pysentimiento 0.1.1

A Transformer-based library for Sentiment Analysis in Spanish

PySentimiento: Sentiment Analysis in Spanish

A simple Transformer-based library for Sentiment Analysis in Spanish (other languages coming soon!).

Just do pip install pysentimiento and start using it:

from pysentimiento import SentimentAnalyzer
analyzer = SentimentAnalyzer()
analyzer.predict("Qué gran jugador es Messi")
# returns 'POS'
analyzer.predict("Esto es pésimo")
# returns 'NEG'
analyzer.predict("Qué es esto?")
# returns 'NEU'

analyzer.predict_probas("Dónde estamos?")
# returns {'NEG': 0.10235335677862167,
# 'NEU': 0.8503277897834778,
# 'POS': 0.04731876030564308}

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"

Trained models so far

• beto-sentiment-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

python bin/train.py "dccuchile/bert-base-spanish-wwm-cased" models/beto-sentiment-analysis/ --epochs 5

3. Upload models to Huggingface's Model Hub

TODO:

• Upload some other models
• Train in other languages
• Write brief paper with description

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)