lowest word-level embedding up to the highest semantic-level representation. Second, it introduces an improved attention scoring function that better utilizes the "history of word" concept. Third, it proposes a fully-aware multi-level attention mechanism to capture the complete information in one text (such as a question) and exploit it in its counterpart (such as context or passage) layer by layer. We apply FusionNet to the Stanford Question Answering Dataset (SQuAD) and it achieves the first position for both single and ensemble model on the official SQuAD leaderboard at the time of writing (Oct. 4th, 2017). Meanwhile, we verify the generalization of FusionNet with two adversarial SQuAD datasets and it sets up the new state-of-the-art on both datasets: on AddSent, FusionNet increases the best F1 metric from 46.6% to 51.4%; on AddOneSent, FusionNet boosts the best F1 metric from 56.0% to 60.7%.