Logo Lanfrica

goobolabs/somali-language-standard

Domain:

natural language processing

Record type:

dataset
Creator:
goo
Host:
The open, machine-readable standard for the Somali language — orthography, grammar, terminology, translation, and AI/benchmark resources, as a versioned, citable RFC-style catalog. # Somali Language Standard (SLS) **The open, machine-readable standard for the Somali language.** SLS defines grammar, orthography, terminology, translation guidance, style conventions, and AI/benchmark resources for Somali — as a versioned, normative, citable standard that any system, human or machine, can implement against. > **Status:** The Milestone 2 proposal package is complete. SLS-0000 through > SLS-0005 are represented at `Proposed`, including the eight-topic SLS-0003 > Grammar set, and the 145-file resources evidence baseline is complete. The > SLS-0002 through SLS-0005 public-comment periods opened in > pull request #7 > on 2026-08-23 and cannot close before 2026-09-06. No standard is `Stable`. See the > Roadmap and M2 readiness record. --- ## Vision SLS aims to be to Somali what an RFC series or a language specification (ECMA-262, Unicode Technical Reports) is to a technical domain: the canonical reference that AI vendors, translation systems, NLP researchers, educators, and government bodies point to when they need an authoritative answer about the Somali language. It is **not a corpus dump**. It is a standard with a compliance surface: every fact SLS asserts — a word, a grammar rule, a terminology mapping — exists as a structured, schema-validated record with provenance. ## Goals - **Normative specifications** for Somali orthography, grammar, style, and translation, published through an RFC-style lifecycle. - **A governed terminology program** that coins and standardizes Somali vocabulary for modern domains (AI, medicine, law, and 17 more) where standard terms often do not yet exist. - **Machine-first data**: a curated lexicon, parallel translation pairs, and corpora — all JSONL, all schema-validated, all carrying provenance. - **AI resources and benchmarks**: prompts, instruction datasets, and evaluation suites kept strictly separate to prevent contamination. - **A numbered standards catalog** (`SLS-XXXX`) so adopters can make precise, falsif …