I am an Indiana University Computational linguistics/Cognitive science student specializing in natural language processing for robotics. I'm usually known as Rehj, both personally and professionally [so I may attempt to apply for a pseudonym soon]. I am a first-year graduate student and have published papers at HRI and RANLP, on the subjects of a speech-input-to-action-output natural language processing system, anaphora resolution, and incremental dependency parsing.
My work is mainly in natural language understanding for human-robot interaction, with an emphasis on dependency parsing. I have also done some personal projects that focused on autonomous physical actions, rather than natural language processing. I'm just now beginning to enter the arena of natural language generation.
My projects always emphasize the use of autonomous physical robots, rather than simulated ones, which many projects use in order to abstract away from the details of debugging and integration that the real world entails. However, sometimes I play with simulated robots for fun.
I have also done work in the past on improving the output of a speech recognizer, but it treated the speech recognizer as a black box, which I now believe was a serious mistake. While I am still interested in this area, my work in this area is currently restricted due to the lack of availability of appropriate and well-transcribed corpora (and lack of time to develop such corpora; I'm waiting till I'm a professor and have students available to work on this ;p ).