A flexible sentiment analysis predictor package supporting multiple pre-trained models, customizable preprocessing, visualization tools, fine-tuning capabilities, and seamless integration with pandas DataFrames.
sentimentpredictor is a Python package designed to classify sentiments in text using various pre-trained models from Hugging Face's Transformers library. This package provides a user-friendly interface for sentiment classification, along with tools for data preprocessing, visualization, fine-tuning, and integration with popular data platforms.
- Multiple Model Support: Easily switch between different pre-trained models.
- Customizable Preprocessing: Clean and preprocess text data with customizable functions.
- Visualization Tools: Visualize sentiment distributions and trends over time.
- Fine-tuning Capability: Fine-tune models on your own datasets.
- User-friendly CLI: Command-line interface for quick sentiment classification.
- Integration with Data Platforms: Seamless integration with pandas DataFrames.
- Extended Post-processing: Additional utilities for detailed sentiment analysis.
You can install the package using pip:
pip install sentimentpredictorHere's an example of how to use the SentimentPredictor to classify a single text:
from sentimentpredictor import SentimentPredictor
# Initialize the predictor with the default model
predictor = SentimentPredictor()
# Classify a single text
text = "I am very happy today!"
result = predictor.predict(text)
print("Sentiment:", result['label'])
print("Confidence:", result['confidence'])You can classify multiple texts at once using the predict_batch method:
texts = ["I am very happy today!", "I am so sad."]
results = predictor.predict_batch(texts)
print("Batch processing results:", results)To visualize the sentiment distribution of a text:
from sentimentpredictor import plot_sentiment_distribution
result = predictor.predict("I am very happy today!")
plot_sentiment_distribution(result['probabilities'], predictor.labels.values())You can also use the package from the command line:
sentimentpredictor --model roberta --text "I am very happy today!"Integrate with pandas DataFrames to classify text columns:
import pandas as pd
from sentimentpredictor import DataFrameSentimentPredictor
df = pd.DataFrame({
'text': ["I am very happy today!", "I am so sad."]
})
predictor = DataFrameSentimentPredictor()
df = predictor.classify_dataframe(df, 'text')
print(df)Analyze and plot sentiment trends over time:
from sentimentpredictor import SentimentAnalysisTrends
texts = ["I am very happy today!", "I am feeling okay.", "I am very sad."]
trends = SentimentAnalysisTrends()
sentiments = trends.analyze_trends(texts)
trends.plot_trends(sentiments)Fine-tune a pre-trained model on your own dataset:
from sentimentpredictor.fine_tune import fine_tune_model
# Define your train and validation datasets
train_dataset = ...
val_dataset = ...
# Fine-tune the model
fine_tune_model(predictor.model, predictor.tokenizer, train_dataset, val_dataset, output_dir='fine_tuned_model')By default, the sentimentpredictor package logs messages at the WARNING level and above. If you need more detailed logging (e.g., for debugging), you can set the logging level to INFO or DEBUG:
from sentimentpredictor.logger import set_logging_level
# Set logging level to INFO
set_logging_level('INFO')
# Set logging level to DEBUG
set_logging_level('DEBUG')You can set the logging level to one of the following: DEBUG, INFO, WARNING, ERROR, CRITICAL.
Run the tests using pytest:
poetry run pytestThis project is licensed under the MIT License - see the LICENSE file for details.
This package uses pre-trained models from the Hugging Face Transformers library.
Contributions are welcome! Please see the CONTRIBUTING file for guidelines on how to contribute to this project.