Hindi as a Second Language: Improving Visually Grounded Speech with Semantically Similar Samples

TL;DR

This paper proposes methods to enhance visually grounded speech models for low-resource languages by leveraging high-resource languages and semantically similar samples, improving cross-modal retrieval performance.

Contribution

It introduces two novel approaches—using a pre-trained high-resource encoder and semantically similar captions—to transfer knowledge from high-resource to low-resource languages in VGS models.

Findings

Low-resource language performance surpasses monolingual models.

Combining high-resource encoder and semantic sampling yields significant improvements.

Methods effectively transfer knowledge across languages in VGS tasks.

Abstract

The objective of this work is to explore the learning of visually grounded speech models (VGS) from multilingual perspective. Bilingual VGS models are generally trained with an equal number of spoken captions from both languages. However, in reality, there can be an imbalance among the languages for the available spoken captions. Our key contribution in this work is to leverage the power of a high-resource language in a bilingual visually grounded speech model to improve the performance of a low-resource language. We introduce two methods to distill the knowledge of high-resource language into low-resource languages: (1) incorporating a strong pre-trained high-resource language encoder and (2) using semantically similar spoken captions. Our experiments show that combining these two approaches effectively enables the low-resource language to surpass the performances of monolingual and bilingual counterparts for cross-modal retrieval tasks.