Written Term Detection Improves Spoken Term Detection

TL;DR

This paper introduces a multitask training method for end-to-end spoken term detection that leverages unpaired text to improve search accuracy and facilitate domain adaptation, addressing limitations of traditional ASR-based systems.

Contribution

It proposes a novel multitask training objective that incorporates unpaired text into end-to-end KWS, enhancing performance without increasing indexing complexity.

Findings

Significant improvements in search performance across multiple languages.

Enhanced document representations for words in unpaired text.

Effective domain adaptation with scarce in-domain data.

Abstract

End-to-end (E2E) approaches to keyword search (KWS) are considerably simpler in terms of training and indexing complexity when compared to approaches which use the output of automatic speech recognition (ASR) systems. This simplification however has drawbacks due to the loss of modularity. In particular, where ASR-based KWS systems can benefit from external unpaired text via a language model, current formulations of E2E KWS systems have no such mechanism. Therefore, in this paper, we propose a multitask training objective which allows unpaired text to be integrated into E2E KWS without complicating indexing and search. In addition to training an E2E KWS model to retrieve text queries from spoken documents, we jointly train it to retrieve text queries from masked written documents. We show empirically that this approach can effectively leverage unpaired text for KWS, with significant improvements in search performance across a wide variety of languages. We conduct analysis which indicates that these improvements are achieved because the proposed method improves document representations for words in the unpaired text. Finally, we show that the proposed method can be used for domain adaptation in settings where in-domain paired data is scarce or nonexistent.