Junior Researcher or Post-Doc for Acoustic Modeling of Under-Resourced Languages

Job description

It is well known that the dimensionality of feature vectors used in state-of-the-art speech recognition systems (typically in the range 30-40) is much larger than the intrinsic dimensionality of speech which is estimated to be 7-10 only. Efforts to make the intrinsic dimensionality smaller have been largely futile as the constraints are too complex for our by and large linear techniques. This inefficiency in basic representation is the main reason why speech recognition systems contain so many hundreds of thousands parameters that are largely redundant and why we need such large corpora to train these parameters. While very large corpora are available for the major languages, this is not the case for smaller languages, making them "under-resourced". This redundancy is also a major cause for lack of robustness in general.
The objective of this project is to apply novel mathematical techniques (e.g. spectral clustering) that can capture constraints - not in the feature space - but in the model space, i.e. in the underlying HMM parameters. Such constraints will lead to lesser requirements on the size of the training databases and should increase robustness in all situations where we don't have large corpora available, such as speaker adaptation, accent adaptation or modeling of under-resourced languages. In this project two test cases of under-resourced languages will be studied: i) "Afrikaans", for which data from Dutch and Flemish can be reused; ii) languages form the Bantu family as spoken in South Africa for which we can only bootstrap from a wide set of rather unrelated languages.

Project Partners

Qualifications

Candidates ideally have a university degree in engineering or computer science. Candidates with a general science degree and excellent programming skills may apply as well. Previous experience in speech recognition is not required but knowledge of or experience in any the following areas form an asset:

• speech recognition and speech modelling
• computational and programming skills (e.g. MATLAB, C, UNIX, Python)
• strong mathematical and statistical background
• some familiarity with Dutch or Afrikaans is a help, though not strictly required

Related Projects

AMODA: Acoustic Modelling for under-resources languages using feature space constraints

Terms of the position

The position is vacant at this moment and should be filled as soon as possible. We prefer to engage a junior researcher who would complete a PhD (4yr) in the course of the project. However, applications for a 2yr post-doc position from recently graduated PhDs will be accepted as well.

Applications

Interested applicants should send their CV to Prof. .