Full Text Work

While we're avoiding full text analysis initially, it is a fruitful area given the system architecture.  As recounted by J.Pitkow in his 1997 GIT thesis:

Letizia manages a fequency based keyword profile for each user based upon visited Web pages. When a user is visiting pages, Letizia searches the pages connected to the current page and uses the keyword profile to determine relevancy.  Recommendations are then made to the user about which pages may be of interest.  This is more of an artful integration of ideas than a methodological contribution. p. 93.

In order to mimize impact on browser response time, MozWho will use a sequence of timeouts from page load.  If the user is still on the page and has shown some interest in it, or it matches in some way, further analysis of the page's links and content may be triggered.

It is also worth note, as also reported in this thesis, the top user reported problem in the mid/late 90s was finding known pages.

|||
Posted at NaN:NaN

Link Appearance

Another fruitful are for online activity is augmenting links.  The simplest approach conveys continuous variables recency of visitation, community ranking, or probability of interest. 

The probability angle was tackled by Chris Olston and
Ed H. Chi in  ScentTrails: Integrating Browsing and Searching on the Web.

Other types of link augmentation abound.
|||
Posted at NaN:NaN

Different Interests

A key challenge is separating a user's interest into distinct domains. JP McGowan, N. Kushmerick, and  B. Smyth offer some insight in Who do you want to be today? Web Personae for personalised information access.

|||
Posted at NaN:NaN