count_durations_for_test_subset.py

# Copyright Amazon.com, Inc. or its affiliates. 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.


import os
import pickle
import sys
from statistics import mean
from preprocessing_scripts import Bin

ref = sys.argv[1]
hyp = sys.argv[2]

path = "./"
correct_dur_pred, incorrect_dur_pred = 0, 0
durations_hyp = []
durations_ref = []

durations_path = './durations_freq_all.pkl'
if os.path.exists(durations_path):
    with open(durations_path, 'rb') as f:
        duration_freq = pickle.load(f)
        print("loaded durations' freq!")
        bin_instance = Bin(duration_freq, n=100)

with open(os.path.join(path, ref)) as f:
    lines = f.readlines()
    sentences = []
    for line in lines:
        line = " ".join(line.split("<eow>"))
        line_segments = line.split('[pause]')
        counter_line = []

        for segment in line_segments:
            counter_segment = 0
            durations = segment.split()
            for i in durations:
                counter_line.append(float(i))
        durations_ref.append(counter_line)

with open(os.path.join(path, hyp)) as f:
    lines = f.readlines()
    sentences = []
    for i, line in enumerate(lines):
        line = " ".join(line.split("<eow>"))
        line_segments = line.split('[pause]')
        counter_line = []
        for segment in line_segments:
            counter_segment = 0
            durations = segment.split()
            for i in durations:
                counter_line.append(float(i))
        durations_hyp.append(counter_line)

errors = []
scores = []
count = 0
count_right = 0
one_pause_or_more = 0
differences = []
sampling_rate = 22050
hop_length = 256
large_dif = 0
small_dif = 0
threshold_in_frames = int(0.3 * sampling_rate/hop_length)
# 0.3s corresponds to 25 frames

for i in range(len(durations_ref)):
        one_pause_or_more += 1
        if len(durations_ref[i]) == len(durations_hyp[i]):
            for j in range(len(durations_ref[i])):
                if durations_ref[i][j] == 0:
                    temp = 1
                else:
                    temp = durations_ref[i][j]
                abs_diff = abs(durations_hyp[i][j] - temp)
                errors.append(abs_diff/temp)
                if durations_hyp[i][j] < temp:
                    scores.append(durations_hyp[i][j]/temp)
                else:
                    scores.append(temp/durations_hyp[i][j])
                differences.append(abs_diff)
                if abs_diff >= threshold_in_frames:
                    large_dif += 1
                else:
                    small_dif += 1
            count_right += 1
        else:
            count += 1
            for j in range(len(durations_ref[i])):
                if durations_ref[i][j] == 0:
                    temp = 1
                else:
                    temp = durations_ref[i][j]
                abs_diff = abs(0 - temp)
                errors.append(abs_diff/temp)
                scores.append(0.0)
                differences.append(threshold_in_frames + 10)
                large_dif += 1

print("Metric 1 is {}".format(1 - mean(errors)))
print("Metric 2 is {}".format(mean(scores)))
print("Metric 3 is {}".format(mean(differences)))

print("Segments with diff > 0.3s: {}".format(large_dif/(small_dif + large_dif)))
print("How many sentences have 1 or more pauses: {}".format(one_pause_or_more))

print("Predicted wrong number of pauses in {} out of {} sentences".format(count, count + count_right))