Dr. Lucian Galescu received his PhD in Computer Science from University of Rochester in 2003. The same year he joined the Florida Institute for Human and Machine Cognition (IHMC) as a Research Scientist. His research expertise is primarily in the areas of statistical natural language processing and dialogue systems.
Selected areas of past and current research:
- language modeling for speech recognition, especially for dialogue systems:
- techniques for rapid development of language models for new application domains, which have led to dramatic decreases in the turnaround time for dialogue systems in new application domains, as well as increases in their portability and adaptability.
- using non-linguistic cues for language model adaptation to increase speech recognition accuracy.
- pronunciation modeling:
- models for the recognition and pronunciation of new words and proper names; his work on grapheme-to-phoneme conversion has resulted in models considered among the best in existence today.
- dialogue systems for various domains:
- human-robot communication and teamwork (funded by ONR and AFRL);
- task learning, planning and execution (funded by DARPA);
- interaction with information systems (funded by NASIC);
- smart homes (funded by NSF);
- providing and eliciting information to/from patients (funded by NIH and Robert Wood Johnson Foundation).
- information extraction:
- using deep natural language processing techniques and complex ontologies for extracting knowledge from clinical documents (funded by NIH, via the Moffitt Cancer Center).
- metaphor detection:
- using lexical ontologies and shallow semantic parsing to detect metaphors (funded by IARPA).
- word sense induction and disambiguation
- using bicorpora to induce word senses and building models for disambiguation and clarification to improve speech-to-speech machine translation (funded by DARPA).
Dr. Galescu is co-author of two patents and has a number of publications in international journals and conferences as well as several book chapters.