In view of the shortcomings of traditional operational knowledge base construction, proposed an automatic hierarchical knowledge extraction model to build the knowledge base on service desk. Firstly, we made quantitative analysis on short text of events worksheet by the establishment of production and operation feature lexicon. Then, we used the improved KNN text classification algorithm to classify the short text of events. Finally, we completed knowledge discovery by the field topic rules. We applied this method to the construction of the knowledge base of enterprise service desk. It can automatically complete the transformation of events worksheet to knowledge, which not only makes up the defect of the hand to create knowledge, but also promotes the automation of the whole operational process.