Community standards

Where these apply:

Telegram, Discord, the mailing list and the LinkedIn group
GitHub issues, pull requests and code review
Workshops, seminars, conference gatherings and social events around them
Anywhere someone is representing EthioNLP, including their own social media when posting as a member

Values

Ethiopians lead the work on Ethiopian languages

Speakers of a language should own the research on it rather than supply data for someone else's. Collaboration from anywhere is welcome on that footing.

Contributions come from a range of backgrounds

Speakers of a language, linguists, students and senior researchers each contribute different knowledge. Students are expected to take part rather than observe.

Work in the open

Datasets, models, code and negative results are published where the licence allows, and the licence is chosen so that it does.

Communities decide about their own data

A language community keeps a say in what is collected from it, how it is used, and whether it is withdrawn.

Training people is part of the output

A thesis supervised, a student running their own experiments, a first paper accepted: these count alongside the datasets and the models, and usually outlast them.

Say what you actually did

Report the method, the size, the limitations and who did the work. Overstated results cost this field more than modest ones.

Conduct

Expected of members

Use welcoming and inclusive language
Respect differing viewpoints and experience
Accept constructive criticism gracefully, and give it kindly
Credit people for their work, including annotators and data collectors
Assume good faith, particularly across a language barrier

Not acceptable

Harassment of any kind, public or private
Sexualised language or imagery, and unwelcome sexual attention
Insults, derogatory comments, and personal or political attacks
Demeaning anyone for their ethnicity, language, religion, gender, disability, age, nationality or level of experience
Publishing anyone's private information without their explicit permission
Sustained disruption of discussions, talks or events

Posting in the groups

Most of this community's day-to-day life happens in the Telegram group. It is a professional space with hundreds of people in it, many of them students, and it is read by people from every part of the country.

Keep ethnic and party politics out

Ethiopia's politics reach into questions about language, and this group is not where they get settled. Discuss a language's data, orthography, speakers and status; do not argue about the standing of the people who speak it. Posts that attack an ethnic or religious group are removed and the poster is removed with them.

Do not forward unverified claims

Screenshots, casualty figures and viral audio spread fastest when they are wrong. If you cannot name the source, do not post it.

Hate speech is our research subject, not our medium

Examples for annotation or discussion belong in the dataset, behind a content warning, with the purpose stated. They do not belong pasted into a general channel.

Ask questions in the open

Beginners' questions are welcome and are not off-topic. Members are at different stages, and basic questions are expected.

No unsolicited private messages

Do not direct-message members you do not know to sell, recruit, canvass or ask them to complete a survey. Post it in the group and let people come to you.

A shared working language, alongside your own

English is the shared working language, so that members who do not read a given language can still follow a thread. Amharic, Afaan Oromo, Tigrinya and others are welcome alongside it; a short gloss keeps the discussion accessible.

Someone else's data is not yours to share

Do not post scraped personal data, unreleased datasets, or anything a collaborator gave you in confidence.

Research ethics

These apply to work published under this community's name and to work listed on this site.

Consent before collection

People whose speech or writing is collected are told what it is for and can decline. Consent to a study is not consent to every future use of it.

Remove what identifies people

Names, phone numbers, addresses and account handles come out of a corpus before it is released, and the removal is documented.

The people who supply the language are collaborators

Annotation, transcription and collection are work, and the people doing them are part of the project rather than a resource it draws on. Name them where they wish to be named, and where a project has funding, pay at a rate that is fair where they live rather than what the local market will bear.

License so the work can be reused

Release under a licence that permits reuse, and check that you hold the rights to what you are releasing.

Say what a model should not be used for

A model trained on a small corpus of one dialect will fail elsewhere. Write that down where the people deploying it will read it.

Reporting a problem

Email ethionlp@googlegroups.com. You do not have to be the person it happened to.

What to do

Email the address above. Any of the maintainers can receive a report; say if you would rather it went to one of them in particular.
Include what happened, where, and roughly when. A screenshot helps but is not required.
You will get an acknowledgement within a few days and be told what came of it, unless you ask not to hear about it again.

What we undertake

Your name is not passed to the person you are reporting without your agreement.
Reporting something in good faith never counts against you, even if the report is not upheld.
Responses range from a private word, through removal of a post, to removal from the group and from events. Serious cases skip the earlier steps.

If your report concerns a maintainer, send it to any other maintainer directly; they will handle it without that person involved.

The conduct rules adapt the Contributor Covenant 1.4, the same text Masakhane uses. The values are adapted from Masakhane's community values. Contributor Covenant 1.4 · Masakhane community documents