diff --git a/rampwf/hyperopt/hyperopt.py b/rampwf/hyperopt/hyperopt.py
index 73383138..51a5654d 100644
--- a/rampwf/hyperopt/hyperopt.py
+++ b/rampwf/hyperopt/hyperopt.py
@@ -250,7 +250,7 @@ def parse_hyperparameters(module_path, workflow_element_name):
 def parse_all_hyperparameters(module_path, workflow):
     """Parse hyperparameters in a submission.
 
-    Load all the the modules, take all Hyperparameter objects, and set the name
+    Load all the modules, take all Hyperparameter objects, and set the name
     of each to the name of the hyperparameter the user chose and the workflow
     element name of each to the corresponding workflow_element_name.
 
diff --git a/rampwf/score_types/brier_score.py b/rampwf/score_types/brier_score.py
index 7509276c..e461cebc 100644
--- a/rampwf/score_types/brier_score.py
+++ b/rampwf/score_types/brier_score.py
@@ -15,7 +15,7 @@ def __init__(self, name='brier_score', precision=3):
     def score_function(self, ground_truths, predictions):
         """A hybrid score.
 
-        It tests the the predicted _probability_ of the second class
+        It tests the predicted _probability_ of the second class
         against the true _label index_ (which is 0 if the first label is the
         ground truth, and 1 if it is not, in other words, it is the
         true probability of the second class). Thus we have to override the
@@ -42,7 +42,7 @@ def __init__(self, name='brier_score', precision=3):
     def score_function(self, ground_truths, predictions):
         """A hybrid score.
 
-        It tests the the predicted _probability_ of the second class
+        It tests the predicted _probability_ of the second class
         against the true _label index_ (which is 0 if the first label is the
         ground truth, and 1 if it is not, in other words, it is the
         true probability of the second class). Thus we have to override the
@@ -77,7 +77,7 @@ def __init__(self, name='brier_score', precision=3,
     def score_function(self, ground_truths, predictions):
         """A hybrid score.
 
-        It tests the the predicted _probability_ of the second class
+        It tests the predicted _probability_ of the second class
         against the true _label index_ (which is 0 if the first label is the
         ground truth, and 1 if it is not, in other words, it is the
         true probability of the second class). Thus we have to override the
@@ -122,7 +122,7 @@ def __init__(self, name='brier_score', precision=3,
     def score_function(self, ground_truths, predictions):
         """A hybrid score.
 
-        It tests the the predicted _probability_ of the second class
+        It tests the predicted _probability_ of the second class
         against the true _label index_ (which is 0 if the first label is the
         ground truth, and 1 if it is not, in other words, it is the
         true probability of the second class). Thus we have to override the
diff --git a/rampwf/score_types/roc_auc.py b/rampwf/score_types/roc_auc.py
index a06b2bbc..e6a20682 100644
--- a/rampwf/score_types/roc_auc.py
+++ b/rampwf/score_types/roc_auc.py
@@ -14,7 +14,7 @@ def __init__(self, name='roc_auc', precision=2):
     def score_function(self, ground_truths, predictions):
         """A hybrid score.
 
-        It tests the the predicted _probability_ of the second class
+        It tests the predicted _probability_ of the second class
         against the true _label index_ (which is 0 if the first label is the
         ground truth, and 1 if it is not, in other words, it is the
         true probability of the second class). Thus we have to override the
diff --git a/rampwf/tests/kits/digits/submissions/starting_kit/batch_classifier.py b/rampwf/tests/kits/digits/submissions/starting_kit/batch_classifier.py
index 6375dd23..89462352 100644
--- a/rampwf/tests/kits/digits/submissions/starting_kit/batch_classifier.py
+++ b/rampwf/tests/kits/digits/submissions/starting_kit/batch_classifier.py
@@ -49,7 +49,7 @@ def fit(self, gen_builder):
             # is set to some number > 1, the neural net will be trained with
             # repetitions of the same data, because the workers are independent
             # and they got through the same generator.
-            # Hence it is necessary to introduce a shared lock between the the
+            # Hence it is necessary to introduce a shared lock between the
             # processes so that they load different data, this can become a bit
             # complicated, so I choose to rather load exactly one chunk at a
             # time using 1 worker (so `workers` have to be equal to 1), but
diff --git a/rampwf/workflows/clusterer.py b/rampwf/workflows/clusterer.py
index 1ea0b7e7..ca3c2b9a 100644
--- a/rampwf/workflows/clusterer.py
+++ b/rampwf/workflows/clusterer.py
@@ -14,7 +14,7 @@
 of `X_array`). It slices up `X_array` into single events, drops the event ids,
 and sends the single event to the `predict_single_event` function implemented
 by the users. This function returns a vector of labels (cluster assignments)
-which is then joined back the the event id column and returned (to be passed
+which is then joined back to the event id column and returned (to be passed
 into `prediction_types.Clustering` and evaluated by
 `score_types.clustering_efficiency`).
 """