- From: Ryan Lee <ryanlee@w3.org>
- Date: Mon, 18 Nov 2002 17:47:54 -0500 (EST)
- To: <www-rdf-logic@w3.org>
- Cc: <swick@w3.org>
A brief summary of the LCS seminar given 2002-11-18.
Goal of project:
- Match web portal users' research interests to some thing with a
'research profile' (originally to help new grad students find a thesis
project)
User experience:
- Standard web portal-type registration, discussion forums, events, etc.
- Novelty comes in with filling out a research profile for users. The
profile is based on a very large XML Schema ontology of computational
and biomedical research interests.
The interface is search/string matching driven, and results are XPath-like
('Computation/AI/') to help users understand the context of results. A
user can add a search result to their profile, and from there,
ontologically related choices for profile addition will also be
displayed.
Behind the scenes:
- Everything of major importance actually has a research profile, not
just users. Events, news, projects, professors; this is to facilitate
the major attraction, which is matching interests, or 'profile
matching'
- Profile matching is accomplished with some algorithm which went by too
quickly, but seems to be based on comparing tree depth (don't take my
word on that)
- Profile matching results in a score, and the highest scored things are
marked as being interesting to a user. Searching for a research
project or advisor is the main use of this, though anything can be
compared (users to users, users to events, etc.)
Other notes:
- Based on .NET
- Not yet live, so no results to speak for its effectiveness
- Also some NLP to build research profiles out of English descriptions,
though getting data into the system seems to be done best as a human
enterprise
- Parts of the system are fairly general and could probably be quickly
redone for a different research field, though with the same goal
--
Ryan Lee
ryanlee@w3.org
Received on Monday, 18 November 2002 17:48:29 UTC