This paper describes the systems developed by the GTM-IRLab team for the Albayzin 2018 Search on Speech evaluation. The system for the spoken term detection task consists in the fusion of two subsystems: a large vocabulary continuous speech recognition strategy that uses the proxy words approach for out-of-vocabulary terms, and a phonetic search system based on the probabilistic retrieval model for information retrieval. The query-by-example spoken term detection system is the result of fusing four subsystems: three of them are based on dynamic time warping search using different representations of the waveforms, namely Gaussian posteriorgrams, phoneme posteriorgrams and a large set of low-level descriptors; and the other one is the phonetic search system used for spoken term detection with some modifications to manage spoken queries.
Cite as: López Otero, P., Docío-Fernández, L. (2018) GTM-IRLab Systems for Albayzin 2018 Search on Speech Evaluation. Proc. IberSPEECH 2018, 240-244, doi: 10.21437/IberSPEECH.2018-50
@inproceedings{lopezotero18b_iberspeech,
author={Paula {López Otero} and Laura Docío-Fernández},
title={{GTM-IRLab Systems for Albayzin 2018 Search on Speech Evaluation}},
year=2018,
booktitle={Proc. IberSPEECH 2018},
pages={240--244},
doi={10.21437/IberSPEECH.2018-50}
}