Travel Track Recognition Based on Natural Language Process

摘要

Travel track surveys have been an effective method to analyze travel rules. The internet is filled with large amounts of text data. This kind of data contains abundant information about travel tracks that can be revealed via natural language processing (NLP). This article presents a new way to recognize travel tracks based on text data, differing from GPS data and cellular signaling data; it is a novel method to reflect the real trace of a person. First, thousands of travel notes was collected and preprocessed as a training set. Second, the paper utilizes Bi-GRU and CRF to train these long text data. Then we used a neural network model to recognize site names and temporal words described in the text. Lastly, the trace emerges after individually mapping the site names and temporal words.