Webis at the CLEF 2017 Dynamic Search Lab

We briefly describe our approach to the query suggestion task at the CLEF 2017 Dynamic Search Lab. The general research idea of our contribution is to evaluate query suggestions in form of keyqueries for clicked documents. A keyquery for a document set D is a query that returns the documents from D among the top-k ranks. Our query suggestion approach derives keyqueries for pairs of documents previously clicked by the user. The assumption then is that the not-already-clicked documents in the top results of the keyqueries could also be interesting to the user. The keyquery suggestions thus focus on retrieving more documents similar to the ones already clicked. Another reasonable approach might instead focus on suggesting queries covering aspects of a user’s information need different to the ones in the already seen documents. However, in our this year’s contribution to the Dynamic Search Lab, we explore the utility of the “similar” suggestions that keyqueries probably produce.