Information Retrieval Agents using Emotional Features

M.-G. Kim and Y.-T. Park (Korea)


emotion, information retrieval, human factor, affection


In this paper, we propose an information retrieval system that uses emotional features of documents. Our system examines each document and generates an emotional vector that is represented by emotion features like HAPPY, SAD, ANGRY, FEAR and DISGUST. Thus, we can retrieve documents for user query include with emotional feature. Through experiment, we have classify DVD title synopsis to comedy and horror area by classification methods which include correlation coefficient, naive Bayesian and k-nearest neighbor method. Also, we have experiment using clustering base on K-means. Experimental results show that using emotion features can improve the performance of text categorization and clustering system. In the other experiment, it shows emotional element variation that is inversely proportional. It means result for better performance in short text than large document.

Important Links:

Go Back