Affinn

Affinn, often written as AFINN, is a sentiment lexicon used in natural language processing to assign polarity scores to English words and phrases. It provides a mapping from lexical items to integer sentiment scores, which can be summed to yield an overall sentiment score for a piece of text. The lexicon is widely used in simple bag-of-words analyses and as a lightweight baseline in sentiment workflows.

Origin and scope: The lexicon was compiled by Finn Årup Nielsen and has been maintained in several versions since its initial release. It covers thousands of English words and short phrases, with scores typically ranging from negative to positive values, commonly between -5 and +5. Data are usually stored in plain text or other simple formats for easy loading into programming environments.

Usage: To apply AFINN, tokenize the text, look up each token in the lexicon, and sum the scores. Optional preprocessing includes lowercasing and handling negations or intensifiers. It is English-focused and works best on general text; domain-specific language may require adaptation or supplemental resources.

Licensing and availability: AFINN data and accompanying materials are distributed as open data through public repositories. Users should consult the license on the project site or repository for exact terms.

Limitations and related tools: The approach is context-insensitive and does not resolve negation, sarcasm, or rhetorical devices. It may misclassify phrases whose sentiment depends on context. For more nuanced analysis, resources such as VADER, SentiWordNet, or transformer-based models are often used alongside AFINN to improve performance.

---