20 lines
535 B
Python
20 lines
535 B
Python
"""Tests for filtering noisy OCR fragments."""
|
|
|
|
from kb.ingest.quality import has_minimum_content
|
|
|
|
|
|
def test_short_ocr_fragments_are_rejected():
|
|
assert not has_minimum_content('"')
|
|
assert not has_minimum_content("B")
|
|
assert not has_minimum_content("12")
|
|
|
|
|
|
def test_short_identifiers_and_real_text_are_retained():
|
|
assert has_minimum_content("BID")
|
|
assert has_minimum_content("00:15")
|
|
assert has_minimum_content("Useful text")
|
|
|
|
|
|
def test_filter_can_be_disabled():
|
|
assert has_minimum_content("B", min_alnum=0)
|