[ GLOSSARY ]

Sentiment analysis

Sentiment analysis is the automatic classification of text as positive, negative or neutral, applied to comments, messages and mentions. Its job in social media is triage: surfacing the angry customer inside five hundred congratulations.

Modern systems use language models rather than word lists, which is why they now survive negation ("not bad at all") far better than the previous generation. Irony, niche slang and mixed feelings in one sentence remain the honest failure cases: expect a classifier to be usefully right, not always right.

The practical use is inbox ordering, not reporting. A sentiment pie chart in a monthly deck changes nothing; the same classifier sorting today's comments so the negatives are read first changes response time, which is the metric customers actually feel.

When evaluating a tool's sentiment feature, test it on YOUR comments: sentiment models are notoriously domain-sensitive, and a model tuned on product reviews may misread creator banter.

Related terms: Social listening · Engagement rate