Social media knows what road it is: quantifying road characteristics with geo-tagged posts

Abstract

Determining traveling routes that provide opportunities to satisfy the various requirements of users in urban areas is still an open problem. This is because it is virtually impossible to manually quantify the characteristics of each street or road and there are few web-based or semantic resources of subjective requirements that describe streets and roads directly. Thus, it is difficult to satisfy all the needs and desires of users that may arise, such as for finding boulevards that are “fashionable” or “beautiful.” The goal of this study is to automatically quantify the characteristics of streets and roads in relation to requirements that can be described using keywords such as “fashionable.” To achieve this goal, we propose a two-stage method that analyzes social media and road networks. First, in estimating the topic distribution (i.e., the characteristics) of each point of interest (POI), our method uses the latent Dirichlet allocation model to analyze geo-tagged texts while considering which users posted useful information for estimating road characteristics. Next, it uses a Markov random field model to estimate the characteristics of each street or road on the basis of those of the POIs and the road networks associated with the POIs. Experiments on real datasets demonstrate that our method achieves statistically significant improvements over baseline methods in terms of ranking quality in information retrieval for 300 roads in three urban areas in Tokyo when given 25 keywords.