Semantic query-by-example speech search using visual grounding
This addresses the challenge of improving speech search systems for users needing semantic relevance, though it is incremental as it builds on existing visually grounded training methods.
The paper tackled the problem of semantic query-by-example speech search, where the goal is to retrieve utterances relevant to a spoken query beyond exact matches, and showed that using an embedding function trained on visually grounded speech data outperforms a purely acoustic system in both exact and semantic retrieval performance.
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.