Spaces:
Running on Zero
Running on Zero
Commit ·
14b11e2
1
Parent(s): fb935de
Gate precision: add nationality adjectives + remaining sentence-starters
Browse filesLive whack-a-mole tail: "Flemish" (nationality) and "Even" (starter)
were still sinking otherwise-grounded narrations. Add a bounded batch of
nationality/cultural-style adjectives and common clause-openers to
_COMMON. Invented venues and the planted Eiffel Tower still fail.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
src/discoverroute/narrate/grounding.py
CHANGED
|
@@ -107,6 +107,18 @@ _COMMON = {
|
|
| 107 |
"remember", "consider", "watch", "listen", "recall", "observe", "sense",
|
| 108 |
"perhaps", "maybe", "indeed", "history", "culture", "tradition", "legend",
|
| 109 |
"myth", "legends", "myths", "everywhere", "somewhere", "anywhere",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 110 |
}
|
| 111 |
|
| 112 |
_TOKEN_RE = re.compile(r"[A-Za-zÀ-ÖØ-öø-ÿ][A-Za-zÀ-ÖØ-öø-ÿ0-9'’.\-]*")
|
|
|
|
| 107 |
"remember", "consider", "watch", "listen", "recall", "observe", "sense",
|
| 108 |
"perhaps", "maybe", "indeed", "history", "culture", "tradition", "legend",
|
| 109 |
"myth", "legends", "myths", "everywhere", "somewhere", "anywhere",
|
| 110 |
+
"even", "still", "yet", "once", "twice", "often", "always", "never",
|
| 111 |
+
"sometimes", "nowhere", "elsewhere", "wherever", "whenever", "however",
|
| 112 |
+
"instead", "throughout", "tonight", "yesterday", "tomorrow", "everyday",
|
| 113 |
+
# nationality / language / cultural-style adjectives — capitalized but never
|
| 114 |
+
# standalone venues (a guide writes "a Flemish façade", "Moorish arches").
|
| 115 |
+
"flemish", "dutch", "belgian", "swiss", "austrian", "portuguese", "greek",
|
| 116 |
+
"turkish", "russian", "polish", "hungarian", "czech", "scandinavian",
|
| 117 |
+
"nordic", "norse", "celtic", "gaelic", "moorish", "byzantine", "ottoman",
|
| 118 |
+
"persian", "arab", "arabic", "hebrew", "jewish", "korean", "thai",
|
| 119 |
+
"vietnamese", "mexican", "brazilian", "cuban", "african", "asian", "slavic",
|
| 120 |
+
"baltic", "iberian", "andalusian", "tuscan", "sicilian", "bavarian",
|
| 121 |
+
"prussian", "mughal", "colonial", "etruscan", "mediterranean",
|
| 122 |
}
|
| 123 |
|
| 124 |
_TOKEN_RE = re.compile(r"[A-Za-zÀ-ÖØ-öø-ÿ][A-Za-zÀ-ÖØ-öø-ÿ0-9'’.\-]*")
|