AFINN

AFINN is a sentiment lexicon that assigns sentiment scores to English words for use in natural language processing. It was created by Danish researcher Finn Årup Nielsen and released around 2011. The lexicon maps individual words to integer sentiment scores, typically in the range from -5 to +5, where negative values indicate negative sentiment and positive values indicate positive sentiment. Instances like love might be scored +3 and hate -3, though exact values vary by version. Words are usually lowercased and tokenized; some versions also include multi-word phrases.

The most commonly cited version is AFINN-111, which contains 111 entries, though other versions exist with more terms. The lexicon is designed for use as a simple, fast baseline for sentiment analysis: to estimate the sentiment of a text, one aggregates the scores of the words present in the text, optionally after preprocessing such as stemming and negation handling.

AFINN is widely used in academic and hobbyist NLP as a starting point for sentiment classification, due to its simplicity and portability. It can be loaded from plain text files or via libraries in Python and other languages.

Limitations include shallow representation, limited vocabulary, and sensitivity to context, sarcasm, negation, and domain-specific language. As a result, AFINN is often combined with other features or used as a stepping stone to more sophisticated methods.