Mention classification in Insights

Insights provides additional information about each mention so that you can understand relevant online conversation on a deeper level.

Insights uses custom statistical classifiers (built in-house by Brandwatch) that predict sentiment and emotion based on a training set of millions of pieces of data, including words, phrases, emoji, and other patterns. In addition, Insights infers gender and location for each mention, where possible.

Sentiment

Insights performs sentiment analysis on every mention, using an AI-driven model based on a training sample of 500,000 mentions. Insights assigns a sentiment of positive, negative, or neutral to each mention.

Team admins can manually change the sentiment assigned to a mention. See Filter and analyze your saved searches for more information.

Emotion

Insights performs emotion analysis on every mention, using an AI-driven model based on a training sample of approximately 2 million tweets, manually labeled with emotional categories.

Insights classifies mentions based on six emotions: anger, disgust, fear, joy, surprise, and sadness. If no emotion is detected, Insights does not assign an emotion to the mention.

Location

Location data is sometimes provided directly by the platform or by the author of the mention. When an author or mention has no explicit location, Insights attempts to infer a location based on the keywords in the author’s bio.

Gender

Insights infers gender for a mention using a statistical method based on the binary gender distribution of author names in census data and other public records.

Supported languages

Insights detects and returns mentions data in the following languages. Languages marked with an asterisk (*) are not supported for sentiment analysis. Emotion analysis is supported for English mentions only.