An Unsupervised Sentence Embedding Method by Mutual Information Maximization

BERT is inefficient for sentence-pair tasks such as clustering or semantic search as it needs to evaluate combinatorially many sentence pairs which is very time-consuming. Sentence BERT (SBERT) attempted to solve this challenge by learning semantically meaningful representations of single sentences, such that similarity comparison can be easily accessed. However, SBERT is trained on corpus with high-quality labeled sentence pairs, which limits its application to tasks where labeled data is extremely scarce. In this paper, we propose a lightweight extension on top of BERT and a novel self-supervised learning objective based on mutual information maximization strategies to derive meaningful sentence embeddings in an unsupervised manner. Unlike SBERT, our method is not restricted by the availability of labeled data, such that it can be applied on different domain-specific corpus. Experimental results show that the proposed method significantly outperforms other unsupervised sentence embedding baselines on common semantic textual similarity (STS) tasks and downstream supervised tasks. It also outperforms SBERT in a setting where in-domain labeled data is not available, and achieves performance competitive with supervised methods on various tasks.

PDF Abstract EMNLP 2020 PDF EMNLP 2020 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Semantic Textual Similarity SICK IS-BERT-NLI Spearman Correlation 0.6425 # 17
Semantic Textual Similarity STS12 IS-BERT-NLI Spearman Correlation 0.5677 # 15
Semantic Textual Similarity STS13 IS-BERT-NLI Spearman Correlation 0.6924 # 17
Semantic Textual Similarity STS14 IS-BERT-NLI Spearman Correlation 0.6121 # 15
Semantic Textual Similarity STS15 IS-BERT-NLI Spearman Correlation 0.7523 # 14
Semantic Textual Similarity STS16 IS-BERT-NLI Spearman Correlation 0.7016 # 14
Semantic Textual Similarity STS Benchmark IS-BERT-NLI Spearman Correlation 0.6921 # 30

Methods