Recommendation as Collaboration in Web Search

The quantity and diversity of the information content that is now available online is without precedent, and its relentless growth is set to continue, if not accelerate, as the web continues to evolve to accommodate new forms of user-generated content (Gulli and Signorini 2005). The popularity of the social web fur-ther escalates this issue by facilitating easy contribution and sharing of information at a pace that places strain on search engines to locate and index such a fast-changing space. The sheer scale and increasing heterogeneity of the modern web make for further significant challenges when it comes to pro-viding individual users and communities with access to the right information at the right time. Despite all of the recent developments in search engine technologies, modern search engines continue to struggle when it comes to providing users with fast and efficient access to information. For example, recent studies have highlighted how even today’s leading search engines fail to satisfy 50 percent of user queries (Smyth et al. 2005). Part of the problem rests with the searchers themselves: with an average of only two to three terms per query (Lawrence and Giles 1998, Spink and Jansen 2004), the typical web search query is often vague with respect to the searcher’s true inten-tions or information needs (Song et al. 2007). Moreover, searchers sometimes choose query terms that are not well rep-resented in the page that they are seeking, and so simply increasing the length of queries will not necessarily improve search performance. Problems with search queries aside, another challenge that faces all mainstream search engines is the ongoing battle with search engine spam through so-called black-hat search engine optimization (SEO) techniques and the rise of the content farms. The former involve the manipulation of search engine rankings in order to promote target pages. At the time of writ