Current Projects
- Program for University Librarians in the
Sciences (PULS) - training science students to be science
librarians
- Machete - bringing
knowledge management to the disk and bench top
- Machete
Wiki
- PDF
Extractor - This prototype extracts the text and figures from PDF
scholarly documents and reconstructs a representation useful for
literature analysis (alá CiteSeer)
and information extraction.
- Entity Extraction -
identification and extraction of 'things' from full text
- Newswire text: named
entities including persons, organizations, places, etc.
- Medical literature:
genes, proteins, organisms, etc.
- Relationship Extraction -
connections between (usually named) entities
- Topic Detection and Tracking
- detect and track stories in news feeds from multiple sources in
multiple languages.
- Question Answering -
Moving beyond the document as the artifact to be retrieved leads to
interesting problems in information retrieval.
- Video Analysis and Retrieval
- Our work so far involves
- Shot Boundary Detection:
Given a video, find all of the boundaries present, and indicate whether
they are cuts or gradual transitions.
- Story Boundary Detection:
Given a news broadcast video, find all of the boundaries between the
stories. Our approach involves a mix of shot boundary detection,
commercial recognition, broadcast anchor recognition, and analysis of
speech recognition transcripts.
- Feature Extraction:
Given a collection of videos already segmented into shots, find shots
containing various features of interest (specific people, events, etc.).
This page is not optimized for any browser!
Last modified Thursday, 17-Feb-2005 13:26:43 PST
