A supervised multinomial classification framework for emotion recognition in textual social data
Abid Hussain Wani, Rana Hashmy · International Journal of Advanced Intelligence Paradigms · 2023
The task of emotion recognition from text has received much attention since the proliferation of online social networking which has woven itself into the fabric of lives of people world-over. This study is aimed at extracting the lexical and contextual information from the text and combining it with semantic information for the detection of the emotional state of a sentence. We propose a supervised framework for recognition of emotions from text in this work. Our framework utilises word embeddings from Word2Vec to extract the set of words which fall in semantic proximity of an affect-bearing word and also takes into account the context in which the words are used. We incorporate class-specific emoticon features in all our experiments as emoticons are commonly used on social media platforms. As the nature of social media text is generally very informal and has an irregular structure, our framework encompasses an appropriate mechanism to handle it. We evaluate our support vector machine-based framework on stance sentiment emotion corpus (SSEC) and Aman's dataset. The classification results achieved are better than state of art techniques currently available.