Activities

Research

Current

• BioTeKS (Biology Text Knowledge Server) - I'm currently working on text analysis and text mining techniques for the Biomedical and Life Sciences domains.  The goal is to support Biomedical research (e.g., genomics, proteomics, drug discovery, etc.) with techniques that facilitate access to the literature in the domain.  One of our first demonstrations of this technology is document clustering, which has been integrated into the BioDictionary built by the Bioinformatics & Pattern Discovery Group at IBM Research.  To see the clustering tool in action, go to the BioDictionary demo, select one of the sample sequences, press "compute", follow the "Relevant PUBMED References" link, then click one of the "Cluster abstracts" links.
• Question Answering - I worked with John Prager, Anni Coden, and Dragomir Radev to develop Predictive Annotation for Question Answering (see our SIGIR paper).  I also developed the core search engine used by our Question Answering system, and I continue to collaborate with John's group in this area.

Recent

• MeetingMiner - The primary goal of this project is to make meetings more productive by automatically analyzing the meeting discussion and providing the meeting participants with relevant information from related knowledge sources.  In particular, the system uses the IBM ViaVoice® speech recognition system to convert the speech to a text transcript, applies a series of analyzers to identify various information elements in the transcript (e.g., questions, named entities, topics, etc.), automatically creates queries, submits those queries to other relevant knowledge repositories, and integrates the results of those searches back into the meeting.  For more information, see the recent Think Research article on MeetingMiner.
• Text Categorization - I have also worked on automatic text categorization, with particular emphasis on Web environments.  The goal is to automatically classify documents into a Yahoo!® style taxonomy.  We developed a system, called WebCat, that uses a k-Nearest Neighbor classifier at its core and exploits many of the attributes specific to Web documents.  For more information on this project, see the article in IBM's Think Research magazine.  

Workshops

• SIGIR 2001 Workshop on Information Retrieval Techniques for Speech Applications (chaired by Anni Coden, Eric Brown, and Savitha Srinivasan.  See the the workshop report in SIGIR Forum 36(1).
• SIGIR 1998 workshop on Hypertext Information Retrieval for the Web (chaired by Eric Brown and Alan Smeaton). See the workshop report in SIGIR Forum 32(2).

Professional

• I am currently the Information Director for ACM SIGIR (Information Retrieval).  If you're interested in information retrieval, web search, clustering, categorization, question answering, text analysis, or anything related, check out the website and consider joining the SIG!
• I typically serve on the program committees for the ACM SIGIR and CIKM conferences, and review articles for journals in the field of information retrieval, such as ACM TOIS. 
• National Engineers Week - I'm involved in the IBM Research program to visit local middle schools and educate students about engineers, including what they do, who they are, and the advantages of pursuing a career in engineering.