| CountryCode-Decide |
| Copyright 2026 Ali Ildan |
|
|
| This product is licensed under the Apache License, Version 2.0 (see LICENSE). |
|
|
| Base model: |
| - jhu-clsp/mmBERT-small: MIT License, https://huggingface.co/jhu-clsp/mmBERT-small |
|
|
| Reference data used to build the lexicon and the training strings (not redistributed as datasets): |
| - Unicode CLDR (via Babel): Unicode License v3, https://www.unicode.org/license.txt |
| - Wikidata: CC0 1.0, https://www.wikidata.org/wiki/Wikidata:Licensing |
| - GeoNames (admin1 and cities15000 names, used to generate training strings): CC BY 4.0, |
| https://www.geonames.org |
|
|
| Register strings (read with aggregate queries: distinct value and count, never records): |
| - Colorado (data.colorado.gov) and Connecticut (data.ct.gov) business registry open data: |
| Public Domain; training and evaluation |
| - UK Companies House: contains public sector information licensed under the Open Government |
| Licence v3.0; training and evaluation |
| - Brønnøysund Register Centre: contains data under the Norwegian licence for Open Government data |
| (NLOD) made available by the Brønnøysund Register Centre; training and evaluation |
| - Oregon (data.oregon.gov) business registry open data: Public Domain (ORS 276A.368); evaluation only |
| - New York (data.ny.gov) corporation filings: OPEN-NY terms; evaluation only |
| - GBIF parsers, countryName.tsv: Apache License 2.0, https://github.com/gbif/parsers; evaluation only |
|
|
| Generators that the code base contains but the released model does not use (they failed the |
| usefulness gate): OCR readings with microsoft/trocr-base-handwritten (MIT) on google/fonts |
| handwriting fonts (SIL OFL 1.1), and LLM paraphrases from Qwen3.5-35B-A3B (Apache-2.0). |
|
|
