A Short Text Similarity Algorithm for Finding Similar Police 110 Incidents

Finding similar police 110 incidents from the incident dataset plays an important role in recognising related cases from which the investigators could find more clues and make a better decision on police deployment. We aim at finding 110 incidents with similar case features and semantic compared against a given incident. A short text similarity algorithm is presented. Our algorithm is developed from a novel semantic similarity algorithm Word Mover'd Distance(WMD). In order to emphasize the significance of case features in incident text, the method introduces the traditional term frequency-inverted document frequency(TF-IDF) as term weights to the WMD. Then the algorithm is verified on the practical dataset of public security department to find similar incidents, and experiments show that the algorithm is effective and can improve the accuracy in finding similar police incidents.