diff --git a/annif/backend/mixins.py b/annif/backend/mixins.py
index 066d5d862..2bb094641 100644
--- a/annif/backend/mixins.py
+++ b/annif/backend/mixins.py
@@ -72,9 +72,14 @@ def initialize_vectorizer(self) -> None:
                 )
 
     def create_vectorizer(
-        self, input: Iterable[str], params: dict[str, Any] = {}
+        self, input: Iterable[str], params: dict[str, Any] = None
     ) -> csr_matrix:
         self.info("creating vectorizer")
+        if params is None:
+            params = {}
+        # avoid UserWarning when overriding tokenizer
+        if "tokenizer" in params:
+            params["token_pattern"] = None
         self.vectorizer = TfidfVectorizer(**params)
         veccorpus = self.vectorizer.fit_transform(input)
         annif.util.atomic_save(
diff --git a/annif/lexical/mllm.py b/annif/lexical/mllm.py
index 37564a76d..ff8bc5894 100644
--- a/annif/lexical/mllm.py
+++ b/annif/lexical/mllm.py
@@ -223,7 +223,7 @@ def _prepare_train_index(
         self._prepare_relations(graph, vocab)
 
         self._vectorizer = CountVectorizer(
-            binary=True, tokenizer=analyzer.tokenize_words
+            binary=True, tokenizer=analyzer.tokenize_words, token_pattern=None
         )
         label_corpus = self._vectorizer.fit_transform((t.label for t in terms))