As an implicit inference mechanism, classification has been successfully applied to solve a wide variety of natural language processing (NLP) problems, such as sentiment analysis, named entity recognition, semantic role labelling, speculation detection, and so on. With text classification techniques, different kind of useful information can be obtained from text, and these information can be further used to support complex tasks like machine translation. In this project, we are exploring how to design effective text classification algorithms for solving different NLP problems. Structure information about datasets and text data is being considered in our design. We experiment with different NLP problems and different datasets.