Conducting Sentiment Analysis and Post-Sentiment Data Exploration through Automated Means

Abstract

One new feature in NVivo 11 Plus, a qualitative and mixed methods research suite, is its sentiment analysis tool; this enables the autocoding of unlabeled and unstructured text corpora against a built-in sentiment dictionary. The software labels selected texts into four categories: (1) very negative, (2) moderately negative, (3) moderately positive, and (4) very positive. After the initial coding for sentiment, there are many ways to augment that initial coding, including theme and subtheme extraction, word frequency counts, text searches, sociogram mapping, geolocational mapping, data visualizations, and others. This chapter provides a light overview of how the sentiment analysis feature in NVivo 11 Plus works, proposes some insights about the proper unit of analysis for sentiment analyses (sentence, paragraph, or cell) based on text dataset features, and identifies ways to further explore the textual data post-sentiment analysis—to create coherence and insight.