Mapping search relevance to social networks.

ABSTRACTThis paper explores how information contained in the structure of the social graph can improve search result relevance on social networking websites. Traditional approaches to search include scoring documents for relevance based on a set of keywords or using the link structure across documents to infer quality and relevance. These approaches attempt to optimally match keywords to documents with little or no information about the searcher and no information about his network. This study analyzes 3.8M profile search queries from a large social networking site in conjunction with the tie structure of a 21M member social graph. The key finding is that a measure of social distance, when used in conjunction with standard search relevance methods, improves the ordering of profiles in search results.