Fixes the download and usage of the downloaded data

src/documents/classifier.py

@@ -306,6 +306,12 @@ class DocumentClassifier:
         from nltk.corpus import stopwords
         from nltk.stem import SnowballStemmer
 
+        import nltk
+
+        # Not really hacky, since it isn't private and is documented, but
+        # set the search path for NLTK data to the single location it should be in
+        nltk.data.path = [settings.NLTK_DIR]
+
         if self.stemmer is None:
             self.stemmer = SnowballStemmer("english")
 

src/paperless/settings.py

@@ -84,6 +84,8 @@ THUMBNAIL_DIR = os.path.join(MEDIA_ROOT, "documents", "thumbnails")
 
 DATA_DIR = __get_path("PAPERLESS_DATA_DIR", os.path.join(BASE_DIR, "..", "data"))
 
+NLTK_DIR = os.path.join(DATA_DIR, "nltk")
+
 TRASH_DIR = os.getenv("PAPERLESS_TRASH_DIR")
 
 # Lock file for synchronizing changes to the MEDIA directory across multiple