Contributing
This guide is a working draft. All seven chapters provide substantive recipes, worked examples, and review worksheets. The complete draft awaits coordinated editorial and technical review. The methodology addresses the reader's artifact class; the Data Snapshot study supplies its running example.
Contributions can be proposed through GitHub issues and pull requests to the AI4Data repository. The cookbook authoring guide governs structure and style.
Criteria for a recipe
- The recipe produces a defined result with an explicit check.
- It names its maturity level, practitioner skills, and indicative time.
- Explanation introduces unfamiliar concepts and shows the input before the procedure.
- A worked example uses the prepared evidence and identifies its source and limits.
- Model proposals and human decisions remain distinguishable.
- Code, when added, is imported from its downloadable source and verified on small inputs.
Evidence and scope
Research results retain their original labels and chronology. A simplified teaching transformation is identified separately from recorded outputs. Fresh model executions are replication attempts and keep their outputs separate. Claims about improvement require comparative evidence.
The original 41-concept assessment belongs to chapter 4. The restoration of panel_title during operational design belongs to chapter 5. Image reuse requires appropriate attribution and permission; public availability alone does not establish redistribution rights.
The guide develops a schema. Production extraction, layout detection, formal ontology construction, and knowledge graph infrastructure remain outside its scope.
Recipe template
The following authoring template describes the structure required for future recipes.
<Recipe title="N.M Imperative title naming the result" level="Foundational" skills="Role and relevant skills" time="Indicative estimate">
**Result:** State the checkable artifact produced by the recipe.
Explain the result and purpose, then show the input artifact, record, or decision before the steps.
1. Describe the action and its reason.
2. Introduce any code or template and explain its fields.
3. State the verification and expected outcome.
**Worked example.** Identify the input, action, output, source, and evidence status.
**What this does not do.** State the limits of the example and procedure.
</Recipe>
Review
Technical review checks evidence fidelity, identities, denominators, links, and executable behavior where implemented. Editorial review checks conceptual clarity, the learning progression, human decision rights, and empirical claims. The standards page identifies reviewed references without implying formal adoption.