gector-bert-es / language-rules.json
Imperatore's picture
gector-bert-es: first public release (CC BY 4.0, base BETO)
571b34f verified
Raw History Blame Contribute Delete
1.05 kB
{
"version": 1,
"language": "es",
"_comment": "Per-bundle language facts (see C# LanguageRules / ADR-004). stripTrailingTerminator true + stripTerminators \".\": strip only the PERIOD (the Italian brittleness the flag exists for) and KEEP a trailing ? / ! — validated at eval 2026-09-06: with the default set the first ES train scored opening_punct 0/8 because the model never saw the ? that says a ¿ is missing. clauseConjunctions: connectors after a comma = independent-clause boundary for GectorAnalyzer.Correct's clause split (see the EN/IT files). 'e'/'u' are the vowel-variant forms of y/o; 'así' is the single-token head of 'así que'. Multi-word connectors (sin embargo, no obstante, por eso) can't be matched by the single-word splitter and are deliberately absent. Validate against lang/es/gold comma_context at eval.",
"stripTrailingTerminator": true,
"stripTerminators": ".",
"clauseConjunctions": ["pero", "y", "e", "o", "u", "sino", "aunque", "porque", "pues", "entonces", "luego", "mientras", "además", "así"]
}