The Use of Zero-Shot Classification in Complex Emotion Detection

Natural language processing techniques have been developing rapidly over the years. Their aim is to better understand what the people are communicating, starting by classifying their messages into sentiments including positive and negative. From there, researchers developed machine learning techniques that could help us not only extract people’s words but also get an abstract meaning out of whole sentences. With those abstractive algorithms, like zero-shot classification, messages as a whole can be better classified into themes and emotions. Furthermore, recent studies have shown that humans do not only have basic but complex emotions, which are summarized up to twenty-eight. Both rapid advancements in psychology and technology fields have opened up a research gap and a technical challenge relating to the use of abstract zero-shot classification in complex emotion detection. This study has found that the new zero-shot classification is significantly more effective than the conventional text classification in detecting complex emotions, contributing to the theoretical understanding of the effectiveness of zero-shot classification, and its practical use for highly-accuracy emotion detection that the current text classification techniques cannot achieve.