New opening for an internship position on "Cross-lingual prosody modeling"
Speech to speech translation is able to translate speech from one language to another. Although the quality varies, we can expect to see the accuracy increase in the coming years. However, the speaking style and emotion in the speech is typically not translated. These aspects of speech are carried mainly by prosody, meaning features such as pitch, intensity and duration.
SP2 is a collaborative project to investigate how to extract, model and
synthesise prosody; it also encompasses prosody transfer across
languages, with the aim of enabling translation of speaking style and
emotion. Project partners are located in Budapest, Skopje and Novi Sad;
the target languages encompass those of the partners: Hungarian,
Macedonian and Serbian, along with English and the main Swiss languages.
The successful candidate will work with researchers at Idiap to detect, model and translate prosodic events. The work is likely to proceed by taking current models and applying them in the context of eastern European languages. However, the goals can be adjusted accordingly, especially if the candidate is already pursuing a doctoral programme elsewhere.
The position would suit a current PhD candidate, or someone wishing to pursue a PhD. Background knowledge in speech and natural language processing would be appreciated. Good command of spoken and written English is required, as are good computer skills. Knowledge of at least one Swiss language and/or another related to the target languages above would be a distinct advantage.
Start date and duration are subject to arrangement, but we expect from around autumn 2014 for around 6 months.