qhchina 0.0.43

A Python package for NLP tasks related to Chinese text.

QHChina Lab

Quantitative Humanities China Lab - A research lab and Python package for NLP tasks related to Chinese text analysis.

Repository Structure

This repository contains:

1. Python Package: The qhchina package for Chinese text analysis in humanities research
2. Lab Website: The QHChina Lab website located in the /docs folder with information about our research, projects, resources, and documentation

Python Package Features

• Collocation Analysis: Find significant word co-occurrences in text
• Corpus Comparison: Statistically compare different corpora
• Word Embeddings: Work with Word2Vec and other embedding models
• Text Classification: BERT-based classification and analysis
• Topic Modeling: Fast LDA implementation with Cython acceleration

Installation

pip install qhchina

Usage Examples

Topic Modeling with LDA

from qhchina.analytics import LDAGibbsSampler

# Each document is a list of tokens
documents = [
    ["word1", "word2", "word3"],
    ["word2", "word4", "word5"],
    # ...
]

# Initialize and train the model
lda = LDAGibbsSampler(
    n_topics=10,
    iterations=500
)
lda.fit(documents)

# Get top words for each topic
for i, topic in enumerate(lda.get_topic_words(10)):
    print(f"Topic {i}: {[word for word, _ in topic]}")

For more examples, see the module documentation.

Website

The lab website is built with Jekyll and includes:

• Information about the QHChina Lab and our research
• Project descriptions and updates
• Resources for Chinese humanities research
• Comprehensive documentation for the qhchina package
• Blog with updates and announcements

To run the website locally:

cd docs
bundle install
bundle exec jekyll serve

Documentation

For complete API documentation and tutorials, visit: https://mcjkurz.github.io/qhchina/

License

This project is licensed under the MIT License - see the LICENSE file for details.

For Developers

Building Wheels with Cython Extensions

The package includes Cython extensions for performance-critical components. When installed from PyPI, precompiled wheels should be available for common platforms.

To build wheels locally:

# Install build dependencies
pip install build wheel setuptools cython numpy

# Build the wheel
python -m build --wheel

The precompiled wheels are built using GitHub Actions and support:

• Linux (for Google Colab and other Linux environments)
• Windows 10/11

If you need to clean up before building:

# Clean while preserving .c files (for wheel building)
python clean_cython.py

# Clean everything including .c files
python clean_cython.py --all