Sentiment is calculated by default for all projects and provides a valuable lens through which to explore your data, especially in cases where numerical ratings are not available. For an overview of how sentiment is calculated please refer here. In this guide we'll be using data from online reviews of a supermarket chain.
We'll start from the analysis Overview page, which shows sentiment totals across all verbatims. By clicking on one the sentiment bars, in this case negative, you'll run a query for that sentiment type.
On the Query page, you'll see a host of visualizations for interpreting the negative sentiment we're exploring, such as the Timeline, Context Network and Comparison Charts. For this guide, we'll focus on the Comparison Charts, which are at the bottom of the page; Clicking on the correlation button will switch display modes to show the concepts and segments that have the highest correlations with negative sentiment.
Accordingly, in the Concept Comparison chart you will see concepts that are the strongest candidates for drivers of the sentiment type you have selected. In this case we see concepts related to calling, emailing and speaking with the manager, as well as returns coming up as potential drivers of negative sentiment. From this point we could dive into the verbatims or run some follow-up queries to further explore this relationship. Please check here for more info on correlations and how they are calculated.
This process is similar for the Segment Comparison chart with an additional detail - by default you'll see the most correlated segments across all variables in your data. By clicking the highlighted dropdown you can narrow the displayed segments down to a particular variable, such as location or age.