Ruka hadi maudhui makuu

Licensing decision tree

A licence can only grant what the people in the data agreed to and what the sources you built on allow. This tree walks those two constraints first, then the community's wishes, and ends at one of five outcomes. Work through it before the project charter is signed, and again before release, because consent terms and source data often change during a project. The reasoning behind each licence is in Data Governance; the practical notes on each type are in Licensing.

Before you start

Have three things in front of you: the consent form your contributors signed or recorded, the licence of every dataset, corpus or model output you reused, and the community's decision on who may benefit from the data (recorded in the charter). The tree asks about each.

Work through the tree​

Printed version of the tree.

  1. Did every contributor consent to public release? No or unclear: do not release; re-consent or drop the items. Research-only consent: go to restricted release (step 6).
  2. Does the dataset include material you did not create? If unchecked: stop and list every source with its licence. If a source is share-alike, non-commercial or competition-only and cannot be removed: your licence inherits the strictest term (CC BY-SA, CC BY-NC, or remove the material).
  3. What did the community decide about benefit? Anyone may reuse with credit: step 4. Benefit must flow back: step 5. Named partners only: step 6. Some material must stay closed: step 7.
  4. Should adaptations stay open under the same terms? Yes: CC BY-SA 4.0. No: CC BY 4.0.
  5. Community licence: NOODL for a community-held corpus, Esethu when a steward will license the data and reinvest the revenue.
  6. Restricted release: gated access with a data-use agreement; publish the datasheet, schema and a consented sample openly.
  7. Split the release: keep the restricted partition closed and run the open remainder through the tree again.

After the tree​

Write the outcome into three places: the licence field of the dataset card, the licensing agreement in Licensing, and the sustainability plan's business-model section, which must respect the licence you chose. Then check the two agreements that grant the project its rights: the contributor agreement and the annotator contract must grant at least what the licence gives away, and no more.

A licence is hard to change after release. Copies already downloaded keep the terms they were released under, so a later move from CC BY to a community licence only binds new downloads. If the community may want to tighten terms later, start with the tighter licence.

Contributor
@shmuhammadd

Join the discussion

Spotted an error, have a question, or want to share what worked on a real project? Sign in with GitHub to add your voice. Every thread lives in the open, powered by GitHub Discussions.

Loading discussion…
Masakhane African Languages Hub logo

Supported by

Masakhane African Languages Hub

The AfriPlaybook is supported by the Masakhane African Languages Hub, a pan-African initiative building open, culturally grounded datasets for African languages. We are grateful for its support.

About the Hub

In collaboration with

  • Bayero University, Kano
  • Bahir Dar University
  • HausaNLP
  • EthioNLP