Jan Hoidekr


2006

pdf bib
Exploiting Linguistic Knowledge in Language Modeling of Czech Spontaneous Speech
Pavel Ircing | Jan Hoidekr | Josef Psutka
Proceedings of the Fifth International Conference on Language Resources and Evaluation (LREC’06)

In our paper, we present a method for incorporating available linguistic information into a statistical language model that is used in ASR system for transcribing spontaneous speech. We employ the class-based language model paradigm and use the morphological tags as the basis for world-to-class mapping. Since the number of different tags is at least by one order of magnitude lower than the number of words even in the tasks with moderately-sized vocabularies, the tag-based model can be rather robustly estimated using even the relatively small text corpora. Unfortunately, this robustness goes hand in hand with restricted predictive ability of the class-based model. Hence we apply the two-pass recognition strategy, where the first pass is performed with the standard word-based n-gram and the resulting lattices are rescored in the second pass using the aforementioned class-based model. Using this decoding scenario, we have managed to moderately improve the word error rate in the performed ASR experiments.

pdf bib
Benefit of a Class-based Language Model for Real-time Closed-captioning of TV Ice-hockey Commentaries
Jan Hoidekr | J.V. Psutka | Aleš Pražák | Josef Psutka
Proceedings of the Fifth International Conference on Language Resources and Evaluation (LREC’06)

This article describes the real-time speech recognition system for closed-captioning of TV ice-hockey commentaries. Automatic transcription of TV commentary accompanying an ice-hockey match is usually a hard task due to the spontaneous speech of a commentator put often into a very loud background noise created by the public, music, siren, drums, whistle, etc. Data for building this system was collected from 41 matches that were played during World Championships in years 2000, 2001, and 2002 and were transmitted by the Czech TV channels. The real-time closed-captioning system is based on the class-based language model designed after careful analysis of training data and OOV words in new (till now unseen) commentaries with the goal to decrease an OOV (Out-Of-Vocabulary) rate and increase recognition accuracy.