Unsupervised acoustic sub-word unit detection for query-by-example spoken term detection

Marijn Huijbregts, Mitchell McLaren, David A. van Leeuwen

2011 · 51 citations · 11 references

DOIFull text

Open access

Concepts

TL;DR

Unsupervised speaker diarization can generate speaker models from a recording without training data. This work proposes a method to automatically generate acoustic sub‑word units that replace conventional phone models in query‑by‑example spoken term detection. The method modifies a speaker diarization system to process single‑speaker speech and lowers the minimum segment duration, enabling detection of speaker‑dependent sub‑word units. The resulting system achieves mean average precisions of 0.28 on broadcast news and 0.38 on war veteran interviews, outperforming a phone‑based system trained only on broadcast data.

Abstract

In this paper we present a method for automatically generating acoustic sub-word units that can substitute conventional phone models in a query-by-example spoken term detection system. We generate the sub-word units with a modified version of our speaker diarization system. Given a speech recording, the original diarization system generates a set of speaker models in an unsupervised manner without the need for training or development data. Modifying the diarization system to process the speech of a single speaker and decreasing the minimum segment duration constraint allows us to detect speaker-dependent sub-word units. For the task of query-by-example spoken term detection, we show that the pro posed system performs well on both broadcast and non-broadcast recordings, unlike a conventional phone-based system trained solely on broadcast data. A mean average precision of 0.28 and 0.38 was obtained for experiments on broadcast news and on a set of war veteran interviews, respectively.

References

11