In the face of massive, heterogeneous, dynamic text information, automatic text classification is of great significance. In recent years, the swarm intelligence theory and method, which has been gradually developed, provides a new intelligent method for text categorization. This paper attempted to introduce the mature particle swarm intelligence algorithm to the text classification field. The text preprocessing model was constructed, which was the foundation of text categorization model. A text categorization model Text PSO-Miner based on PSO was constructed and tested and compared on the vector space matrix of text set. Text PSO-Miner performance indicators were better than the classic classification model（SVM,KNN,NB） and ACO based text classification model. The results showed that Text PSO-Miner can be better applied to text categorization.

This paper introduces the current trend of the transformation of the domestic intelligence agencies to the think tanks. Based on the new paradigm of knowledge service of intelligent ecology in the environment of "Internet +", and Combed the relevant research at home and abroad, the paper constructs knowledge service system of the science and technology intelligence institute for think tank transformation. This complex knowledge service system puts forward four levels of knowledge organization、 knowledge management、knowledge flow and knowledge service. It is connected by six layers: network layer、data layer、team layer、product layer、service layer and user layer. It is composed of technical staff、 intelligence analyst、Industry experts and customers. It ultimately provides prospective、effective and professional intelligence services to governments and enterprises.