Product Taxonomy & Graph Cleanup

I maintain a referral network that relies on a multi-thousand-item product taxonomy and a relationship graph. Right now the data shows three main pain points, with correcting miscategorized items at the top of the list. Inconsistent tags appear everywhere—22 categories and more than 200 sub-categories—so each pass you make must scale yet still catch edge cases.

You’ll work from my written rules and the issue log, using a combination of Python scripts (working with JSON and CSV exports) and careful manual review. Scripts can handle bulk renaming, tag normalization, and simple validation; your judgement comes in when a product seems to straddle two categories or when a referral link just feels off.

Deliverables
• A cleaned taxonomy file (CSV or JSON) with corrected categories and harmonised tags
• An updated relationship graph with improbable links removed or flagged
• A short change log explaining any grey-area decisions

Acceptance is met when no product appears in more than one leaf category, tag spelling is 100 % consistent, and every referral edge passes our documented business rules. I’m available for quick clarifications—ask rather than guess.