|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
ABSTRACT
Work on evaluating and improving the relevance of web search engines typically use human relevance judgments or clickthrough data. Both these methods look at the problem of learning the mapping from queries to web pages. In this paper, we identify some issues with this approach, and suggest an alternative approach, namely, learning a mapping from web pages to queries. In particular, we use human computation games to elicit data about web pages from players that can be used to improve search. We describe three human computation games that we developed, with a focus on Page Hunt, a single-player game. We describe experiments we conducted with several hundred game players, highlight some interesting aspects of the data obtained and define the 'findability' metric. We also show how we automatically extract query alterations for use in query refinement using techniques from bitext matching. The data that we elicit from players has several other applications including providing metadata for pages and identifying ranking issues. REFERENCES
Note: OCR errors may be found in this Reference List extracted from the full text article. ACM has opted to expose the complete List rather than only correct and linked references.
INDEX TERMS
Primary Classification:
Additional Classification:
General Terms:
Keywords:
Collaborative Colleagues:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||