Semantic query-by-example speech search using visual grounding

TL;DR

This paper presents a semantic query-by-example speech search method that leverages visual grounding during training, improving retrieval of relevant utterances beyond exact matches.

Contribution

The study introduces a visually grounded speech embedding approach for semantic QbE, outperforming traditional acoustic methods in retrieval tasks.

Findings

Visually grounded embeddings improve semantic retrieval accuracy.

The method outperforms purely acoustic QbE systems.

Semantic retrieval includes relevant but non-exact utterances.

Abstract

A number of recent studies have started to investigate how speech systems can be trained on untranscribed speech by leveraging accompanying images at training time. Examples of tasks include keyword prediction and within- and across-mode retrieval. Here we consider how such models can be used for query-by-example (QbE) search, the task of retrieving utterances relevant to a given spoken query. We are particularly interested in semantic QbE, where the task is not only to retrieve utterances containing exact instances of the query, but also utterances whose meaning is relevant to the query. We follow a segmental QbE approach where variable-duration speech segments (queries, search utterances) are mapped to fixed-dimensional embedding vectors. We show that a QbE system using an embedding function trained on visually grounded speech data outperforms a purely acoustic QbE system in terms of both exact and semantic retrieval performance.