Merge pull request #931 from mvalentsev/fix/i18n-entity-metadata

fix: use i18n candidate patterns for entity extraction in miner and palace
2026-04-16 15:54:01 -03:00
parent c5e249bba8 cde0f5b9e7
commit 55a004fe1e
4 changed files with 56 additions and 5 deletions
@@ -656,7 +656,9 @@ class EntityRegistry:
        Find capitalized words in query that aren't in registry or common words.
        These are candidates for Wikipedia research.
        """
-        candidates = re.findall(r"\b[A-Z][a-z]{2,15}\b", query)
+        from .palace import _candidate_entity_words
+
+        candidates = _candidate_entity_words(query)
        unknown = []
        for word in set(candidates):
            if word.lower() in COMMON_ENGLISH_WORDS: