Recommendation types

RISE – Recommendation Types

The RISE MyRecommendations search system is going to provide three types of recommendations:

1. Course-Based “People on your course(s) viewed”
This type of recommendation is designed to show the user what people studying their module have viewed recently.  At the moment this only picks up the first module that a student is studying but we are planning a future enhancement that will include all the modules that are being studied with a feature to allow users to flag which module they are currently looking for resources for.  The recommendations are generated by analyzing the resources most viewed by people studying a particular module.

2. Relationship Based “These resources may be related to others you viewed recently”
These recommendations are generated for a resource-pair. For example, if users commonly visit resource B after viewing resource A, the system considers this to be a relationship, and will recommend resource B to people viewing resource A in the future.  As the system doesn’t host resources internally, it instead looks at a user’s previously viewed resources (most recent), and then checks for the most often viewed resources by users who’ve also viewed the same (most recent) resources.

3. Search Based “People using similar search terms often viewed”
We have limited data on search terms used, from the EZProxy logfiles so we are using the searches carried out in MyRecommendations to build search recommendations. Using this we associate search terms used with the resources most often visited as a result of such a search. For example, if people searching for ‘XYZ’ most often visit the 50^th result returned from Ebsco, this part of the recommendation algorithm will pick up on this.  Hence in future when people search for ‘XYZ’, that particular result will appear top of the list of recommendations for users in a “People searching for similar search terms often viewed” section.