Recommendations

When users browse through a web site they are usually looking for items they find interesting. Interest items can consist of a number of things. For example, textual information can be considered as interest items or an index on a certain topic could be the item a user is looking for. Another example, applicable for a web vendor, is to consider purchased products as interest items. Whatever the items consist of, a web site can be seen as a collection of these interest items.

Dynamically adding hyperlinks is often used for personalization and is the only approach that will be considered here. Recommender systems can present their recommendations in other ways however. Amazon.com for example, also delivers recommendations through email. Another approach is to display the average rating of an item from people who are correlated with the user.

Factors

Several factors can be considered in determining which documents should be suggested to the user:

  • The similarity between a document and the user profile.
  • The novelty of a document is determined by the existence of information in a document that is new to the user.
  • The proximity of a document is determined by the minimal number of links it takes to navigate from the current page to a page that presents the document.
  • Some recommender systems also check if a document is relevant to the information shown on the current page.

Adaptive Interfaces

In general, the better the web site is organized, the harder it will be to personalize the site. Ironically enough, many information filtering techniques can be used to improve the structure of a web site. An example are recommendations that Amazon.com makes when visitors select certain books. A related book is suggested if other visitors have purchased it along with the selected book. These recommendations are not personalized but are the same for every visitor.

Search engines exist that uses document clustering techniques to organize a directory of web sites. This directory is formed by partitioning web pages into domains via clustering. Most web sites, especially larger ones, however can never be perfectly optimized for all users. Users have different interests and personalizing a web site could help them find information faster as they otherwise would have or wouldn’t have found at all.

Leave a Reply