Benjamin_Boenisch/breakpilot-lehrer
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix Step 5h: reject grammar patterns in slash-IPA, convert trailing variants

Browse Source 
- Reject /.../ matches containing spaces, parens, or commas (e.g. sb/sth up)
- Second pass converts trailing /ipa2/ after [ipa1] (double pronunciation)
- Validate standalone /ipa/ at start against same reject pattern

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Benjamin Admin
2026-03-20 12:40:28 +01:00
parent 7fafd297e7
commit 04092a0a66
2 changed files with 57 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

29
klausur-service/backend/grid_editor_api.py
View File

@@ -2014,6 +2014,9 @@ async def _build_grid_core(session_id: str, session: dict) -> dict:
)
# Standalone slash IPA at start of text (headword on previous line)
_STANDALONE_SLASH_IPA_RE = re.compile(r'^/([^/]{2,})/')
# IPA between slashes never contains spaces, parentheses, or commas.
# Reject matches that look like grammar: "sb/sth up a) jdn/"
_SLASH_IPA_REJECT_RE = re.compile(r'[\s(),]')
slash_ipa_fixed = 0
for z in zones_data:
for cell in z.get("cells", []):
@@ -2025,6 +2028,10 @@ async def _build_grid_core(session_id: str, session: dict) -> dict:
nonlocal slash_ipa_fixed
headword = m.group(1)
ocr_ipa = m.group(2) # includes slashes
inner_raw = ocr_ipa.strip("/").strip()
# Reject if inner content has spaces/parens/commas (grammar)
if _SLASH_IPA_REJECT_RE.search(inner_raw):
return m.group(0)
# Strip superscript digits for lookup
clean_hw = re.sub(r'[²³¹\d]', '', headword).strip()
ipa = _lookup_ipa(clean_hw, "british") if clean_hw else None
@@ -2032,9 +2039,7 @@ async def _build_grid_core(session_id: str, session: dict) -> dict:
slash_ipa_fixed += 1
return f"{headword} [{ipa}]"
# Fallback: keep OCR IPA but convert slashes to brackets
inner = ocr_ipa.strip("/").strip()
# Strip leading ' (OCR stress marker)
inner = inner.lstrip("'").strip()
inner = inner_raw.lstrip("'").strip()
if inner:
slash_ipa_fixed += 1
return f"{headword} [{inner}]"
@@ -2042,11 +2047,27 @@ async def _build_grid_core(session_id: str, session: dict) -> dict:
new_text = _SLASH_IPA_RE.sub(_replace_slash_ipa, text)
# Second pass: convert remaining /ipa/ after [ipa] from first pass.
# Pattern: [ipa] /ipa2/ → [ipa] [ipa2] (second pronunciation variant)
_AFTER_BRACKET_SLASH = re.compile(r'(?<=\])\s*(/[^/]{2,}/)')
def _replace_trailing_slash(m: re.Match) -> str:
nonlocal slash_ipa_fixed
inner = m.group(1).strip("/").strip().lstrip("'").strip()
if _SLASH_IPA_REJECT_RE.search(inner):
return m.group(0)
if inner:
slash_ipa_fixed += 1
return f" [{inner}]"
return m.group(0)
new_text = _AFTER_BRACKET_SLASH.sub(_replace_trailing_slash, new_text)
# Handle standalone /ipa/ at start (no headword in this cell)
if new_text == text:
m = _STANDALONE_SLASH_IPA_RE.match(text)
if m:
inner = m.group(1).strip().lstrip("'").strip()
inner = m.group(1).strip()
if not _SLASH_IPA_REJECT_RE.search(inner):
inner = inner.lstrip("'").strip()
if inner:
new_text = "[" + inner + "]" + text[m.end():]
slash_ipa_fixed += 1

38
klausur-service/backend/tests/test_grid_editor_api.py
View File

@@ -823,24 +823,42 @@ class TestSlashIpaConversion:
r"(/[^/]{2,}/)"
)
_STANDALONE_SLASH_IPA_RE = re.compile(r'^/([^/]{2,})/')
_SLASH_IPA_REJECT_RE = re.compile(r'[\s(),]')
def _replace(m):
headword = m.group(1)
ocr_ipa = m.group(2)
inner_raw = ocr_ipa.strip("/").strip()
if _SLASH_IPA_REJECT_RE.search(inner_raw):
return m.group(0)
clean_hw = re.sub(r'[²³¹\d]', '', headword).strip()
ipa = _lookup_ipa(clean_hw, "british") if clean_hw else None
if ipa:
return f"{headword} [{ipa}]"
inner = ocr_ipa.strip("/").strip().lstrip("'").strip()
inner = inner_raw.lstrip("'").strip()
if inner:
return f"{headword} [{inner}]"
return m.group(0)
new_text = _SLASH_IPA_RE.sub(_replace, text)
# Second pass: trailing /ipa/ after [ipa]
_AFTER_BRACKET_SLASH = re.compile(r'(?<=\])\s*(/[^/]{2,}/)')
def _replace_trailing(m):
inner = m.group(1).strip("/").strip().lstrip("'").strip()
if _SLASH_IPA_REJECT_RE.search(inner):
return m.group(0)
if inner:
return f" [{inner}]"
return m.group(0)
new_text = _AFTER_BRACKET_SLASH.sub(_replace_trailing, new_text)
if new_text == text:
m = _STANDALONE_SLASH_IPA_RE.match(text)
if m:
inner = m.group(1).strip().lstrip("'").strip()
inner = m.group(1).strip()
if not _SLASH_IPA_REJECT_RE.search(inner):
inner = inner.lstrip("'").strip()
if inner:
new_text = "[" + inner + "]" + text[m.end():]
return new_text
@@ -865,21 +883,19 @@ class TestSlashIpaConversion:
assert "/und/" not in result
def test_sb_sth_not_matched(self):
"""sb/sth should NOT be treated as IPA (too short or grammar)."""
"""sb/sth should NOT be treated as IPA (contains space/parens)."""
text = "(tie sb/sth up) jdn/etwas anbinden"
result = self._run_step_5h(text)
# /sth up) jdn/ has length > 2 but the headword is "sb" which is
# not a real word — the regex would match, but "sb" won't be in dict
# and the inner text would contain grammar, not IPA.
# Key assertion: "jdn/etwas" is not corrupted
assert "etwas" in result
# The inner content "sth up) jdn" has spaces and parens → rejected
assert result == text # unchanged
def test_double_ipa(self):
"""times/taimz/ /tamz/ → both converted."""
def test_double_ipa_both_converted(self):
"""times/taimz/ /tamz/ → times [tˈaɪmz] [tamz] (both converted)."""
result = self._run_step_5h("times/taimz/ /tamz/ Präp")
assert "[tˈaɪmz]" in result
# Second /tamz/ is standalone after first replacement
assert "[tamz]" in result
assert "/taimz/" not in result
assert "/tamz/" not in result
def test_standalone_slash_ipa_at_start(self):
"""/tam/ Nomen → [tam] Nomen (no headword in cell)."""