Exploiting social media information toward a context-aware recommendation system

Abstract

The rise of the social networks during the last few years has provided a vast amount of knowledge in several domains. Among them, route planning and point-of-interest recommendation have significantly benefited. Seen from the side of a tourist, they consist two challenging and time-consuming tasks since they may rely on many parameters and are limited by several constraints, such as time and budget available, user preferences, etc. In this paper we present Xenia, a context-aware system that works toward solving the aforementioned problems. More specifically, it aims to automatically construct travel routes, i.e., ordered visits to various places-of-interest. The user (tourist) indicates an initial and an ending point and her/his available time budget and the system proposes travel routes that maximize her/his travel experience, while adhering to the aforementioned limitations. This particular route planning problem is widely known as the “Tourist Trip Design Problem,” having several variations. In this work we solve this problem by modeling it through the “Orienteering Problem.” We harvest geo-tagged photos from the well-known social network Flickr and using the user-generated textual metadata that accompany them we extract areas-of-interest within a given city along with their underlying semantics. Moreover, by utilizing both the timestamps and the geo-tags of the photos we are able to identify the trajectory patterns of tourists, to detect popular places-of-interest and finally to estimate the average visit duration. Using this historical data we propose travel routes for four of the most popular Greek cities. The effectiveness of our approach is validated upon a twofold validation consisting by (a) a comparison versus the most typical baselines that have adopted by state-of-the-art works and (b) an empirical evaluation by real-life users.