Abstract: In order to analyze the personnel's work plan and actual implementation, and provide the basis for personnel assessment, based on text mining technology, this paper carried out the performance analysis of railway security supervisors and the text similarity calculation method was designed. BiLSTM-CRF algorithm combined with Bidirectional Long Short Time Memory(BiLSTM) network and Conditional Random Field(CRF) was applied to implement the extraction of named entities in the personnel performance plan and the realistic text, and the conceptual similarity calculation method based on the Knowledge Network was adopted to calculate the similarity between the same entities, so as to implement the matching calculation between the plan and the actual reality in the personnel performance. Through the experimental analysis of the work plan and realistic text data of the work supervisors in a Railway Administration, the BILSTM-CRF algorithm has an accuracy rate of over 90% for each named entity. The accuracy of personnel planning and realistic matching is 83%. The experiment proves that text computing method based on BiLSTM-CRF and concept similarity is feasible in personnel performance analysis, and can also provide a reference method for similarity calculation of other texts in the railway field.