Natural Languange | Department of Intelligent Systems

Natural Languange

In the field of speech and language technologies we work on speech synthesis, forensic speaker recognition, semantic analysis of text and question answering.

Together with the Amebis company, we developed a new speech synthesizer for Slovene. In cooperation with the national television and radio, RTV Slovenia, we developed a phonetically rich and balanced speech database for corpus-based speech synthesis.

In speaker recognition we investigate the correlation between the speech quality in telephony and the performance of automatic speaker verification.

In question answering, a machine-learning classifier that classifies questions by the expected answer type was developed. We developed a new version of the virtual assistant, which uses state-of-the-art web technology to improve the user experience. Robi, the Institute’s virtual assistant, received some handy new functions, for example the ability to search the employees’ contact data. We deployed the Little Dragon, who answers questions of the visitors to the Municipality of Ljubljana website. We also developed an assistant for the use in an intelligent home.