Abstract: A text mining based identification model for railway infrastructure risk was proposed in this paper. The model used segmentation algorithm based on Cascaded Hidden Markov Model (CHMM) to deal with data in the form of long text, which recorded railway infrastructure quality problems. Then, the word frequency was calculated and the railway infrastructure management risk was identified. The analysis result was intuitively and clearly displayed by using the visualization technology of word cloud. The proposed model was experimentally verified by using 4 662 records of railway infrastructure quality problems in Lanzhou Railway Administration between January 2012 to April 2016.