Users of information-retrieval systems are now-days spoiled by the simplicity of having to provide a short query in a plain search box, and expecting the retrieval system to do all the guesswork and come up with all the right answers. The results are often very good but there is a lot of room for improvement, after all in the scale of scientific age, information retrieval is only a baby.
In particular, some queries are inherently ambiguous. For example, a user may give the query "salsa" with Mexican cuisine in mind, or perhaps she is looking for the article about "Celia Cruz", the famous salsa singer, instead. In the MOTIF demo, we are experimenting with a "what if" question, namely, what can we gain if the system has a little more information of what the user is looking for. So, we have built a retrieval system that attempts to answer a query provided that it also knows the "context" of the query.
Returning to the example above, the appropriate context for the first case would be "nutrition", while the context "music" gives more relevant results for the latter case. Our demo, which is built around Wikipedia, lets users to search for Wikipedia articles. In addition to a query, they are also asked to provide a context for that query. To simplify matters, we try to make a guess for a set of concepts that are most relevant to a given query, however, to get the full experience, we invite users to come up with their own contexts.
Obviously the concept is general and it may be used in various other settings besides Wikipedia. The underlying algorithm is described in the article of Ukkonen et al. The purpose of the demo is to explore the idea of contextual search. It is not optimized for efficiency, and should not be considered as a product ready for serious use.
Feedback and suggestions to improve the demo are most welcome at firstname.lastname@example.org
"Searching the wikipedia with contextual information", Proceedings of the 17th ACM Conference on Information and Knowledge Management, CIKM 2008.