Deep learning framework for crash detection using Twitter data

Traffic crashes are one of the leading causes of deaths in the United States. Detection of traffic incidents will not only help traffic authorities deal with problems like congestion but also help provide faster emergency responses and improve travel time predictions. In this report, we propose a framework that can be used to build real-time incident detection models for various cities using Twitter data. We use various NLP techniques and deep learning architectures to find the model that best fits the data for each of these cities. Finally, we show that these city-specific models are computationally inexpensive, capture more accident related data for those cities and have better evaluation metrics than that of the combined incident detection model