feat: ability to skip English language specific checks with env var (#224)

* add language env var

* update docs

* version and bump change log

Author: MthwRobinson

Diff for: CHANGELOG.md

@@ -1,7 +1,8 @@
-## 0.4.9-dev1
+## 0.4.9-dev2
 
 * Added ingest modules and s3 connector
 * Default to `url=None` for `partition_pdf` and `partition_image`
+* Add ability to skip English specific check by setting the `UNSTRUCTURED_LANGUAGE` env var to `""`.
 
 ## 0.4.8
 

Diff for: docs/source/bricks.rst

@@ -280,6 +280,7 @@ for consideration as narrative text. The function performs the following checks
   ``non_alpha_ratio`` kwarg or the ``UNSTRUCTURED_NARRATIVE_TEXT_NON_ALPHA_RATIO`` environment variable.
   The environment variables takes precedence over the kwarg.
 * The cap ratio test does not apply to text that is all uppercase.
+* If you use the ``language=""`` kwarg or set the ``UNSTRUCTURED_LANGUAGE`` environment variable to ``""``, the function will skip the verb check and the English word check.
 
 
 Examples:
@@ -320,6 +321,8 @@ for consideration as a title. The function performs the following checks:
 * Narrative text must contain at least one English word (if ``language`` is set to "en")
 * If a title contains more than one sentence that exceeds a certain length, it cannot be a title. Sentence length threshold is controlled by the ``sentence_min_length`` kwarg and defaults to 5.
 * If a segment of text ends in a comma, it is not considered a potential title. This is to avoid salutations like "To My Dearest Friends," getting flagged as titles.
+* If you use the ``language=""`` kwarg or set the ``UNSTRUCTURED_LANGUAGE`` environment variable to ``""``, the function will skip the English word check.
+
 
 
 Examples:

Diff for: test_unstructured/partition/test_text_type.py

@@ -50,6 +50,31 @@ def test_is_possible_narrative_text(text, expected, monkeypatch):
     assert is_possible_narrative is expected
 
 
+def test_text_type_handles_non_english_examples():
+    narrative_text = "Я говорю по-русски. Вы тоже?"
+    title = "Риски"
+
+    assert text_type.is_possible_narrative_text(narrative_text, language="en") is False
+    assert text_type.is_possible_narrative_text(narrative_text, language="") is True
+
+    assert text_type.is_possible_narrative_text(title, language="en") is False
+    assert text_type.is_possible_narrative_text(title, language="") is False
+
+    assert text_type.is_possible_title(title, language="en") is False
+    assert text_type.is_possible_title(title, language="") is True
+
+
+def test_text_type_handles_non_english_examples_with_env_var(monkeypatch):
+    monkeypatch.setenv("UNSTRUCTURED_LANGUAGE", "")
+
+    narrative_text = "Я говорю по-русски. Вы тоже?"
+    title = "Риски"
+
+    assert text_type.is_possible_narrative_text(narrative_text) is True
+    assert text_type.is_possible_narrative_text(title) is False
+    assert text_type.is_possible_title(title) is True
+
+
 @pytest.mark.parametrize(
     "text, expected",
     [

Diff for: unstructured/__version__.py

@@ -1 +1 @@
-__version__ = "0.4.9-dev1"  # pragma: no cover
+__version__ = "0.4.9-dev2"  # pragma: no cover

Diff for: unstructured/partition/text_type.py

@@ -49,6 +49,7 @@ def is_possible_narrative_text(
         logger.debug(f"Not narrative. Text is all numeric:\n\n{text}")
         return False
 
+    language = os.environ.get("UNSTRUCTURED_LANGUAGE", language)
     if language == "en" and not contains_english_word(text):
         return False
 
@@ -67,7 +68,7 @@ def is_possible_narrative_text(
     if under_non_alpha_ratio(text, threshold=non_alpha_threshold):
         return False
 
-    if (sentence_count(text, 3) < 2) and (not contains_verb(text)):
+    if (sentence_count(text, 3) < 2) and (not contains_verb(text)) and language == "en":
         logger.debug(f"Not narrative. Text does not contain a verb:\n\n{text}")
         return False
 
@@ -118,6 +119,7 @@ def is_possible_title(
     if text.endswith(","):
         return False
 
+    language = os.environ.get("UNSTRUCTURED_LANGUAGE", language)
     if language == "en" and not contains_english_word(text):
         return False