Abstract: For it’s important to exploit the spatio-temporal coherence of trajectory for analysing road network structures and human movement patterns, we present road2 vec, a visual analysis method for urban roads based on taxi GPS data. First we used word embedding technique to model the trajectory data and obtain vector representation of roads. Then we calculated the position relationship between vectors to get the similarity between roads. Finally we exploited the location information and connection relationship between roads in trajectories by similarity. Additionally, we have designed a visual analysis system to support the exploration. Based on taxi GPS data of Wenzhou City, the case study shows that our method can help us to effectively find the human movement patterns in urban road network.