سند علمی

Dealing with the Cold Start Problem when Providing Personalized Enterprise Content Recommendations

چکیده

We demonstrate how we can take advantage of employee information and digital traces of interaction in order to provide personalized recommendations of content to enterprise users. In particular, we focus on the cold start problem encountered when a service lacks relevant data to base recommendations. Our recommendation service – Steer – first seeks out user preference data. In the absence of such data it utilizes a wide range of sources to create a user‟s enterprise interest profile, which is used to provide recommendations. Users that Steer cannot generate an interest profile for are given generic recommendations based on a novel algorithm. We describe the algorithm that drives Steer, how the service scales, and takes into consideration content that could be outdated. We conclude with a discussion of an evaluation plan for Steer and future directions we wish to explore.

اسناد مشابه