diff --git a/neofuzz/process.py b/neofuzz/process.py
index c049575..d91cadb 100644
--- a/neofuzz/process.py
+++ b/neofuzz/process.py
@@ -76,7 +76,8 @@ def index(self, options: Iterable[str]):
         n_dimensions = vectors.shape[1]
         self.nearest_neighbours = AnnoyIndex(n_dimensions, self.metric)
         for i_option, vector in enumerate(vectors):
-            self.nearest_neighbours.add_item(i_option, vector)
+            array = vector.toarray()[0,:]
+            self.nearest_neighbours.add_item(i_option, array)
         self.nearest_neighbours.build(self.n_trees, n_jobs=self.n_jobs)
 
     def query(
@@ -115,8 +116,9 @@ def query(
         indices = []
         distances = []
         for query_vector in search_matrix:
+            query_array = query_vector.toarray()[0,:]
             ind, dist = self.nearest_neighbours.get_nns_by_vector(
-                query_vector, limit, include_distances=True
+                query_array, limit, include_distances=True
             )
             indices.append(ind)
             distances.append(dist)
@@ -236,7 +238,11 @@ def ratio(self, s1: str, s2: str) -> int:
                 " please index before getting ratios."
             )
         v1, v2 = self.vectorizer.transform([s1, s2])
-        distance = pairwise_distances(v1, v2, metric=self.metric)
+        if self.metric == 'angular':
+            metric = 'cosine'
+        else:
+            metric = self.metric
+        distance = pairwise_distances(v1, v2, metric=metric)
         distance = np.ravel(distance)[0]
         score = (1 - distance) * 100
         return int(score)
@@ -288,7 +294,7 @@ def from_disk(cls, save_dir: Union[str, Path]):
 def char_ngram_process(
     ngram_range: Tuple[int, int] = (1, 5),
     tf_idf: bool = True,
-    metric: str = "cosine",
+    metric: str = "angular",
     refine_levenshtein: bool = False,
 ) -> Process:
     """Basic character n-gram based fuzzy search process.
@@ -300,7 +306,7 @@ def char_ngram_process(
         n-grams.
     tf_idf: bool, default True
         Flag signifying whether the features should be tf-idf weighted.
-    metric: str, default 'cosine'
+    metric: str, default 'angular'
         Distance metric to use for fuzzy search.
     refine_levenshtein: bool, default None
         Indicates whether results should be refined with Levenshtein distance