chinese-herbal-sentiment 0.1.0

Chinese Herbal Medicine E-commerce Sentiment Analysis System

Chinese Herbal Medicine Sentiment Analysis System

A comprehensive Natural Language Processing (NLP) toolkit specifically designed for analyzing customer reviews and evaluating supply chain quality in Chinese herbal medicine e-commerce platforms.

🎯 Features

🔍 Sentiment Analysis

• Dictionary-based Analysis: Traditional sentiment analysis using Chinese sentiment dictionaries
• Machine Learning Models: SVM, Naive Bayes, and Logistic Regression classifiers
• Deep Learning Models: LSTM, TextCNN, and BERT-based sentiment analysis
• Graph-based Analysis: TextRank algorithm for sentiment analysis

🔑 Keyword Extraction

• TF-IDF: Term Frequency-Inverse Document Frequency for keyword extraction
• TextRank: Graph-based algorithm for keyword ranking
• LDA: Latent Dirichlet Allocation for topic-based keyword extraction

📊 Supply Chain Evaluation

• Multi-dimensional Analysis: Upstream (raw materials), midstream (processing), downstream (distribution)
• Quality Metrics: Comprehensive evaluation of supply chain quality indicators
• Visualization: Rich visualizations for analysis results

🛠️ Utility Features

• Data Processing: Efficient handling of large-scale review datasets
• Visualization Tools: Comprehensive plotting and charting capabilities
• Command-line Interface: Easy-to-use CLI for batch processing
• Modular Design: Flexible and extensible architecture

🚀 Quick Start

Installation

# Basic installation
pip install chinese-herbal-sentiment

# With deep learning support
pip install chinese-herbal-sentiment[deep_learning]

# With development tools
pip install chinese-herbal-sentiment[dev]

# Complete installation
pip install chinese-herbal-sentiment[all]

Basic Usage

import pandas as pd
from chinese_herbal_sentiment import SentimentAnalyzer, KeywordExtractor

# Sample data
data = pd.DataFrame({
    '评论内容': [
        '这个中药质量很好，效果不错',
        '包装很差，质量一般',
        '服务态度很好，物流快'
    ]
})

# Sentiment analysis
analyzer = SentimentAnalyzer()
sentiment_results = analyzer.analyze_all_methods(data)

# Keyword extraction
extractor = KeywordExtractor()
keyword_results = extractor.extract_all_methods(data, num_keywords=10)

print("Sentiment Results:", sentiment_results.head())
print("Keywords:", keyword_results.head())

Command Line Usage

# Analyze sentiment
chinese-herbal-analyze data/reviews.xlsx --method all --output results.csv

# Extract keywords
chinese-herbal-keywords data/reviews.xlsx --method tfidf --num_keywords 20

# Full analysis
chinese-herbal-full data/reviews.xlsx --mode all --output_dir results/

📚 Documentation

API Reference

SentimentAnalyzer

from chinese_herbal_sentiment import SentimentAnalyzer

analyzer = SentimentAnalyzer()

# Single method analysis
results = analyzer.analyze_sentiment(data, method='svm')

# All methods analysis
results = analyzer.analyze_all_methods(data)

Methods:

• dictionary: Dictionary-based sentiment analysis
• svm: Support Vector Machine classifier
• naive_bayes: Naive Bayes classifier
• logistic_regression: Logistic Regression classifier
• all: All available methods

KeywordExtractor

from chinese_herbal_sentiment import KeywordExtractor

extractor = KeywordExtractor()

# Single method extraction
keywords = extractor.extract_keywords(data, method='tfidf', num_keywords=20)

# All methods extraction
keywords = extractor.extract_all_methods(data, num_keywords=20)

Methods:

• tfidf: TF-IDF keyword extraction
• textrank: TextRank algorithm
• lda: Latent Dirichlet Allocation
• all: All available methods

Deep Learning Models

from chinese_herbal_sentiment import BERTSentimentAnalyzer, TextCNNSentimentAnalyzer

# BERT analysis
bert_analyzer = BERTSentimentAnalyzer()
bert_results = bert_analyzer.analyze_sentiment(data)

# TextCNN analysis
textcnn_analyzer = TextCNNSentimentAnalyzer()
textcnn_results = textcnn_analyzer.analyze_sentiment(data)

