Sentiment analysis

Sentiment analysis is the use of natural language processing to automatically classify a piece of text, a review, a social post, or a support ticket, as expressing a positive, negative, or neutral opinion, sometimes with a numeric score in between.

Simple approaches count positive and negative words against a predefined lexicon; more accurate systems use trained classifiers or large language models that account for context, negation, and sarcasm. Output is often aggregated into a net sentiment score across a batch of text. This differs from topic modeling, which finds what people are talking about, whereas sentiment analysis judges how they feel about it.

Sentiment analysis underpins social listening and brand health tracking, letting teams monitor opinion at a scale no human reviewer could match. Its main weaknesses are sarcasm, mixed sentiment within one message, and domain-specific language: a model trained on product reviews can misread financial or political text. Neutral classification is also handled inconsistently across tools, so scores should be compared within the same tool over time rather than benchmarked across vendors.