Multi-label Emotion Classification using Content-Based Features in Twitter

Abstract

Multi-label Emotion Classification is a supervised classification problem that aims to classify multiple emotion labels from a given text. Recently, Multi-label Emotion Classification has appealed to the research community due to possible applications inE-learning, marketing, education, and health care, etc. We applied content-based methods (words and character n-grams) on tweets to show how our purposed content-based method can be used for the development and evaluation of the Multi-label Emotion Classification task. The results achieved after our extensive experimentation demonstrate that content-based word unigram surpassed other content-based features (Multi-label Accuracy = 0.452, MicroF1 = 0.573, MacroF1 = 0.559, Exact Match = 0.141, Hamming Loss= 0.179).