Introducing active learning on Text to Emotion Analyzer

Now-a-days, online interpersonal communications have become more preferable than face-to-face interactions. However, emotions play a significant role in online communication. Automatic extraction of emotions from the text is a hot research issue because it minimizes the communication gap and misunderstanding between users. To become emotionally more intelligent, our previous text to emotion analyzing system should communicate with experts for suggestions of possible emotional state if it fails to analyze the text. In this research, we augment our previous system by introducing active learning approach which allows to query experts for emotional label of the given text. It makes our training dataset enriched. To build a classification model by analyzing the training dataset, we employ Naive Bayes classification technique. Our classifier updates the emotional database automatically. We also develop a prototype of our system named TEA: Text-to-Emotion-Analyzer. Our experiment and evaluation section exhibits satisfactory results in terms of recall-precision over our previous system as well as other method namely Vector Space Model (VSM).