The Future of Search - Direction

Tim Berners-Lee's vision was of a participatory web. His idea was that a browser could also edit and the whole web would be collaborative. Web 2.0, however inappropriate a term it may be is a reality. Social networks such as MySpace, Digg, YouTube, Flickr, del.icio.us are hugely popular and their sheer weight of numbers renders them as powerful forces in the online media environment. The problem which these social networks pose to engines is that there is a marked difference between what users profess to wanting and what they actually use. Social bookmarking is big business. Votes, comments, tags, blogs, shared user behaviour and submitted choices allow users to collectively mould search to their needs. Yahoo! have been strongly advertising their increased emphasis on social search. Their Yahoo! Answers are being heavily marketed as a way of bringing the raw materials of search together to produce search ‘art’ through the application of "collective wisdom". Yahoo! have purchased both del.icio.us and Flickr recently and their published goal is "to change the game of search" and "tap the untapped authority" of users' (vii) .

Whatever the overt position offered by the engines is, user behaviour, whether measured explicitly or implicitly, is the future of search relevancy. An algorithm cannot ever recognise spam as well as a human, but the principles of social search are inherently flawed in two ways. Popularity does not necessarily indicate accuracy of information and users are notoriously bad at recognising their own behaviour patterns (viii) . This is likely to mean that, while social search is going to continue to feature heavily in assisting algorithms, it will be reduced to a smaller proportion of those calculations.

Social bookmarking has many excellent uses. In particular, it allows algorithms to pick up on dialect-specific expressions. For example, a British page about [lifts] may not ever turn up without an element of human intervention for an American user looking for the search term [elevators]. People who use similar bookmarks may well be looking for similar items. We already see this in ecommerce sites, such as Amazon’s hugely successful recommendations engine: 'Products tagged "bees" are also tagged:' (followed by a list of associated tags) (ix) and increasingly within SERPs, for example with Google OneBox (x) Results or Ask's Smart Answers (xi) , specific answers are supplied to the question most often associated with a search.

As search engines move more into other media channels, such as digitisation of books, papers and magazines, cable TV, streaming radio, computer game chatter, instant messaging and software chat, activities in offline media may come to influence search results. For example, a heavily advertised product in magazines may well be the result a searcher is looking for online rather than the result which might have otherwise been top of the SERPs.

Additionally, increased AI (Artificial Intelligence) and advances in neural net technologies coupled with the increased personal nature of search, may well allow cross pollination of trend analysis methods. This will allow the searches of groups of users who communicate or who follow similar patterns of usage both online and within other monitored media channels to have SERPS that are adaptive to their usage and that of their peers, thus tailoring what they see to more closely match what the majority of users and more specifically what their associates or “similar” classifications of users prefer, resulting in a truly organic and self adaptive search paradigm.

<< previousnext >>

Notes:
vii Jeff Weiner
viii Jakob Nielsen
ix A search for the tag [bees]
x As described by Google
xi As described by Ask