Emotion analysis and support system for Blog

Abstract

Recently blog has become much popular and many have used it to express their feeling and emotion. Some organizations even tried to found frustrated by searching blogs. As those blogs are emotion rich, lots of emotion analysis researches have been done with blog. We are interested if the emotion analysis can used with emotion support system for the blog users. In this project we evaluate the effectiveness of emotion analysis can be done on local (Hong Kong) blogs, and experiment the way of giving emotion support on blogs based on the emotion analysis results. We investigate on emotion analysis and its implementation for blog, and if emotion analysis can be more accurate with trend recording. We also look into issues of emotion analysis involved with Chinese and local (Hong Kong) blog data and try out the way of giving emotion support by building a simple blog system with emotion analysis and support features. This project finds out that the effectiveness of classification is highly affected by the training dataset size, and emotion analysis on Hong Kong local (Hong Kong) blogs is feasible but larger and better corpora should be used. By using local words lists for trendy/Cantonese words in local (Hong Kong) the classification effectiveness can be increased, but it is not significant. It also find out that by turning the new blog posts in the dataset can help to increase the training dataset size and thus its accuracy. In this project it also finds out that the emotion analysis is feasible to run online as a real time web application in term of response time.