Allows disabling NLTK, adds it as a consideration for low power devices

docs/setup.rst

@@ -774,6 +774,8 @@ configuring some options in paperless can help improve performance immensely:
     OCR results.
 *   If using docker, consider setting ``PAPERLESS_WEBSERVER_WORKERS`` to
     1. This will save some memory.
+*   Consider setting ``PAPERLESS_ENABLE_NLTK`` to false, to disable the more
+    advanced language processing, which can take more memory and processing time.
 
 For details, refer to :ref:`configuration`.
 

src/documents/classifier.py

@@ -312,7 +312,7 @@ class DocumentClassifier:
         content = re.sub(r"[^\w\s]", " ", content)
 
         # If the NLTK language is supported, do further processing
-        if settings.NLTK_LANGUAGE is not None:
+        if settings.NLTK_LANGUAGE is not None and settings.NLTK_ENABLED:
 
             import nltk
 

src/paperless/settings.py

@@ -709,6 +709,10 @@ ENABLE_UPDATE_CHECK = os.getenv("PAPERLESS_ENABLE_UPDATE_CHECK", "default")
 if ENABLE_UPDATE_CHECK != "default":
     ENABLE_UPDATE_CHECK = __get_boolean("PAPERLESS_ENABLE_UPDATE_CHECK")
 
+###############################################################################
+# Machine Learning                                                            #
+###############################################################################
+
 
 def _get_nltk_language_setting(ocr_lang: str) -> Optional[str]:
     """
@@ -735,4 +739,6 @@ def _get_nltk_language_setting(ocr_lang: str) -> Optional[str]:
     return iso_code_to_nltk.get(ocr_lang, None)
 
 
+NLTK_ENABLED: Final[bool] = __get_boolean("PAPERLESS_ENABLE_NLTK", "yes")
+
 NLTK_LANGUAGE: Optional[str] = _get_nltk_language_setting(OCR_LANGUAGE)