Geometric and Statistical Analysis of Emotions and Topics in Corpora

Abstract

NLP techniques can enrich unstructured textual data, detecting topics of interest and emotions. The task of understanding emotional similarities between different topics is crucial, for example, in analyzing the Social TV landscape. A measure of how much two audiences share the same feelings is required, but also a sound and compact representation of these similarities. After evaluating different multivariate approaches, we achieved these goals by applying Multiple Correspondence Analysis (MCA) techniques to our data. In this paper we provide background information and methodological reasons to our choice. MCA is especially suitable to analyze categorical data and detect the main contrasts among them: NLP-annotated data can be transformed and adapted to this framework. We briefly introduce the semantic annotation pipeline used in our study and provide examples of Social TV analysis, performed on Twitter data collected between October 2013 and February 2014. The benefits of examining emotions shared in social media using multivariate statistical techniques are highlighted: using additional dimensions, instead of "simple" polarity of documents, allows to detect more subtle differences in the reactions to certain shows.