# Doc Extraction Hybrid Classification Strategy Wrk 1188 Learning

> Sub-skill of doc-extraction: Hybrid Classification Strategy (WRK-1188 Learning).

- Skill: `vamseeachanta/doc-extraction-hybrid-classification-strategy-wrk-1188-l-2` (Agent Skill)
- Install (CLI): `npx skillmds@latest add vamseeachanta/doc-extraction-hybrid-classification-strategy-wrk-1188-l-2`
- Raw SKILL.md: https://api.skillmd.com/api/skills/vamseeachanta/doc-extraction-hybrid-classification-strategy-wrk-1188-l-2/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: vamseeachanta (https://skillmd.com/u/vamseeachanta)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/vamseeachanta/doc-extraction-hybrid-classification-strategy-wrk-1188-l-2

---


# Hybrid Classification Strategy (WRK-1188 Learning)

## Hybrid Classification Strategy (WRK-1188 Learning)


For large homogeneous collections, prefer deterministic classifiers over LLM:

| Collection | Strategy | Cost | Accuracy |
|-----------|----------|------|----------|
| ASTM (25,537 docs) | Designation prefix → discipline | $0 | 86% vs LLM |
| API/DNV/ISO (1,062) | LLM (Claude Haiku CLI) | ~$2 | Baseline |
| Unknown org (484) | LLM (Claude Haiku CLI) | ~$1 | Baseline |

**Rule**: If org has predictable title/designation patterns, write a deterministic
classifier first. Validate with 100-doc LLM sample. Accept if >85%.

See `data/document-index-pipeline` skill for full pipeline orchestration.