Ditto: A Simple and Efficient Approach to Improve Sentence Embeddings

Qian Chen, Wen Wang, Qinglin Zhang, Siqi Zheng, Chong Deng, Hai Yu, Jiaqing Liu, Yukun Ma, Chong Zhang

Submission Type: Regular Short Paper

Submission Track: Semantics: Lexical, Sentence level, Document Level, Textual Inference, etc.

Keywords: Sentence embeddings, BERT, Transformer, Self-attention

TL;DR: Introducing Ditto: a learning-free approach that uses model-based importance estimations to weight words and compute sentence embeddings from pre-trained model representations.

Abstract: Prior studies diagnose the anisotropy problem in sentence representations from pre-trained language models, e.g., BERT, without fine-tuning. Our analysis reveals that the sentence embeddings from BERT suffer from a bias towards uninformative words, limiting the performance in semantic textual similarity (STS) tasks. To address this bias, we propose a simple and efficient unsupervised approach, Diagonal Attention Pooling (Ditto), which weights words with model-based importance estimations and computes the weighted average of word representations from pre-trained models as sentence embeddings. Ditto can be easily applied to any pre-trained language model as a postprocessing operation. Compared to prior sentence embedding approaches, Ditto does not add parameters nor requires any learning. Empirical evaluations demonstrate that our proposed Ditto can alleviate the anisotropy problem and improve various pre-trained models on the STS benchmarks.

Submission Number: 1619