Review Salience → PCA → K-means Customer Segmentation

Full pipeline from a review corpus to customer segments:

1. Score every review on every attribute (salience 0–7) → review × attribute matrix
2. Reduce attributes to latent dimensions via PCA
3. Segment reviews into customer groups via iterative K-means

Each stage is independently useful. Run only the stages the user needs.

For the Excel output format, read references/xlsx-format.md.
For PCA and K-means implementation details, read references/pca-kmeans.md.
For a worked example (safety-eyewear, 923 reviews, 30 attributes, 4 clusters), read references/worked-example.md.

Salience measures how prominently a reviewer mentions an attribute — not sentiment. Each cell is an integer 0–7: