survey-seed-harvest

Survey Seed Harvest

Bootstrap taxonomy seeds from existing survey/review papers inside your retrieved set.

This is an accelerator for the early structure stage: it should make taxonomy-builder easier, not replace it.

Inputs

Outputs

Workflow (heuristic)

Uses: papers/papers_dedup.jsonl .

Find likely survey/review papers:
title/abstract contains “survey”, “review”, “systematic”, “meta-analysis”
Extract candidate topic terms and group them into:
~4–10 top-level nodes (“chapters”)
2–6 children per node (mappable leaves)
Write short, actionable descriptions:
what belongs here / what does not
(optional) list 2–5 representative titles as seeds
Treat the result as a starting point:
pass it to taxonomy-builder for domain-meaningful rewriting and scope alignment.

Quality checklist

outline/taxonomy.yml exists and is valid YAML.
Taxonomy has at least 2 levels (children used) and every node has a description.
Avoid generic placeholder nodes like “Overview/Benchmarks/Open Problems” unless they are truly content-based for your domain.

Script (optional helper)

Quick Start

python .codex/skills/survey-seed-harvest/scripts/run.py --help
python .codex/skills/survey-seed-harvest/scripts/run.py --workspace <workspace_dir>

All Options

Examples

More conservative term selection:
python .codex/skills/survey-seed-harvest/scripts/run.py --workspace <ws> --top-k 80 --min-freq 3

Notes

This helper is keyword-based; treat the output as seeds and refine with taxonomy-builder .

Troubleshooting

Issue: no survey/review papers are detected in the set

Fix:

Broaden retrieval (add “survey”, “review”, “benchmark” variants) or manually seed a few known surveys, then rerun.

Issue: taxonomy seeds look like generic buckets

Fix:

Keep seeds concrete (named methods/benchmarks/tasks) and rely on taxonomy-builder to rewrite under the actual scope.

Source Transparency