admin/Spanish
1
0
Fork 0
Code Issues 3 Pull Requests Actions Packages Projects Releases Wiki Activity

Issue #32 cleanup — drop the last 5 mis-oriented vocab pairs

Browse Source 
Two small fixes after the LLM-vision pass:

1. merge_pdf_into_book.py — when the LLM classifies an image as 'hybrid'
   but extracts zero pairs (e.g., a conjugation table whose only English
   text is on the section header that was excluded by the prompt rules),
   respect that decision instead of falling through to the bbox/heuristic
   pipeline. Previously: 1 chapter-2 estar conjugation table generated
   4 bad pairs from the heuristic fallback.

2. fix_vocab.py language_score — recognize Spanish present-perfect
   ('he tenido', 'He andado por este pueblo') as Spanish. The classifier
   was treating the auxiliary 'he'/'has'/'ha' as English subject pronouns,
   producing false-positive mis-orientation flags on 4 chapter-15/20/23
   present-perfect example tables.

Result: mis-oriented vocab pairs across the book go from 5 → 0.
textbookDataVersion bumped to 14.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Trey T
2026-05-03 18:52:53 -05:00
parent f368c24ad6
commit 05a0cc0d17
5 changed files with 21 additions and 223 deletions
Download Patch File Download Diff File Expand all files Collapse all files
Conjuga/Conjuga/Services/DataLoader.swift
+1 -1
View File
@@ -6,7 +6,7 @@ actor DataLoader {
static let courseDataVersion = 7
static let courseDataKey = "courseDataVersion"
static let textbookDataVersion = 13
static let textbookDataVersion = 14
static let textbookDataKey = "textbookDataVersion"
/// Quick check: does the DB need seeding or course data refresh?
Conjuga/Conjuga/textbook_data.json
+1 -1
View File
File diff suppressed because one or more lines are too long
Conjuga/Conjuga/textbook_vocab.json
-216
View File
@@ -1297,78 +1297,6 @@
"section": "Subject Pronouns",
"sourceImage": "f0014-01.jpg"
},
{
"front": "vosotros estáis",
"back": "he is",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Estar (to be)",
"sourceImage": "f0015-01.jpg"
},
{
"front": "ellos están",
"back": "she is",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Estar (to be)",
"sourceImage": "f0015-01.jpg"
},
{
"front": "ellas están",
"back": "you are",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Estar (to be)",
"sourceImage": "f0015-01.jpg"
},
{
"front": "Uds. están",
"back": "nosotros estamos",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Estar (to be)",
"sourceImage": "f0015-01.jpg"
},
{
"front": "La comida está buena.",
"back": "The meal is (tastes) good.",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Estar (to be)",
"sourceImage": "f0016-02.jpg"
},
{
"front": "El pescado está delicioso.",
"back": "The fish is (tastes) delicious.",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Estar (to be)",
"sourceImage": "f0016-02.jpg"
},
{
"front": "La sopa está sabrosa.",
"back": "The soup is (tastes) delicious",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Estar (to be)",
"sourceImage": "f0016-02.jpg"
},
{
"front": "Ella está hermosa hoy.",
"back": "She is (looks) pretty today.",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Estar (to be)",
"sourceImage": "f0016-02.jpg"
},
{
"front": "Él está guapo.",
"back": "He is (looks) handsome.",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Estar (to be)",
"sourceImage": "f0016-02.jpg"
},
{
"front": "¿cómo?",
"back": "how?",
@@ -1513,150 +1441,6 @@
"section": "Adjectives",
"sourceImage": "f0017-03.jpg"
},
{
"front": "La casa es roja.",
"back": "The house is red.",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Ser (to be)",
"sourceImage": "f0019-02.jpg"
},
{
"front": "El libro es azul.",
"back": "The book is blue.",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Ser (to be)",
"sourceImage": "f0019-02.jpg"
},
{
"front": "Los carros son viejos.",
"back": "The cars are old.",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Ser (to be)",
"sourceImage": "f0019-02.jpg"
},
{
"front": "Somos simpáticos.",
"back": "We are nice.",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Ser (to be)",
"sourceImage": "f0019-02.jpg"
},
{
"front": "¿ Es la flor amarilla?",
