Greenlens/audits/semantic-search/README.md

Semantic Search Audit

Generated: 2026-03-12T14:15:29.567Z

Files:

• summary.json: category counts and suggested audit order
• all-plants-categories.csv: one row per plant with its full category list
• master.csv: all category assignments with blank evidence columns
• suspicious.csv: entries that require elevated review based on rule flags
• categories/*.csv: per-category audit sheets

Suggested audit order:

• pet_friendly (15)
• air_purifier (12)
• medicinal (32)
• low_light (16)
• bright_light (42)
• sun (172)
• easy (132)
• high_humidity (51)
• hanging (34)
• tree (59)
• large (35)
• patterned (31)
• flowering (157)
• succulent (46)

Workflow:

1. Review one category CSV at a time.
2. Fill audit_status, evidence_source, evidence_url, and notes.
3. Apply only high-confidence source-tag corrections to the lexicon batch files.
4. Rebuild the server catalog from batches after source edits.