Semantic sentence similarity: size does not always matter

TL;DR

This paper demonstrates that visually grounded speech recognition models can effectively learn sentence semantics with smaller datasets, emphasizing the importance of data quality and paraphrasing over sheer size.

Contribution

It shows that smaller, well-structured datasets with paraphrasing can outperform larger datasets in training semantic speech models.

Findings

Smaller datasets can produce embeddings correlating well with human judgments.

Multiple captions per image improve semantic learning even with fewer images.

Dataset quality and paraphrasing are crucial, not just size.

Abstract

This study addresses the question whether visually grounded speech recognition (VGS) models learn to capture sentence semantics without access to any prior linguistic knowledge. We produce synthetic and natural spoken versions of a well known semantic textual similarity database and show that our VGS model produces embeddings that correlate well with human semantic similarity judgements. Our results show that a model trained on a small image-caption database outperforms two models trained on much larger databases, indicating that database size is not all that matters. We also investigate the importance of having multiple captions per image and find that this is indeed helpful even if the total number of images is lower, suggesting that paraphrasing is a valuable learning signal. While the general trend in the field is to create ever larger datasets to train models on, our findings indicate other characteristics of the database can just as important important.