New opening for a PhD position in machine learning for spoken term detection

The Idiap Research Institute seeks qualified candidates for one PhD position in machine learning for spoken term detection.

The position is funded by the Hasler Stiftung. The duration of the appointment is for a maximum of 4 years with a competitive PhD salary. The position is available from December 2011 onwards. Starting date is negotiable.


The majority of state-of-the-art speech recognition systems heavily rely on carefully "hand-crafted" features; usually generative models are then used to estimate likelihood of subword units (typically phonemes); dynamic programming methods are then applied to recognize the word sequence under various constraints, such as lexical constraints or language model constraints. This type of approach divided in several independent steps has shown great advantages, in term of accuracy performance or computational cost.

It raises however several questions: (1) Can we do better than existing "hand-crafted" features? In particular, can we learn the right features for our task? (2) A problem solved in several steps has good chances to be sub-optimal. Can we find ways to reduce the number of steps involved in speech processing systems? In this context, the student will investigate new end-to-end learning systems for speech processing, following recent work on deep learning techniques. Leveraging the outcomes of his or her research, he or she will design a practical spoken term detection (keyword spotting) system.




*** This position has been filled ***


More information about Idiap open position can be found on  the page Education & Jobs