Search and destroy
Microsoft's new
search engine, Bing, could be improved at some point through the use of an
online game called Page Hunt.
Boffins at Microsoft Research have released a
new paper which suggests that the game, called Page Hunt, could be used to
refine queries and search results. Bing currently sits in third in the
U.S. search-engine market behind Yahoo and Google, even as Microsoft pumps
between $80 million and $100 million into the initial marketing. Page
Hunt, provides users with a random Web page, and then asks them to input the
search terms that will put that page within a search engine’s top five
search results.
Depending on how close to the top of the rankings their
queries put the Web page, players are awarded points. In order to sweeten
the experience, the game adds animations, a top-score list, bonus points,
and other "game-like" features. Page Hunt can be found on this site is
based around Redmond's Silverlight and it exists entirely as a research
project, with no direct connection to Bing.
The results it generates
could contribute mightily to the extraordinarily complex task of refining
the search-engine process. Microsoft needs to do this if it is going to get
any advantage over Google and Yahoo in the online search space. According
to the report, which has the catchy title "Page Hunt: Improving Search
Engines Using Human Computation Games" was written by Raman Chandrasekar and
Chris Quirk of Microsoft Research, with Abhishek Gupta of Digital Media and
Hao Ma of the Chinese University of Hong Kong. Using data from users gathered
in human computation games can improve search, the abstract of the paper
reads.
The original pilot experiment conducted by Chandrasekar and company
involved 341 Microsoft employees playing Page Hunt over a 10-day period,
generating 14,400 labels for the 744 Web pages in the system. The
researchers extracted the queries that corresponded to winning trials,
generated all pairs of queries as bitext data, and applied the bitext matching algorithm.