Detecting emotion from dialogs and creating personal ambient in a context aware system

Abstract

This paper presents a personal ambient creation systme, IlluMe, which detects users' emotion from their chatting context in instant messages and then analyze them to recommend suitable lighting and music to create a personal ambient. The system includes a mechanism for recording users' feedback of the provided ambient to learn their preference. The aim of the proposed system is to link human language and emotion with the computer created environment seemlessly. To achieve this, we propose four apporaches to calculate emotion scores of words: Topical Approach, Emotional Approach, Retrieval Approach and Lexicon Approach. Natural language processing techniques such as normalization, part of speech tagging, word bigram utilization, and sentiment dictionaries lookup are incorporated to enhance system performance. Experiments results are shown and discussed, from which we find the system satisfactory and several future research directions are inspired. © 2013 Springer-Verlag Berlin Heidelberg.