|
Logo
Draft for Public Comment

6. Automated recommender systems for repositories

1 As a user, I want to receive recommendations about content that is of potential interest to me and related to my work, so I increase my knowledge in my field. Recommendation systems can significantly help users who are navigating large-scale research collections and datasets. Recommender systems are believed to be one of the key functionalities making academic social network sites popular. As a repository user, I want to be able to discover new research outputs related to my interest, both pro-actively when browsing as well as in the form of notifications, regardless of the place in which they are stored. Equally, I want to be able to discover and identify important people, relevant scientific methods, conference/journal/meetup venues, funding opportunities, etc. in the research field I am interested in. Recommendation systems today are typically based on related content or user-interaction based. The most successful systems use a combination of the two. Repositories currently only make it possible to build content-based recommender systems. Leave a comment on line 1 0

2 To enable the creation of state-of-the-art recommender systems for repositories, they will need to: Leave a comment on line 2 0

  • Offer a machine interface providing access to anonymised user-interaction logs, in the form of a triple <user, activity, time>, where user is either an anonymous session or an anonymized (e.g. hashed) user id and where activity is, for example, a download or view event.
  • Enable a secure cross-repository user profile (for personalised recommendations and notification) holding the user’s interests which can be user specified or derived based on previous interactions with any of the repository systems in the global network. Such profiling can also be based partly on an existing service.

3   Leave a comment on line 3 0

Page 12

Source: http://comment.coar-repositories.org/7-6-automated-recommender-systems-for-repositories/