pysentimiento: A Python toolkit for Sentiment Analysis and Social NLP tasks

A Transformer-based library for SocialNLP classification tasks.

Currently supports:

• Sentiment Analysis (Spanish, English)
• Emotion Analysis (Spanish, English)

Just do pip install pysentimiento and start using it:

from pysentimiento import SentimentAnalyzer
analyzer = SentimentAnalyzer(lang="es")

analyzer.predict("Qué gran jugador es Messi")
# returns SentimentOutput(output=POS, probas={POS: 0.998, NEG: 0.002, NEU: 0.000})
analyzer.predict("Esto es pésimo")
# returns SentimentOutput(output=NEG, probas={NEG: 0.999, POS: 0.001, NEU: 0.000})
analyzer.predict("Qué es esto?")
# returns SentimentOutput(output=NEU, probas={NEU: 0.993, NEG: 0.005, POS: 0.002})

analyzer.predict("jejeje no te creo mucho")
# SentimentOutput(output=NEG, probas={NEG: 0.587, NEU: 0.408, POS: 0.005})
"""
Emotion Analysis in English
"""

emotion_analyzer = EmotionAnalyzer(lang="en")

emotion_analyzer.predict("yayyy")
# returns EmotionOutput(output=joy, probas={joy: 0.723, others: 0.198, surprise: 0.038, disgust: 0.011, sadness: 0.011, fear: 0.010, anger: 0.009})
emotion_analyzer.predict("fuck off")
# returns EmotionOutput(output=anger, probas={anger: 0.798, surprise: 0.055, fear: 0.040, disgust: 0.036, joy: 0.028, others: 0.023, sadness: 0.019})

Also, you might use pretrained models directly with transformers library.

from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("finiteautomata/beto-sentiment-analysis")

model = AutoModelForSequenceClassification.from_pretrained("finiteautomata/beto-sentiment-analysis")

Preprocessing

pysentimiento features a tweet preprocessor specially suited for tweet classification with transformer-based models.

from pysentimiento.preprocessing import preprocess_tweet

# Replaces user handles and URLs by special tokens
preprocess_tweet("@perezjotaeme debería cambiar esto http://bit.ly/sarasa") # "[USER] debería cambiar esto [URL]"

# Shortens repeated characters
preprocess_tweet("no entiendo naaaaaaaadaaaaaaaa", shorten=2) # "no entiendo naadaa"

# Normalizes laughters
preprocess_tweet("jajajajaajjajaajajaja no lo puedo creer ajajaj") # "jaja no lo puedo creer jaja"

# Handles hashtags
preprocess_tweet("esto es #UnaGenialidad")
# "esto es una genialidad"

# Handles emojis
preprocess_tweet("🎉🎉", lang="en")
# '[EMOJI] party popper [EMOJI][EMOJI] party popper [EMOJI]'

Trained models so far

Check CLASSIFIERS.md for details on the reported performances of each model.

Spanish models

• beto-sentiment-analysis
• beto-emotion-analysis

English models

• bertweet-base-sentiment-analysis
• bertweet-base-emotion-analysis

Instructions for developers

1. First, download TASS 2020 data to data/tass2020 (you have to register here to download the dataset)

Labels must be placed under data/tass2020/test1.1/labels

2. Run script to train models

Check TRAIN_EVALUATE.md

3. Upload models to Huggingface's Model Hub

Check "Model sharing and upload" instructions in huggingface docs.

Citation

If you use pysentimiento in your work, please cite this paper

@misc{perez2021pysentimiento,
      title={pysentimiento: A Python Toolkit for Sentiment Analysis and SocialNLP tasks}, 
      author={Juan Manuel Pérez and Juan Carlos Giudici and Franco Luque},
      year={2021},
      eprint={2106.09462},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

TODO:

• Upload some other models
• Train in other languages

Suggestions and bugfixes

Please use the repository issue tracker to point out bugs and make suggestions (new models, use another datasets, some other languages, etc)