The AMI System for the Transcription of Speech in Meetings

Abstract

This paper describes the AMI transcription system for speech in meetings developed in collaboration by five research groups. The system includes generic techniques such as discriminative and speaker adaptive training, vocal tract length normalisation, heteroscedastic linear discriminant analysis, maximum likelihood linear regression, and phone posterior based features, as well as techniques specifically designed for meeting data. These include segmentation and cross-talk suppression, beam-forming, domain adaptation, web-data collection, and channel adaptive training. The system was improved by more than 20% relative in word error rate compared to our previous system and was usd in the NIST RT'06 evaluations where it was found to yield competitive performance.

Recognition and understanding of meetings the AMI and AMIDA projects

“…In the following we briefly address issues of the domain, followed by a brief description of the essential components of a meeting transcription system and the performance in recent NIST evaluations. For a more elaborate description of the systems, the interested reader is referred to [14].…”

Automatic analysis of multiparty meetings

“…Acoustic segmentation and speech/non-speech detection remains an important problem, with nearly 10% of errors in our current system resulting from errors in the speech/non-speech detection component. A feature of the systems developed for meeting recognition is the use of multiple recognition passes, cross-adaptation and model combination (Hain et al 2007). In particular successive passes make use of more detailed-and more diverse-acoustic and language models.…”

Recognition of Dialogue Acts in Multiparty Meetings Using a Switching DBN

“…Automatic transcriptions of the AMI meeting corpus were obtained using the AMI-ASR system [13]. This LVCSR system is based on decision tree clustered crossword triphone hidden Markov models, and a trigram language model.…”