pysentimiento

Why we included this project

pysentimiento gives you pretrained transformer models for sentiment, emotion, hate speech, and irony detection in Spanish, English, Italian, and Portuguese, so you can skip the training step and go straight to classifying. A single create_analyzer call returns predictions with full probability distributions, which makes it easy to drop into a moderation pipeline or a research script. Its tweet preprocessor handles the messy parts of real social text, replacing handles and URLs, shortening repeated characters, normalizing laughter, and expanding emojis, the kind of cleanup that usually eats up time before classification. It also covers NER and POS tagging for Spanish and English, so opinion-mining projects can stay in one library. One caveat: the models are trained on third-party datasets with non-commercial restrictions, so check those licenses before using it in a commercial product.