Research On Text Classification Technology Based On Railway Text

Railway safety supervision texts play an important role in ensuring the safe operation of railway.Relevant departments carry out a lot of safety supervision every day and form corresponding text records.On the basis of safety supervision record the severity of the problem described in the text,such as different responsibility Duan Wu railway professional knowledge,focus on the text classification under the direction of natural language processing technology,the research for railway safety supervision and the key technology of text records in accordance with the classification rules,the road network operation and safety harness big data more category has important research significance and application value.Based on the classification of railway safety supervision records,this paper studies the classification of railway safety supervision records based on the demand of railway safety supervision records.The main work is as follows:(1)It is put forward for railway safety supervision of text feature extraction methods: against traditional TF-IDF and Word2 Vec method based on the lexical characteristics used in professional vocabulary rich and complex railway safety supervision and the limitations and shortcomings,on the record Level a Character feature extraction method is proposed CLW2V(Character Level-Word2Vec),the method can be used to extract to the more granular text features.(2)The text classification method of railway safety supervision based on CLW2 V was proposed: using the feature extraction method realized by CLW2 V,combining with the convolutional neural network to carry out model pre-training according to different categories of railway safety problem.Library and generate a pre-training classification model,the multi-category classification of railway safety supervision text was realized.(3)Based on railway safety supervision text classification method of text mining technology,through the classification strategy based on railway safety supervision text classification method to railway safety supervision text reference to different extent,professionalism,responsibility and event type category more category matching,cascade realization of railway safety supervision text more category classification.The auxiliary function of classification matching is demonstrated to provide an efficient auxiliary function for the related work flow in railway safety supervision.The experimental results on the obtained data set of railway safety supervision show that the above methods have achieved the expected results,and solve the problem that the traditional methods rely on the word segmentation in advance and the feature extraction effect is not good,and give the application path in the text of railway safety supervision based on the above contributions.