"back": "Is the flower yellow?",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Ser (to be)",
"sourceImage": "f0019-02.jpg"
},
{
"front": "El vino es de Portugal.",
"back": "The wine is from Portugal",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Ser (to be)",
"sourceImage": "f0020-01.jpg"
},
{
"front": "La cerveza es de México.",
"back": "The beer is from Mexico.",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Ser (to be)",
"sourceImage": "f0020-01.jpg"
},
{
"front": "El café es de Brazil.",
"back": "The coffee is from Brazil.",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Ser (to be)",
"sourceImage": "f0020-01.jpg"
},
{
"front": "Somos amigos.",
"back": "We are friends.",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Ser (to be)",
"sourceImage": "f0020-02.jpg"
},
{
"front": "José y Eduardo son hermanos.",
"back": "Joe and Ed are brothers",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Ser (to be)",
"sourceImage": "f0020-02.jpg"
},
{
"front": "Pablo es español.",
"back": "Paul is Spanish.",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Ser (to be)",
"sourceImage": "f0020-02.jpg"
},
{
"front": "¿ Eres tú cubano?",
"back": "Are you Cuban?",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Ser (to be)",
"sourceImage": "f0020-02.jpg"
},
{
"front": "Fila es católica.",
"back": "She is Catholic",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Ser (to be)",
"sourceImage": "f0020-02.jpg"
},
{
"front": "La mesa es de madera.",
"back": "The table is of wood.",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Ser (to be)",
"sourceImage": "f0020-03.jpg"
},
{
"front": "La bolsa es de plástico.",
"back": "The bag is of plastic.",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Ser (to be)",
"sourceImage": "f0020-03.jpg"
},
{
"front": "Los zapatos son de cuero.",
"back": "The shoes are of leather:",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Ser (to be)",
"sourceImage": "f0020-03.jpg"
},
{
"front": "Las ventanas son de vidrio.",
"back": "The windows are of glass",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Ser (to be)",
"sourceImage": "f0020-03.jpg"
},
{
"front": "La casa es de piedra.",
"back": "The house is of stone.",
"chapter": 2,
"chapterTitle": "Estar, Ser, and Subject Pronouns",
"section": "Ser (to be)",
"sourceImage": "f0020-03.jpg"
},
{
"front": "la alcoba",
"back": "the bedroom",
Conjuga/Scripts/textbook/fix_vocab.py
+13 -2
View File
@@ -46,6 +46,9 @@ SPANISH_ARTICLES = {"el", "la", "los", "las", "un", "una", "unos", "unas"}
ENGLISH_STARTERS = {"the", "a", "an", "to", "my", "his", "her", "our", "their"}
HABER_FORMS = {"he", "has", "ha", "hemos", "habéis", "han"}
def language_score(s: str) -> "tuple[int, int]":
"""Return (es_score, en_score) for a string."""
es = 0
@@ -56,9 +59,17 @@ def language_score(s: str) -> "tuple[int, int]":
if not words:
return (es, en)
first = words[0].strip(",.;:")
if first in SPANISH_ARTICLES:
second = words[1].strip(",.;:") if len(words) > 1 else ""
# Spanish present-perfect ("he tenido", "Ha andado") starts with a haber
# form followed by an -ado/-ido past participle. Recognise this pattern
# before the bare-pronoun check so "he" isn't mistaken for English "he".
if first in HABER_FORMS and (
second.endswith(("ado", "ido", "to", "cho", "sto", "esto"))
):
es += 3
elif first in SPANISH_ARTICLES:
es += 2
if first in ENGLISH_STARTERS:
elif first in ENGLISH_STARTERS:
en += 2
# Spanish-likely endings on later words
for w in words:
Conjuga/Scripts/textbook/merge_pdf_into_book.py
+6 -3
View File
@@ -307,10 +307,13 @@ def main() -> None:
# Choose pair source. For reference_only (Spanish-only tables)
# we deliberately produce no cards — the UI will fall back to
# rendering the flat OCR lines as a reference list.
if llm_kind == "reference_only":
# rendering the flat OCR lines as a reference list. Same for
# hybrid images where the LLM determined no genuine pair rows
# exist (e.g. estar conjugations with English glosses on the
# header row only).
if llm_kind == "reference_only" or (llm_kind == "hybrid" and not llm_pairs):
cards_for_block = []
pair_source = "llm-reference"
pair_source = "llm-no-pairs"
elif llm_pairs:
cards_for_block = [
{"front": p["es"], "back": p["en"]}