Emotional analysis of the COLUMNAS.HUMOR corpus: A mixed approach

Abstract

The study of emotions and polarity in language has been the focus of increased attention over recent years due to its relevance in areas such as artificial intelligence and the analysis of sentiment on social networks, advertising, and communication in general, given that it seeks to understand how people interact and relate through language. Studies of this type are fundamental towards being able to interpret the feelings and attitudes of users and customers, as well as to improve human-machine interaction and the user experience in a variety of fields of activity. Polarity and emotions in language are especially relevant in the analysis of humour, since understanding the emotional patterns associated with humour can be useful in the development of artificial intelligence systems capable of detecting, understanding and generating humour effectively. The present study analyses data from the COLUMNAS HUMOR corpus (full texts and humorous sequences) with the aim of observing the polarity and predominant emotions therein. The methodology is corpus-based, with the use of digital tools to extract data on polarity and emotions and the application of quantitative and qualitative analysis. Drawing on different types of frequency-based findings, a qualitative analysis is conducted to interpret the data in each corpus under study. This exhaustive analysis has allowed us to detect differences in sentiment and emotional verbalisation in each corpus, as well as to contrast the results by corpus and by the gender of the studied journalists.