def test_e2e():
with open("tests/data/lucier_golden.json") as f:
golden = json.load(f)
with open("tests/data/lucier.txt") as f:
transcript = f.read()
ret = lm_transcribe("tests/data/lucier.mp3", transcript, "PROTO_LANGDIR",
"data/nnet_a_gpu_online")
score = alignment_score(golden['words'], ret['words'])
assert_greater(score['correct'], 0.85)
assert_less(score['error'], 0.35)
def test_e2e():
with open("tests/data/lucier_golden.json") as f:
golden = json.load(f)
with open("tests/data/lucier.txt") as f:
transcript = f.read()
ret = lm_transcribe(
"tests/data/lucier.mp3",
transcript,
"PROTO_LANGDIR",
"data/nnet_a_gpu_online")
score = alignment_score(golden['words'], ret['words'])
assert_greater(score['correct'], 0.85)
assert_less(score['error'], 0.35)
def transcribe():
transcript = request.form['transcript']
audio = request.files['audio']
_, extension = os.path.splitext(audio.filename)
audio_file = tempfile.NamedTemporaryFile(suffix=extension)
audio.save(audio_file)
proto_langdir = app.config['proto_langdir']
nnet_dir = app.config['nnet_dir']
aligned = lm_transcribe(audio_file.name, transcript,
proto_langdir, nnet_dir)
return jsonify(transcript=aligned)