Advanced Usage

Custom Analysis Pipeline

from chinese_herbal_sentiment import DataAnalyzer, Visualizer

# Load and preprocess data
data_analyzer = DataAnalyzer()
data = data_analyzer.load_data('reviews.xlsx', sample_size=10000)

# Perform analysis
sentiment_results = analyzer.analyze_all_methods(data)
keyword_results = extractor.extract_all_methods(data)

# Generate visualizations
visualizer = Visualizer()
visualizer.plot_sentiment_distribution(sentiment_results, save_path='sentiment.png')
visualizer.plot_keyword_cloud(keyword_results, save_path='keywords.png')

Supply Chain Quality Evaluation

from chinese_herbal_sentiment.utils.keyword_mapping import KeywordMapper

# Map keywords to supply chain dimensions
mapper = KeywordMapper()
supply_chain_results = mapper.map_keywords_to_dimensions(keyword_results)

# Analyze quality indicators
quality_metrics = mapper.calculate_quality_metrics(supply_chain_results)

📊 Output Examples

Sentiment Analysis Results

评论内容	dictionary_sentiment	svm_sentiment	naive_bayes_sentiment	logistic_regression_sentiment
质量很好，效果不错	positive	positive	positive	positive
包装很差，质量一般	negative	negative	negative	negative
服务态度很好	positive	positive	positive	positive

Keyword Extraction Results

keyword	score	method
质量	0.85	TF-IDF
包装	0.72	TF-IDF
服务	0.68	TF-IDF
效果	0.65	TextRank
物流	0.58	TextRank

🔧 Configuration

Data Format

The package expects data in the following format:

# Excel/CSV file with columns:
data = pd.DataFrame({
    '评论内容': ['review text 1', 'review text 2', ...],
    '评分': [5, 4, 3, ...],  # Optional
    '时间': ['2024-01-01', '2024-01-02', ...],  # Optional
    '用户ID': ['user1', 'user2', ...]  # Optional
})

Model Configuration

# Custom model parameters
analyzer = SentimentAnalyzer(
    vectorizer_params={'max_features': 5000},
    classifier_params={'C': 1.0}
)

extractor = KeywordExtractor(
    tfidf_params={'max_features': 1000},
    textrank_params={'window_size': 4}
)

🧪 Testing

# Run all tests
pytest

# Run with coverage
pytest --cov=chinese_herbal_sentiment

# Run specific test file
pytest tests/test_sentiment_analysis.py

📈 Performance

Accuracy Comparison

Method	Accuracy	Precision	Recall	F1-Score
Dictionary	0.72	0.71	0.72	0.71
SVM	0.85	0.84	0.85	0.84
Naive Bayes	0.82	0.81	0.82	0.81
Logistic Regression	0.87	0.86	0.87	0.86
BERT	0.91	0.90	0.91	0.90
TextCNN	0.89	0.88	0.89	0.88

Processing Speed

• Small dataset (< 1K reviews): ~1-2 seconds
• Medium dataset (1K-10K reviews): ~10-30 seconds
• Large dataset (> 10K reviews): ~2-5 minutes

🤝 Contributing

We welcome contributions! Please see our Contributing Guidelines for details.

Development Setup

# Clone the repository
git clone https://github.com/chenxingqiang/chinese-herbal-sentiment.git
cd chinese-herbal-sentiment

# Install in development mode
pip install -e .[dev]

# Run tests
pytest

# Format code
black chinese_herbal_sentiment tests

# Lint code
flake8 chinese_herbal_sentiment tests

📄 License

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

🙏 Acknowledgments

• Research Foundation: Based on master's thesis research on Chinese herbal medicine e-commerce supply chain quality evaluation
• Open Source Libraries: Built on top of scikit-learn, transformers, PyTorch, and other excellent open-source projects
• Academic Community: Inspired by research in sentiment analysis and supply chain management

📞 Support

• Documentation: GitHub Wiki
• Issues: GitHub Issues
• Email: [e-mail]

🔄 Changelog

v0.1.0 (2024-12-XX)

• Initial release
• Basic sentiment analysis (dictionary, SVM, Naive Bayes, Logistic Regression)
• Keyword extraction (TF-IDF, TextRank, LDA)
• Deep learning models (BERT, TextCNN, TextRank)
• Command-line interface
• Comprehensive documentation and examples

---

Note: This package is designed specifically for Chinese herbal medicine e-commerce review analysis and supply chain quality evaluation. For general sentiment analysis tasks, consider using more general-purpose NLP libraries.