Description: The general aim of the joint project in terms of scientific research is to advance prosody extraction, processing and modeling by involving a large portfolio of European languages (English, French, German, Hungarian, Italian, Serbian, Macedonian) where beside Western languages, Central- and Eastern European languages are also represented. Detection of salient prosodic events and investigation of crosslingual transfer is also of interest. Salient events are especially frequent in conversational / spontaneous speech, the latter being an active research topic nowadays. This aim is similar to that of the SIWIS project, however, within the present context, experiments are extended to several new languages, like Slavic Serbian and Macedonian (Indo-European family) and FinnoUgrian Hungarian (the most spoken European language of non Indo-European origin). The goal is to advance conversational modeling, prosody modification techniques and crosslingual transfer of prosody. Another aspect assessed is prosody — and especially stress – prediction based on text. Although numerous approaches exist for this purpose, there is a lack of methods that would assign stress patterns to textual input obtained from human speech in a prosody transplantation way. For this task, it is needed to create new databases, as well as integrate available methods and advanced models of the stress patterns of human speech.

Contact person: Prof. dr Vlado Delić
Period of realization: 2014 – 2016