Exploring survey data through sentiment-weighing of questions

Abstract

Sentiment analysis allows a methodology to sharpen its scope of analysis to the emotion conveyed or extracted from a text. Using the data from the International Social Survey Programme (ISSP) 2015 Work Orientations IV survey, we implement sentiment analysis to measure the sentiment conveyed by the questions. We then used principal component analysis (PCA) and cosine similarity to compare responses across countries. We find that using sentiment analysis decreases the number of eigenvectors of the responses and results in a wider spectrum when comparing responses. Rather than making a general analysis of survey responses, we show the ability to highlight variances in the responses right away when we extract the sentiment aspect of the responses.