Microsoft is tackling what it considers to be one of the most difficult computation problems associated with a search engine, namely relevance, through a simple gaming approach. In this context, Microsoft Research has introduced Page Hunt. According to the Redmond company, page hunt is a Human Computational Game set up to help with the evaluation process of the relevance of Bing. Essentially, the software giant is eliciting data from players in order to improve the search engine that killed off Live Search. “The data elicited using Page Hunt has several applications including providing metadata for pages, providing query alterations for use in query refinement, and identifying ranking issues,” the company explained.
Page Hunt is already live and can be accessed by users. The purpose of the game is to serve player random pages from the Internet. The company is looking to “get the player to come up with a query that would bring up this page in the top few results on a search engine. The web page being ‘hunted’ is shown in the background. The player types in queries and looks at results returned in the floating operating panel, initially in the lower right corner. This panel turns almost transparent when not in focus. The border of the panel provides feedback through animation and color changes,” the software giant added.
Microsoft noted that ahead of being released to web the game was tested by 341 users. The pilot took 10 days and produced more than 14,400 labels on 744 web pages. “On average every player contributed 42 labels, and every page has about 19 labels. Using a session limit of 10 minutes of inactivity, we had a total of 681 sessions. Of these, 18% (123 sessions) were by anonymous players. About 47% of the sessions were from people who played 2 or more sessions. 16% of the people had 5 or more sessions, and only 240 sessions (35% of the total number of sessions) were single session players,” the company explained.
Following the initial experiment, Microsoft approximates that some 27% of the webpages that were served to players had a findability score of 100%. This means that all persons participating in the test were able to bring the page in the top 5 list of results returned by Bing. But at the same time, 26% of all pages tested had 0% findability. “We investigated this further and found that as the length of URL of the page (in characters) increases, the pages are harder to find,” the Redmond company revealed.