Cross-lingual Matryoshka Representation Learning across Speech and Text

TL;DR

This paper presents a bilingual speech-text embedding model for French-Wolof that enables efficient retrieval and generalizes to other tasks, with analysis of cost-accuracy trade-offs and modality fusion strategies.

Contribution

It introduces the first bilingual speech-text Matryoshka embedding model for under-represented languages, along with new benchmarks and data curation pipelines.

Findings

Modality fusion within a frozen text model performs best.

The model generalizes well to speech intent detection.

Information is concentrated in few components, enabling efficiency improvements.

Abstract

Speakers of under-represented languages face both a language barrier, as most online knowledge is in a few dominant languages, and a modality barrier, since information is largely text-based while many languages are primarily oral. We address this for French-Wolof by training the first bilingual speech-text Matryoshka embedding model, enabling efficient retrieval of French text from Wolof speech queries without relying on a costly ASR-translation pipelines. We introduce large-scale data curation pipelines and new benchmarks, compare modeling strategies, and show that modality fusion within a frozen text Matryoshka model performs best. Although trained only for retrieval, the model generalizes well to other tasks, such as speech intent detection, indicating the learning of general semantic representations. Finally, we analyze cost-accuracy trade-offs across Matryoshka dimensions and ranks, showing that information is concentrated only in a few components, suggesting potential for efficiency improvements.