Adding a language

Teaching the parser to read language names written in a new language means adding one wordlist. No code changes are required. LANGS is discovered from the filesystem at import time.

Steps

1. Create the directory and file:

   ovos_lang_parser/res/<lang>/langs.json
   

   where <lang> is the BCP-47 code of the language the names are written in (for example sv for Swedish).

2. Fill in {code: name} pairs. Key each entry by the BCP-47 code of the target language; the value is that language's name in <lang>:

   {
     "en": "engelska",
     "de": "tyska",
     "pt": "portugisiska",
     "pt-br": "brasiliansk portugisiska"
   }
   

   • The first name for a code is canonical. It is what pronounce_lang returns. List the most natural spelling first.
   • Provide both base and regional tags where they differ (pt and pt-br). A regional tag with no entry falls back to the base name automatically.

3. Add spelling variants with templates when several forms should all resolve to one code. Use (a|b) alternation in a single string value:

   { "el": "(nygrekiska|grekiska)" }
   

   Both "nygrekiska" and "grekiska" become accepted names for el. Alternations can appear anywhere in the string and nest.

That is the whole contribution. Drop the file in, and the new code appears in LANGS automatically.

Verifying your wordlist

import ovos_lang_parser
from ovos_lang_parser import extract_langcode, pronounce_lang, get_lang_data

assert "<lang>" in ovos_lang_parser.LANGS

# names resolve back to codes
assert extract_langcode("<a name from your file>", "<lang>")[0] == "<expected code>"

# codes resolve back to your canonical names
print(pronounce_lang("de", "<lang>"))

# inspect the whole table
print(len(get_lang_data("<lang>")), "names loaded")

Conventions

• Codes are normalized to modern lowercase form on load, so you may use legacy aliases (iw, jw, mo). They merge onto the current code (he, jv, ro). Prefer the modern code when authoring.
• Keep the file UTF-8 and valid JSON. The loader reads it directly.
• Names are matched case-insensitively at runtime, so casing in the file is only cosmetic. Keep it natural for pronounce_lang output.

---

← Coverage · Home