In this paper, we examined click patterns produced by users of Yahoo! search engine when prompting definition questions. Regularities across these click patterns are then utilized for constructing a large and heterogeneous training corpus for answer ranking. In a nutshell, answers are extracted from clicked web-snippets originating from any class of web-site, including Knowledge Bases (KBs). On the other hand, nonanswers are acquired from redundant pieces of text across web-snippets. The effectiveness of this corpus was assessed via training two state-of-the-art models, wherewith answers to unseen queries were distinguished. These testing queries were also submitted by search engine users,.
