Semantic similarity prediction is better than other semantic similarity measures

TL;DR

This paper demonstrates that directly predicting semantic similarity with a fine-tuned model yields more accurate and aligned results than traditional overlap or embedding-based measures.

Contribution

It introduces the STSScore approach, a fine-tuned model specifically designed for semantic similarity prediction, outperforming existing methods on benchmark tasks.

Findings

STSScore aligns better with semantic similarity expectations

Fine-tuned models outperform traditional overlap and embedding methods

Results show improved correlation with human judgments

Abstract

Semantic similarity between natural language texts is typically measured either by looking at the overlap between subsequences (e.g., BLEU) or by using embeddings (e.g., BERTScore, S-BERT). Within this paper, we argue that when we are only interested in measuring the semantic similarity, it is better to directly predict the similarity using a fine-tuned model for such a task. Using a fine-tuned model for the Semantic Textual Similarity Benchmark tasks (STS-B) from the GLUE benchmark, we define the STSScore approach and show that the resulting similarity is better aligned with our expectations on a robust semantic similarity measure than other approaches.