A C LUSTERING TECHNIQUE FOR E MOTION D ETECTION FROM TEXT

Abstract

Emotion detection from text is a critical area of research in natural language processing, especially with the rise of social media platforms like X (Twitter) and Facebook. These platforms generate vast amounts of short text, where users frequently express their emo- tions. By analyzing these brief posts, unsupervised learning techniques can identify and categorize feelings from short text. Our work delves into the significance of emojis and keywords and their influence on interpreting emotions positively on X. It considers emotion detection from text on X, focusing on both English and the Algerian dialects. By utilizing ensemble clustering methods, the research aims to automatically identify and categorize emotions according to Ekman’s six basic emotions: happiness, sadness, anger, disgust, fear, and surprise. Ensemble clustering combines multiple clustering algorithms to improve the robustness and accuracy of the results, making it particularly useful for the informal and diverse nature of social media content. Our analysis shows that ensemble clustering per- forms better than single clustering methods. The silhouette score, a measure of clustering quality, is 0.82 for English data and 0.728 for Arabic data. Our findings suggest that ensem- ble clustering methods improve emotion detection in X text for both English and Algerian dialect.