QueryTracker was misbehaving due to high load on the previous machine on which it was hosted. I moved it to a dedicated box over the 01/30/05 weekend. Everyone's queries shoould still be in the system, but all training and explicitly submitted examples were lost. Hope that's not too big of a deal :) |
QueryTracker is a continuous query system implemented on top of the Google search engine. |
Once you're logged in, QueryTracker will ask you to name your queries using a text string identifier, and enter the corresponding query terms. Each query will be run through the search engine every 24 hours, and new links identified automatically. Documents you've already seen will also be identified, but only if new and relevant text has been added to them. |
Within 24 hours of adding a new query, results should be available for viewing. When you view results, QueryTracker will allow you to enter feedback on which links are or are not relevant to your query. QueryTracker uses relevant feedback documents to build a query profile which helps improve relevance predictions for new documents (and changes to existing ones) encountered in the future. Documents marked as non-relevant are not used as such; mark them just to let QueryTracker know you've seen the document and didn't think it was useful. If no feedback is given at all for a document, QueryTracker assumes it hasn't been read and will consider it to be new the next time it encounters it. |
Click to log in, or create a new account here! |