Microsoft's contributions at the International Association for Computing Machinery's Special Interest Group on Information Retrieval (ACM SIGIR) conference in Seattle Washington has been the highlight of the conference. Microsoft was the largest contributor this year and, in relation to the subject discussed here, submitted two papers that examine an improvement to search engine results and ranking by integrating user browsing behaviours and patterns.
In considering research that was carried out by iProspect in conjunction with Jupiter Research in April this year - titled the Search Engine User Behaviour Study - it is clear to see that user behaviours are changing. The results from this and its preceding, comparable, studies undertaken in 2004 and 2002 demonstrate that users are now less inclined to navigate past the first page of results and even less so after the third page (only 10 percent) than in 2002.
Microsoft has clearly been reading up on such reports, as their researchers are clearly on the case. The first Microsoft paper of interest from the SIGIR is Improving Web Search Ranking by Incorporating User Behaviour Information, by Eugene Agichtein, Eric Brill and Susan Dumais, which stated that:
"Incorporating user behaviour data can significantly improve ordering of top results in real web search setting... [I]ncorporating implicit feedback can augment other features, improving the accuracy of a competitive web search ranking algorithms by as much as 31% relative to the original performance."
This thinking could mean reaching some of the 90 percent of users who don't navigate past the third page by altering or greatly improving your website's ranking to one of the first three pages by strengthening its relevance to the user's search, relying perhaps on previous search result success for that particular user or similar users.
Agichtein, Brill, and Dumais believe that ranking search results is a "fundamental problem in information retrieval". The current retrieval process is considered "two-dimensional", considering only the search query and the content displayed in the link structure of the search results page. An additional dimension, one that these researchers hope to incorporate, is the user: the "third dimension". The pattern of the user's browsing will be taken into consideration in the search results. This means that, for example, previous searches and conversions will alter rankings more dramatically than they currently might.
The research studies and results rely on "click-through and browsing patterns across a large number of users, [from which] we are able to learn a great deal about how people interact with search techniques and can thereby improve our accuracy dramatically."
The second paper, Learning User Interaction Models for Predicting Web Search Result Preference, by Agichtein et al, and Robert Ragno additionally, looks at evaluating user behaviour to create a model to predict what those users' search result preferences might be.
Publishing what they understand to be the first "post-search user behaviour" paper, their second paper achieves a model which they consider a success in "higher prediction accuracy". Their first paper, as mentioned above, which considers "implicit feedback" will ensure that ranking is greatly improved.
It's worth keeping on top of, since as user behaviour changes so shall search engine result accuracy, relevance and ranking. It's already happening to an extent for user behaviour tracking through search engine toolbars sending back behavioural information, but this increased emphasis could be the next step to truly intelligent SERP rankings.
