Sentence Meta-Embeddings for Unsupervised Semantic Textual Similarity

TL;DR

This paper introduces a method to improve unsupervised semantic textual similarity by combining multiple pre-trained sentence encoders into meta-embeddings, achieving state-of-the-art results.

Contribution

It extends and evaluates meta-embedding techniques from word embeddings to sentence embeddings, setting new unsupervised benchmarks for STS tasks.

Findings

Achieved new unsupervised state-of-the-art on STS Benchmark.

Meta-embeddings outperform single-source systems by 3.7% to 6.4% in Pearson's r.

Demonstrated effectiveness of ensemble methods in semantic similarity tasks.

Abstract

We address the task of unsupervised Semantic Textual Similarity (STS) by ensembling diverse pre-trained sentence encoders into sentence meta-embeddings. We apply, extend and evaluate different meta-embedding methods from the word embedding literature at the sentence level, including dimensionality reduction (Yin and Sch\"utze, 2016), generalized Canonical Correlation Analysis (Rastogi et al., 2015) and cross-view auto-encoders (Bollegala and Bao, 2018). Our sentence meta-embeddings set a new unsupervised State of The Art (SoTA) on the STS Benchmark and on the STS12-STS16 datasets, with gains of between 3.7% and 6.4% Pearson's r over single-source systems.