You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
PaddleSpeech/paddlespeech/s2t/utils/asr_utils.py

53 lines
1.9 KiB

# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# Reference espnet Apache 2.0 (http://www.apache.org/licenses/LICENSE-2.0)
import json
import numpy as np
__all__ = ["label_smoothing_dist"]
def label_smoothing_dist(odim, lsm_type, transcript=None, blank=0):
"""Obtain label distribution for loss smoothing.
:param odim:
:param lsm_type:
:param blank:
:param transcript:
:return:
"""
if transcript is not None:
with open(transcript, "rb") as f:
trans_json = json.load(f)["utts"]
if lsm_type == "unigram":
assert transcript is not None, (
"transcript is required for %s label smoothing" % lsm_type)
labelcount = np.zeros(odim)
for k, v in trans_json.items():
ids = np.array([int(n) for n in v["output"][0]["tokenid"].split()])
# to avoid an error when there is no text in an uttrance
if len(ids) > 0:
labelcount[ids] += 1
labelcount[odim - 1] = len(transcript) # count <eos>
labelcount[labelcount == 0] = 1 # flooring
labelcount[blank] = 0 # remove counts for blank
labeldist = labelcount.astype(np.float32) / np.sum(labelcount)
else:
logging.error("Error: unexpected label smoothing type: %s" % lsm_type)
sys.exit()
return labeldist