Sentence Embeddings in NLI with Iterative Refinement Encoders

TL;DR

This paper introduces an iterative refinement encoder architecture using BiLSTM and max pooling for sentence embeddings, achieving state-of-the-art results in NLI and strong transfer learning performance.

Contribution

It proposes a hierarchical BiLSTM and max pooling model with iterative refinement, improving sentence embedding quality for NLP tasks.

Findings

State-of-the-art results on SciTail dataset

Outperforms InferSent on 7/10 transfer tasks

Outperforms InferSent on 8/10 linguistic probing tasks

Abstract

Sentence-level representations are necessary for various NLP tasks. Recurrent neural networks have proven to be very effective in learning distributed representations and can be trained efficiently on natural language inference tasks. We build on top of one such model and propose a hierarchy of BiLSTM and max pooling layers that implements an iterative refinement strategy and yields state of the art results on the SciTail dataset as well as strong results for SNLI and MultiNLI. We can show that the sentence embeddings learned in this way can be utilized in a wide variety of transfer learning tasks, outperforming InferSent on 7 out of 10 and SkipThought on 8 out of 9 SentEval sentence embedding evaluation tasks. Furthermore, our model beats the InferSent model in 8 out of 10 recently published SentEval probing tasks designed to evaluate sentence embeddings' ability to capture some of the important linguistic properties of sentences.