Joint discriminative language modeling and utterance classification
2005 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP '05, Philadelphia, PA, Amerika Birleşik Devletleri, 18 - 23 Mart 2005, cilt.I, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Cilt numarası: I
- Doi Numarası: 10.1109/icassp.2005.1415175
- Basıldığı Şehir: Philadelphia, PA
- Basıldığı Ülke: Amerika Birleşik Devletleri
- Boğaziçi Üniversitesi Adresli: Evet
Özet
This paper investigates discriminative language modeling in a scenario with two kinds of observed errors: errors in ASR transcription and errors in utterance classification. Using the perception algorithm, we train joint language and class models either independently or simultaneously, under various parameter update conditions. On a large vocabulary customer service call-classification application, we show that simultaneous optimization of class, n-gram, and class/n-gram feature weights results in a significant WER reduction over a model using just n-gram features, while additionally significantly outperforming a deployed baseline in classification error rate. A range of parameter update approaches for the various feature sets are presented and evaluated. The resulting models are encoded as weighted finite-state automata, and are used by intersecting the model with word lattices. © 2005 IEEE.