Unsupervised Acoustic Model Training Research Articles

Lithuanian Broadcast Speech Transcription Using Semi-supervised Acoustic Model Training

This paper reports on an experimental work to build a speech transcription system for Lithuanian broadcast data, relying on unsupervised and semi-supervised training methods as well as on other low-knowledge methods to compensate for missing resources. Unsupervised acoustic model training is investigated using 360hours of untranscribed speech data. A graphemic pronunciation approach is used to simplify the pronunciation model generation and there-fore ease the language model adaptation for the system users. Discriminative training on top of semi-supervised training is also investigated, as well as various types of acoustic features and their combinations. Experimental results are provided for each of our development steps as well as contrastive results comparing various options. Using the best system configuration a word error rate of 18.3% is obtained on a set of development data from the Quaero program.

Enhancing ASR Systems for Under-Resourced Languages through a Novel Unsupervised Acoustic Model Training Technique

Statistical speech and language processing techniques, requiring large amounts of training data, are currently state-of-the-art in automatic speech recognition. For high-resourced, int ...

Lightly supervised and unsupervised acoustic model training

The last decade has witnessed substantial progress in speech recognition technology, with today’s state-of-the-art systems being able to transcribe unrestricted broadcast news audio data with a word error of about 20%. However, acoustic model development for these recognizers relies on the availability of large amounts of manually transcribed training data. Obtaining such data is both time-consuming and expensive, requiring trained human annotators and substantial amounts of supervision. This paper describes some recent experiments using lightly supervised and unsupervised techniques for acoustic model training in order to reduce the system development cost. The approach uses a speech recognizer to transcribe unannotated broadcast news data from the Darpa TDT-2 corpus. The hypothesized transcription is optionally aligned with closed-captions or transcripts to create labels for the training data. Experiments providing supervision only via the language model training materials show that including texts which are contemporaneous with the audio data is not crucial for success of the approach, and that the acoustic models can be initialized with as little as 10 min of manually annotated data. These experiments demonstrate that light or no supervision can dramatically reduce the cost of building acoustic models.