Dan Chester's Research Interests and Expertise

Parsing Languages with Free Word Order
While many parsing strategies have been developed for parsing sentences in languages like English and French, which have highly restricted word order, new strategies are needed to efficiently parse the many languages that allow considerable freedom in the order of words. We seek a grammar formalism and parsing strategy that parses the constrained-word-order and free-word-order aspects of languages such as Russian, Latin, Korean and Walbiri with equal efficiency. Several approaches to identifying dependencies between words are being investigated and applied to the parsing of Esperanto, a highly inflected, agglutinative language, which is over a hundred years old, has a large community of speakers, and is probably the easiest free-word-order language to study.

Image Query System
This project explores knowledge representation issues in a context other than robot arms. Workstations that can show photographic quality pictures are now common. The image query system displays such pictures and then answers questions about them. This ties together the problems of representing the meanings of natural language with those of representing the information content of images. (Think of this as a robot with disabilities; it can't move, but it can communicate in natural language and it can see pictures, either stored in files or live from a video camera.) While meaning representations of natural language appear as sharp, symbolic expressions, the information content of images is usually fuzzy, both because the image is fuzzy and because it is often not clear where one object or event ends and another begins. The focus of this project is on the mapping between the symbolic and image representations.

Intelligent Advising Language System
This project attempts to tie together the natural language work by various faculty and graduate students in the department into a coherent system that takes natural language input and produces natural language output. My emphasis in this project is on parsing natural language input into meaning representations suitable for use by the modules written by the other natural language researchers in the department. Since much of the natural language work done here centers around the task of giving advice via a dialog with a user in need, the resulting product will be an intelligent advising system. It will provide a testbed vehicle for demonstrating and evaluating many of the other natural langauge processing projects in the department.


last updated March 24, 2003
return to my homepage