Automatically Recognizing Emotions in Text Using Prediction by Partial Matching (PPM) Text Compression Method

Abstract

In this paper, we investigate the automatic recognition of emotion in text. We perform experiments with a new method of classification based on the PPM character-based text compression scheme. These experiments involve both coarse-grained classification (whether a text is emotional or not) and also fine-grained classification such as recognising Ekman’s six basic emotions (Anger, Disgust, Fear, Happiness, Sadness, Surprise). Experimental results with three datasets show that the new method significantly outperforms the traditional word-based text classification methods. The results show that the PPM compression based classification method is able to distinguish between emotional and nonemotional text with high accuracy, between texts involving Happiness and Sadness emotions (with 80% accuracy for Aman’s dataset and 76.7% for Alm’s datasets) and texts involving Ekman’s six basic emotions for the LiveJournal dataset (87.8% accuracy). Results also show that the method outperforms traditional feature-based classifiers such as Naive Bayes and SMO in most cases in terms of accuracy, precision, recall and F-measure.