From hate speech recognition to happiness indexing: critical issues in datafication of emotion in text mining

Abstract

One prominent application of computational methods is the identification of affectivity and emotions in textual data, commonly known as sentiment analysis. In this chapter, we explore the datafication of affective language by focusing on operationalization and translation involved in the analysis processes behind common methods to identify affectivity or specific emotions in text. We draw examples from popular cases and from our own empirical studies that apply and develop sentiment and hate speech analysis. We suggest that sentiment analysis is a fruitful case for discussing the role of and the tensions involved in applying computational techniques in the automated analysis of meaning-laden phenomena. We highlight that any application of sentiment analysis techniques to investigate emotional expression in texts amounts to an effort of constructing sentiment measurements - a process essentially driven by judgments made by researchers in an attempt to reconcile diverging conventions and conceptions of good/proper research practices.