Reformat the output of Amazon Transcribe so it's readable.

speakers.py

"""
Reformat the output of Amazon Transcribe

usage: python3 speakers.py asrOutput.json
"""
import json
import sys

# read in the JSON output of Amazon Transcribe
with open(sys.argv[1]) as f:
    t = json.load(f)

i = 0

# items seem to be either individual words or puctuation
items = t['results']['items']

# segments are series of items spoken by a speaker, in other words, somebody
# said something.
segments = t['results']['speaker_labels']['segments']

# step through all the segments joining up the items that make up the segment
for segment in segments:
    print('\n', segment['speaker_label'])

    contents = []
    for seg_item in segment['items']:
        assert seg_item['speaker_label'] == segment['speaker_label']

        item = items[i]
        if item['type'] == 'pronunciation':
            contents.append(' ')
            assert item['start_time'] == seg_item['start_time'], 'item=' + str(item) + '\n' + 'seg_item=' + str(seg_item)
            assert item['end_time'] == seg_item['end_time'], 'item=' + str(item) + '\n' + 'seg_item=' + str(seg_item)
        contents.append(item['alternatives'][0]['content'])
        i += 1

        item = items[i]
        if item['type'] == 'punctuation':
            contents.append(item['alternatives'][0]['content'])
            i += 1

    print(''.join(contents))