A customer intention aware system for document analysis

Document classification tasks can be divided into two sorts: supervised document classification and unsupervised document classification. Supervised learning algorithm always has a better performance than unsupervised learning. However, it is very difficult to assign enough teacher signal. In this study, we developed a customer intention aware system for document analysis. The system starts from an unlabeled document set, give out several cluster results. The user could fine tune the classifier by modifying some key documents´ labels. After several circles of learning and feedback, the system will finally understand the users intention and generates a suitable expert system. This is a kind of semi-supervised learning. For clustering, we use weighted comparative advantage (WCA) algorithm for clustering and supervised WCA for classification algorithm, respectively.