diff --git a/tests/test_pipelines_automatic_speech_recognition.py b/tests/test_pipelines_automatic_speech_recognition.py
index eebdc477b0dd18..6d391eee9a934c 100644
--- a/tests/test_pipelines_automatic_speech_recognition.py
+++ b/tests/test_pipelines_automatic_speech_recognition.py
@@ -34,14 +34,14 @@ def test_pt_defaults(self):
     def test_torch_small(self):
         import numpy as np
 
-        nlp = pipeline(
+        speech_recognizer = pipeline(
             task="automatic-speech-recognition",
             model="facebook/s2t-small-mustc-en-fr-st",
             tokenizer="facebook/s2t-small-mustc-en-fr-st",
             framework="pt",
         )
         waveform = np.zeros((34000,))
-        output = nlp(waveform)
+        output = speech_recognizer(waveform)
         self.assertEqual(output, {"text": "C'est ce que j'ai fait à ce moment-là."})
 
     @require_datasets
@@ -50,21 +50,21 @@ def test_torch_small(self):
     def test_torch_large(self):
         import numpy as np
 
-        nlp = pipeline(
+        speech_recognizer = pipeline(
             task="automatic-speech-recognition",
             model="facebook/wav2vec2-base-960h",
             tokenizer="facebook/wav2vec2-base-960h",
             framework="pt",
         )
         waveform = np.zeros((34000,))
-        output = nlp(waveform)
+        output = speech_recognizer(waveform)
         self.assertEqual(output, {"text": ""})
 
         from datasets import load_dataset
 
         ds = load_dataset("patrickvonplaten/librispeech_asr_dummy", "clean", split="validation")
         filename = ds[0]["file"]
-        output = nlp(filename)
+        output = speech_recognizer(filename)
         self.assertEqual(output, {"text": "A MAN SAID TO THE UNIVERSE SIR I EXIST"})
 
     @slow