R-Rec: A rule-based system for contextual suggestion using tag-description similarity
This work addresses the problem of personalized recommendation for users in location-based services, but it is incremental as it builds on existing rule-based and similarity techniques.
The authors tackled the problem of contextual suggestion for points-of-interest by proposing R-Rec, a rule-based system that ranks suggestions based on similarity to liked places and dissimilarity to disliked ones, achieving efficacy as demonstrated on the TREC-Contextual Suggestion 2015 dataset.
Contextual Suggestion deals with search techniques for complex information needs that are highly focused on context and user needs. In this paper, we propose \emph{R-Rec}, a novel rule-based technique to identify and recommend appropriate points-of-interest to a user given her past preferences. We try to embody the information that the user shares in the form of rating and tags of any previous point(s)-of-interest and use it to rank the unrated candidate suggestions. The ranking function is computed based on the similarity between a suggestion and the places that the user like and the dissimilarity between the suggestion and the places disliked by the user. Experiments carried out on TREC-Contextual Suggestion 2015 dataset reveal the efficacy of our method.