Pocketsphinx produces poor results

Question

I am trying to use Pocketsphinx (for Python) with the code below:

from pocketsphinx import AudioFile, get_model_path
for phrase in AudioFile("test.wav", 
                hmm=get_model_path("en-us/en-us"), 
                lm=get_model_path("en-us/en-us.lm.bin"), 
                dict=get_model_path("en-us/cmudict-en-us.dict")):
    print(phrase)

I tested it with a sample audio clip (available at https://drive.google.com/file/d/1FZVMG5TDaeXq23M-IjNlS_SpZH25AhCY/view?usp=sharing) that reads: 'This is a librivox recording. Our librivox recordings are in the public domain. For more information or to volunteer, please visit lirivox.org.'

However, the result is very poor:

thune upton what i what we
or that a a when compared to what if the that met with what what are the what what he'd room for and what about when he
corrupt and laugh when he heard what are you that when the the
when the fun of

Any idea about what am I doing wrong?

Thanks in advance