We investigate the benefit of augmenting with geo-location information the language model used in speech recognition for voice-search. We observe reductions in perplexity of up to 15% relative on test sets obtained from both typed query data, as well as transcribed voice search data; on a subset of the test data consisting of “local” queries — search results displaying a restaurant, some address, or similar — the reduction in perplexity is even higher, up to 30% relative. Automatic speech recognition experiments confirm the utility of geo-location information for improved language modeling. Significant reductions in word error rate are observed both on general voice search traffic, as well as “local” traffic, up to 2% and 8% relative, respectively.
Cite as: Chelba, C., Zhang, X., Hall, K. (2015) Geo-location for voice search language modeling. Proc. Interspeech 2015, 1438-1442, doi: 10.21437/Interspeech.2015-344
@inproceedings{chelba15_interspeech,
author={Ciprian Chelba and Xuedong Zhang and Keith Hall},
title={{Geo-location for voice search language modeling}},
year=2015,
booktitle={Proc. Interspeech 2015},
pages={1438--1442},
doi={10.21437/Interspeech.2015-344}
}