R&D
The goal of this project is to develop efficient algorithms for recommending arbitrary Web content (e.g. web pages, blog posts, videos, photos, etc.) to social network users, while taking both location and time into consideration. The idea is to extend the traditional recommender systems' methods that are solely based on user-item similarities (typically known as collaborative filtering), with geographic and temporal data. When a user uploads a photo to Flickr, for example, that photo has a location (i.e. where it occurred) and a temporal (i.e. when it occurred) dimension, which can help the recommender engine to better filter and contextualize the recommendations. The proposed algorithms will be evaluated against standard collaborative filtering (CF), and some algorithms developed at HP Labs borrowing from economic models of attention. The best performing algorithms will also be implemented and tested in the HP Gloe recommender system for on-line personalized suggestions of local Web content... [more]