To the fault diagnosis of radar reconnaissance equipment described with natural language,a fault diagnosis approach based on text classification and support vector machine was proposed.First,a fault feature lexicon was built up by analyzing the fault text sets and extracting the fault features.Then,the text vectors were represented with the Boolean model to construct the fault vector library.Finally,the diagnosis classification model was established by One-Against-One method of the SVM classification,and the parameters were optimized by using grid search method.And thus the fault diagnosis was implemented.The experimental result shows the effectiveness and the validity of the approach,and the maximum recognition accuracy of fault diagnosis could be improved to 90% ultimately.