Myr
My thesis research deals with representing lexical semantic properties of words using semantic features and learning semantic features(properties) of words with statistical methods. I recently defended a thesis proposal, proposing to develop a set of algorithms to learn and utilize lexical semantic feature sets in parsing. Among other things, I'm planning to use a log-linear model in conjunction with expectation maximization to learn feature sets associated with words and dependencies within feature sets.
I'm actively involved in the Conversational Interaction and Spoken Dialogue research group , developing multimodal dialogue systems. Currently we are working on the TRIPS system , an interactive planing assistant capable of understanding spoken language. My most recent project is developing an algorithm to automatically adapt generic system lexicon to the application domains.
I spent the summer of 2001 at Mitsubishi Electric Research Laboratories working with Candace Sidner to get a small robot gesture naturally synchronized in speech. I connected it to COLLAGEN, a collaborative agent developed at MERL.
My first project was on robust parsing in the system, but we quickly discovered that the methods successfully used previously in the smaller TRAINS domain were not working properly in a larger domain, and the lexical information in the lexicon was insufficient to extend them appropriately. Since then I worked on the new design for TRIPS lexicon that would improve parsing, support robustness capability, and would make it easier to port the system to new domains. I was mainly concentrating on lexical semantics and ontology design. If you're currently working with an existing ontology, I'd love to talk to you!
I'm supervised a project developing a GUI to the TRIPS lexicon in Java. We created a flexible API for parsing and outputting the lexicon files, and a set of viewers that integrate information from different files in easy to view fashion, which I use actively to help me edit the lexicon. At present, we are developing the editor interface. You can look at the current documentation
September 1997- Department of Computer Science, University of Rochester. PhD in Computer Science (expected Fall 2003)
1992-1997 Ivan Franko Lviv State University, Applied Mathematics and Informatics department
June 1997 Diploma in applied mathematics
| Fall 1997- now | TRAINS research group, Department of Computer Science, University of Rochester |
| Robust understanding in spoken dialogue, lexicon and ontology development for TRIPS system. Domain adaptation in parsing. Statistical learning of semantic features for words. | |
| Summer 2000 | Mitsubishi Electric Research Labs, Boston, MA |
| Implementation of gestures synchronized with speech for a small robot attached to the COLLAGEN system. | |
| 1995-1997 | Research laboratory of new optics materials, Institute of Applied Physics of Ivan Franko Lviv State University. |
| Calculation of the parameters of scintillators. | |
| 1994-1995 | Chair of Programming, Ivan Franko Lviv State University, Lviv, Ukraine |
| Representations of trees and algorithms on them. |
My diploma says "applied math", but this is only because of the foolish Ukrainian government classification system, which doesn't include Computer Science, my actual major.
My department gave me good background in both mathematics and computer science, though it was more oriented to computational math and numerical methods. Thus, most of my undergraduate research was done in scientific computation as the only research field supported by the department. My initial interest in NLP stems from the general interest in foreign languages and the ways their grammars work. In addition to Ukrainian, I freely speak Russian and English, and also know some Polish and French. I hope that one day we will build the systems that will be able to understand all these languages.