More

More on my research

Regarding speech technology, I have been active particularly within text-to-speech-conversion development. Two Swedish MBROLA diphone databases have been developed at the department (I supervised one of them), and NLP components for text processing, pronunciation and prosody have been implemented using the Festival (http://www.festvox.org) framework. A demo is available (http://person2.sol.lu.se/JohanFrid/festival/festival.php), but the current state of the system is more a test implementation than a practical product and many aspects of the system could be improved. Other synthesis-related demos are available as well: Letter-to-Sound rules for Swedish (http://person2.sol.lu.se/JohanFrid/festival/ltsr.php) and an MBROLA interface (http://person2.sol.lu.se/JohanFrid/festival/mbrola.php).