Fraunhofer-Gesellschaft

Publica

Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Efficient subword lattice retrieval for German spoken term detection

 
: Mertens, T.; Schneider, D.

:

IEEE Signal Processing Society:
IEEE International Conference on Acoustics, Speech, and Signal Processing 2009. Proceedings. Vol.8 : April 19 - 24, 2009, Taipei International Convention Center, Taipei, Taiwan
Piscataway/NJ: IEEE, 2009 (2009 IEEE International Conference on Acoustics, Speech, and Signal Processing 8)
ISBN: 978-1-4244-2353-8
ISBN: 978-1-4244-2354-5
pp.4885-4888
International Conference on Acoustics, Speech, and Signal Processing (ICASSP) <34, 2009, Taipei>
English
Conference Paper
Fraunhofer IAIS ()
spoken term detection; spoken document retrieval; speech recognition; speech search

Abstract
We present a lattice-based STD method for German broadcast news data and compare it to a previously proposed fuzzy search. Due to the important out-of-vocabulary (OOV) problem in German, we evaluate suitable subword indexing units for lattice retrieval. Hybrid lattice retrieval of words and subwords is investigated because of the robust nature of words as an indexing unit. We show that by using efficient lattice graph and score pruning techniques, precision of subword retrieval is increased by 8% absolute with only a small loss in recall. Additionally, a speed-up of up to 6 times can be observed.

: http://publica.fraunhofer.de/documents/N-100732.html