Tracking and recognizing emotions in short text messages from online chatting services

To automate the process of emotion recognition, in this study, we develop a computational approach for continuously tracking and analyzing users’ emotions while chatting online. Our work has several unique features: it provides relative probabilities of possible emotions for a word, constructs a distribution for each chatting message accordingly, performs a clustering procedure for the message distribution, and aggregates the emotions of continuous chatting sentences to draw the conclusion. To evaluate the proposed approach, we conducted experiments in two phases. The first phase was to evaluate the effectiveness of the proposed computational approach in analyzing the chatting sentences. The participants were asked to focus on tagging emotions toward each sentence for a pre-designed dialogue. The second phase involves a real-time chatting between two online users. The participants were asked to choose topics and freely chat with each other. The messages were analyzed, and the results were provided to the users for their evaluations. The results show that our approach is both effective and efficient in tracking the emotions of chatting users. Additional analyses and further discussions were carried out to further evaluate the quantitative experimental results. All the findings confirmed the usefulness and feasibility of the presented approach.