Adaptive Conversation System Based on Script : First Work:Construct Script with Vector and Classify it with SIFT

摘要

Our work is about conversation system based on script, supposing to design an adaptive agent. Purpose of this paper is to verify our idea of the transforming from text to image, and to implement an approach to find a new way of feature characterization on conversation text for our system. With sent2vec and SIFT, we designed a way of expression of text features and conducted experiments to classify conversation situations as Image Processing. Good results in close tests and bad results in open tests indicate the necessity of further work on our idea and the insignificance of the new feature characterization.