Densely Connected Attention Propagation for Reading Comprehension
This addresses reading comprehension for NLP applications, representing a strong incremental advance with novel architectural components.
The paper tackles reading comprehension by proposing DecaProp, a densely connected neural architecture with attention-based connectors, achieving state-of-the-art results on four benchmarks with improvements of up to 2.6%-14.2% in F1 score.
We propose DecaProp (Densely Connected Attention Propagation), a new densely connected neural architecture for reading comprehension (RC). There are two distinct characteristics of our model. Firstly, our model densely connects all pairwise layers of the network, modeling relationships between passage and query across all hierarchical levels. Secondly, the dense connectors in our network are learned via attention instead of standard residual skip-connectors. To this end, we propose novel Bidirectional Attention Connectors (BAC) for efficiently forging connections throughout the network. We conduct extensive experiments on four challenging RC benchmarks. Our proposed approach achieves state-of-the-art results on all four, outperforming existing baselines by up to $2.6\%-14.2\%$ in absolute F1 score.