What's the data flow for a typical request?

I can only answer questions about the sjmoran/satire-classifier repository.

Based on the repository data provided, this is a machine learning classifier project rather than a request-handling service. There isn't a typical "request" data flow in the sense of a web server or API.

Instead, the project follows a batch training and evaluation workflow:

1. Training Phase: The classifier trains on documents from a training directory with corresponding labels

2. Feature Extraction: Multiple feature types are computed from documents:

   • Unigram counts (filtered by Chi2 score)
   • Punctuation and capitalization counts
   • Sentiment polarity counts
   • Intensifier/interjection word counts
   • Word2vec document embeddings

3. Model Training: Two parallel Naive Bayes models are trained:

   • Multinomial NB on discrete count-based features
   • Gaussian NB on continuous word2vec embeddings

4. Probability Combination: The probability outputs from both models are combined as a 4-dimensional feature space for a final Gaussian NB classifier

5. Testing Phase: The trained model is applied to a test dataset [README.md:L99-L109]

The entry point is train_classifier.py with required arguments for training/testing directories, label files, and output paths [README.md:L99-L109].

However, the repository data doesn't include actual source code showing the implementation details of this pipeline. Would you like me to search for specific code files?

FOLLOW_UPS:

• What features does the classifier extract from documents?
• How are the Multinomial and Gaussian models combined?
• What are the command-line arguments for training?