Multi-task learning in under-resourced Dravidian languages

Eigentrigraphemes for under-resourced languages

Grapheme-based modeling has an advantage over phone-based modeling in automatic speech recognition for under-resourced languages when a good dictionary is not available. Recently we proposed a new method for parameter estimation of context-dependent hidden Markov model (HMM) called eigentriphone modeling. Eigentriphone modeling outperforms conventional tied-state HMM by eliminating the quantiza...

Joint Bayesian Morphology learning for Dravidian languages

In this paper a methodology for learning the complex agglutinative morphology of some Indian languages using Adaptor Grammars and morphology rules is presented. Adaptor grammars are a compositional Bayesian framework for grammatical inference, where we define a morphological grammar for agglutinative languages and morphological boundaries are inferred from a plain text corpus. Once morphologica...

Useful technique for low-resourced/under-resourced languages

Acoustic modelling for under-resourced languages

Over the past decades research in the field of automatic speech recognition has lead to systems with a sufficiently high grade of maturity that makes them suitable for use in real-life applications. However, such recognition systems have been developed only for very few languages. Languages addressed are mainly those with a large population, a high economic power, or for which a high political ...

Multi-Task Learning Using Mismatched Transcription for Under-Resourced Speech Recognition

It is challenging to obtain large amounts of native (matched) labels for audio in under-resourced languages. This could be due to a lack of literate speakers of the language or a lack of universally acknowledged orthography. One solution is to increase the amount of labeled data by using mismatched transcription, which employs transcribers who do not speak the language (in place of native speak...