Journal Article
Towards Evolutionary Named Group Recommendations
Fulltext URI
Document type
Text/Journal Article
Additional Information
Date
43435
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Springer
Abstract
When sharing information, a common tactic for reducing the cost of choosing recipients is to form named groups of users. These groups are then selected as recipients in lieu of or in addition to users. However, keeping named groups up to date is a difficult and error-prone task when conducted manually. Past schemes automating this task make different tradeoffs and can be distinguished based on several factors including the types of named groups they consider, whether they evolve a specific group or a set of multiple groups, and how integrated they are with techniques for predicting initial groups. We analyze these approaches and identify a design space of potential evolutionary approaches. Using this analysis, we introduce a novel approach for automatically suggesting a sub-type of evolution, evolutionary growth. This approach (a) requires no prior knowledge of which groups change, (b) composes, and therefore interoperates, with an existing engine for recommending named groups, and (c) extracts groups from the social graph of multiple types of applications regardless of whether the graph are explicit or derived implicitly from message communication. Our evaluation considers social graphs created using explicit and implicit connections, and identifies the conditions under which the approach outperforms baseline techniques.