In my work, I selected queries manually to assess my information retrieval model. These queries are supposed given by users. Now I'm looking for an automatic method that generates queries from a set of keywords. This query must have a meaning, so why not have it be based on the content of one or a set of articles.
You can try generating the queries based on term weight. Nouns or noun phrases would make good candidates. However, I think the real problem is in generating the relevance judgment. Unless you have a perfect retrieval system, you can not completely rely on the query result.
How about adding a sidebar to your system presenting a partial result from the generated queries and asking your users to mark the relevant documents?
You can try by using real queries people use on Google. They are available in Google Analytics and moreover on Webmaster Tools. If you or any colleague have a site related to the topic your studying, you cann use it.
Sorry for my "Spanglish".
Thank for your answer, but i had tried to have a dataset based on theses tools but as you know it's impossible if you haven't a direct access to these websites. I'm looking for queries in New websites such CNN, Independnt, Aljazeera ... If you have any idea about public datasets please don't hesitate to tell me :) Thank you !
Thank you very much but if all these queries all in spanish they will not be very usefull for me.
This is my email: [email protected]
Thank you very much !
- Badges
- Science topic