Sentiment score
A sentiment score is a number that expresses how positive or negative a piece of text is: below zero for negative, above zero for positive and zero for neutral, often on a scale from -1 to +1. Sentiment analysis produces it for a comment or a whole set.
How it works
Each text gets a score from a sentiment model. Scores can then be averaged across comments, tracked over time or compared by product, region or topic. Some teams report the share of positive, neutral and negative comments instead of an average, because an average hides polarization: a set of love-it and hate-it comments can average to neutral.
Teams check a model's scores against a hand-coded sample, watch for drift when language changes, and read the underlying comments before acting on a movement in the score. Tracked week by week, a sudden fall tells a team which comments to read first.
A worked example
In two weeks, the average sentiment score on a streaming service's app reviews drops from +0.42 to +0.18, and the share of negative reviews rises from 11 to 27 percent. Reading them points to one cause, an update that signs viewers out of the TV app after every episode.
Related
Automate this in Pulse
A sentiment score averaging near zero can hide customers who love a feature and customers who hate it, and Pulse's themes, grouped by what people said, put each camp's quotes under its own heading.