Task 72f50712 (debate transcript causal claim extractor, merged 2026-04-24) processed ~80 sessions and yielded ~1,498 KG edges on 2026-04-16 (the largest single-day KG growth event). But 841 total sessions exist — ~760 remain unprocessed.
Current KG: 2,316 edges. Estimated post-completion: 16,000+ edges (~7x increase).
Step 1: Identify which sessions were already processed by task 72f50712 (check kg_edges metadata or debate_session flags).
Step 2: Apply the causal claim extraction logic from task 72f50712 to remaining sessions in batches of 25.
- For each session: parse transcript_json for mechanistic claims (A activates B, C inhibits D, etc.)
- Extract entity pairs + relation type; normalize against existing KG nodes
- Write kg_edges rows with relation_type from the claim
Step 3: Quality check — spot-check 20 edges for accuracy, verify no duplicates.
Step 4: Commit extraction logic to codebase for future reuse.
Target: ≥500 additional KG edges added (conservative; 760 sessions × avg 19 edges = ~14K potential).
Do NOT re-extract already-processed sessions. Do NOT emit edges for vague claims without specificity.
Spec: docs/planning/specs/quest_atlas_debate_kg_extraction_complete_spec.md