ABSTRACT One important class of recommender system involves people as both the subject and object of the recommendation. Some examples are: employment web sites, which help a job seeker and employer and the right employer nd each other; dating web sites; mentor-mentee matching systems.
This paper denes reciprocal recommenders, showing their similarities to and differences from more conventional recommenders, identifying key approaches for creating this class of recommender.
One contribution of this work is the denition of the properties of, and promising approaches, for a new class of recommender, the reciprocal recommender.
Traditionally these systems have been used to recommend items to users.
We evaluated our approach by implementing the reciprocal recommender approach into RECON and evaluating it on a major Australian dating website.
We report analyses of the impact of dierent elements of the approach.
The contributions of this paper are the de nition of this class of recommendation system, the identi cation of the particular personalisation challenges for them, the proposition of some promising techniques to address these challenges.
We illustrate these concepts with a case study in online dating.
Some online dating sites had been using Behavioural Bidirectional Recommendation Engines for years, like Plenty Of Fish, and they could not outperform compatibility Matching Methods based on personality profiling.