TSummary: A Traffic Summarization System Using Semantic Words.

Due to the popularity of GPS devices, road information and trajectory data are being generated in large quantities and accumulated rapidly. However, raw trajectory data is usually stored in the form of timestamps and locations. As a result, the raw data does not make much sense to humans without semantic representation, because humans cannot get an intuitive view of a group of trajectories that move on the same road, namely traffic of the road. In this work, we propose a partitionand-summarization framework which can automatically generate short text summaries for traffic description, aiming to advance human understanding of traffic information. In the partition phase, we first define a group of features that can be described in summaries, and then partition the roads. The purpose of the partitioning is to make the features of road segments in each partition as homogeneous as possible. In the summarization phase, the most interesting features for each partition are selected, and short text description are generated. For empirical study, experiments are conducted with a real trajectory dataset, and the experiment results have proven that the short text generated by the proposed framework can effectively reflect the important information of traffic.