Relevance Ranking for Vertical Search Engines

List of Figures

Number	Figure	Page
2.1	Scatter plot of CTR versus editor’s relevance judgments.	11
2.2	Intuitive illustration of the proposed Joint Relevance and Freshness Learning (JRFL) model. The user-issued query and the corresponding returned URLs are represented by their features on the left part. Dashed-arrow lines in the middle indicate the assumed user’s judging process before she clicks. The check boxes on the right record the clicked URLs.	13
2.3	Coordinate descent for JRFL.	15
2.4	Scatter plot of CTR versus JRFL’s prediction. (a) Object function value update. (b) Pairwise error rate (PER) update. (c) Query weight αQ update.	20
2.5	Scatter plot of CTR versus JRFL’s prediction.	26
2.6	The architecture of the news search result clustering.