Most of existing recommender systems are based on collaborative filtering or association rule, to build models based on occurrence.
For example, if I always open a dvd(B) page after I have visited a page of book A, they may learn patterns like:
if A visited, then B will be the next
The pattern is based on specific pages(urls), in other words, it not be applied on a brand new site, such as a Web site in China.
What EzSeer doing is in a different way, our model is not based on any specific information viewed, but how the user searches information.
For example, no matter where I am on the Web, either in Canada or in China, I use the same strategy to browse the Web, that is, use my ownbehavior model to search information.
Thus behavior based search technology could help the users no matter whether they are on the Web, and what they are working on.
No comments:
Post a Comment