Question Answering
I am currently working on Question Answering as my
PhD thesis topic. Question answering (QA) takes
information retrieval to the next level by allowing
a user to specify questions to which an answer
should be returned. Current technology, such as Google are
keyword-based, meaning they typically don't
understand what kind of answer you want and simply
return documents containing the keyword. Also,
current search engines return entire documents (or
document snippets) which the user must manually
search through to find an answer.
At the present moment in time, I'm looking for a
good method for answer type evaluation. For any
given question, only a certain subset of things
could possibly be an answer. If I were to ask the
question "Which city has ...", I would only expect
answers that are, in fact, cities. Automatically
discovering which answers are valid would improve
the performance of a QA system by filtering out
closely-related but incorrect answers should they be
of an invalid type.
For a bit more info on the state of the art in QA
have a look at TREC.
Metacomputing
My Master's
thesis work involved designing and implementing
a scheduling system that schedules a workload across
multiple computers so that the load is balanced and
the workload finishes as quickly as possible. For a
little more information, have a look at this page.
Other Natural Language Processing
I did some work on text summarization during the 2000 -
2001 academic year. This work culminated in a paper
being presented at the 2001
Document Understanding Conference (DUC 2001). You
can find the related paper on my publications page.
| 
