Recent Advances of Affect Detection from Arabic Text

Abstract

Emotion Detection (ED) from text has been an active research field recently. It has attracted the attention of researchers as it can measure the emotional contexts while humans interact with computers. Humans could express their emotion in various ways; using typed text, facial expressions, speech, gestures, and physiological measures. ED is considerably different from sentiment analysis SA, where SA goal is to detect polarity from text such as positive, negative or neutral. On the other hand, ED aims to recognize emotions from input text. Emotions can be modeled as discrete categories, e.g. Ekmans six basic emotions (angry, fear, joy, disgust, surprise and sadness). On the other hand there is the dimensional model that express emotions as valence, arousal and dominance values. Social media provides a rich source of emotional text, e.g. Twitter and Facebook. In this paper, we provide a review of recent work on ED from Arabic text. We discuss approaches (lexicons, machine learning, deep neural networks and ensemble approaches), tools for text processing, and We also discuss description of the most popular datasets in this domain.