A disfluency study for cleaning spontaneous speech automatic transcripts and improving speech language models.
Martine Adda-DeckerBenoit HabertClaude BarrasGilles AddaPhilippe Boula de MareüilPatrick ParoubekPublished in: DiSS (2003)
Keyphrases
- spontaneous speech
- language model
- automatic speech recognition
- speech recognition
- human machine interaction
- spoken language
- word error rate
- spoken document retrieval
- probabilistic model
- language modeling
- test collection
- speech signal
- pattern recognition
- document retrieval
- information retrieval
- n gram
- linguistic features
- broadcast news
- retrieval model
- natural language
- machine learning
- spoken term detection