Query-by-Example Spoken Term Detection: A Systematic Review
Manisha Naik Gaonkar, Veena Thenakanidiyoor, A. D. Dileep · International Journal of Electronics and Communication Engineering · 2025
Massive collections of multimedia data have been created as a result of the Internet’s recent exponential expansion. Effective management of these repositories necessitates matching spoken queries with the audio content of these videos. One powerful technique that has emerged to address this issue is QbE-STD. Nevertheless, QbE-STD encounters many difficulties, including age sensitivity, dialect differences, and computational complexity. The objective of this study is to address these issues. In this area, review publications are scarce and mostly lack an in-depth study of feature representations and matching techniques. This paper presents a detailed analysis of different approaches and developments in QbE-STD. It covers feature representations, similarity metrics, matching methods, datasets, evaluation measures, and benchmarking platforms. The paper delves into the intricacies of various feature representations and scrutinizes similarity metrics. These metrics are analyzed for their advantages and disadvantages in computing a matching matrix between a query and an utterance. Furthermore, the paper highlights how machine learning and deep learning architectures are increasingly integrated into QbE-STD. Finally, the paper discusses a few challenges associated with QbE-STD, which provide an opportunity for future research in this field.