"""Produce Word Document transcriptions using the automatic speech recognition from AWS Transcribe."""

# Initiate Document

document = Document()

# A4 Size

document.sections[0].page_width = Mm(210)

document.sections[0].page_height = Mm(297)

# Font

font = document.styles['Normal'].font

font.name = 'Calibri'

# Load Transcription output from command line input

# eg: python3 application.py 'output.json'

data = json.load(open(file))

click.echo(f"{click.format_filename(file)} opened...")

# Document title and intro

title = f"Transcription of {data['jobName']}"

document.add_heading(title, level=1)

# Set thresholds for formatting later

threshold_for_grey = 0.98

# Intro

document.add_paragraph('Transcription using AWS Transcribe automatic speech recognition.')

document.add_paragraph(datetime.datetime.now().strftime('Document produced on %A %d %B %Y at %X.'))

document.add_paragraph() # Spacing

document.add_paragraph(f"Grey text has less than {int(threshold_for_grey * 100)}% confidence.")

# Stats dictionary

stats = {

'timestamps': [],

'accuracy': [],

'9.8': 0, '9': 0, '8': 0, '7': 0, '6': 0, '5': 0, '4': 0, '3': 0, '2': 0, '1': 0, '0': 0,

'total': len(data['results']['items'])}

# Confidence count

click.echo('Producing stats...')

for item in data['results']['items']:

if item['type'] == 'pronunciation':

stats['timestamps'].append(float(item['start_time']))

stats['accuracy'].append(int(float(item['alternatives'][0]['confidence']) * 100))

if float(item['alternatives'][0]['confidence']) >= 0.98: stats['9.8'] += 1

elif float(item['alternatives'][0]['confidence']) >= 0.9: stats['9'] += 1

elif float(item['alternatives'][0]['confidence']) >= 0.8: stats['8'] += 1

elif float(item['alternatives'][0]['confidence']) >= 0.7: stats['7'] += 1

elif float(item['alternatives'][0]['confidence']) >= 0.6: stats['6'] += 1

elif float(item['alternatives'][0]['confidence']) >= 0.5: stats['5'] += 1

elif float(item['alternatives'][0]['confidence']) >= 0.4: stats['4'] += 1

elif float(item['alternatives'][0]['confidence']) >= 0.3: stats['3'] += 1

elif float(item['alternatives'][0]['confidence']) >= 0.2: stats['2'] += 1

elif float(item['alternatives'][0]['confidence']) >= 0.1: stats['1'] += 1

else: stats['0'] += 1

# Display confidence count table

table = document.add_table(rows=1, cols=3)

table.style = document.styles['Light List Accent 1']

table.alignment = WD_ALIGN_PARAGRAPH.CENTER

hdr_cells = table.rows[0].cells

hdr_cells[0].text = 'Confidence'

hdr_cells[1].text = 'Count'

hdr_cells[2].text = 'Percentage'

row_cells = table.add_row().cells

row_cells[0].text = str('98% - 100%')

row_cells[1].text = str(stats['9.8'])

row_cells[2].text = str(round(stats['9.8'] / stats['total'] * 100, 2)) + '%'

row_cells = table.add_row().cells

row_cells[0].text = str('90% - 97%')

row_cells[1].text = str(stats['9'])

row_cells[2].text = str(round(stats['9'] / stats['total'] * 100, 2)) + '%'

row_cells = table.add_row().cells

row_cells[0].text = str('80% - 89%')

row_cells[1].text = str(stats['8'])

row_cells[2].text = str(round(stats['8'] / stats['total'] * 100, 2)) + '%'

row_cells = table.add_row().cells

row_cells[0].text = str('70% - 79%')

row_cells[1].text = str(stats['7'])

row_cells[2].text = str(round(stats['7'] / stats['total'] * 100, 2)) + '%'

row_cells = table.add_row().cells

row_cells[0].text = str('60% - 69%')

row_cells[1].text = str(stats['6'])

row_cells[2].text = str(round(stats['6'] / stats['total'] * 100, 2)) + '%'

row_cells = table.add_row().cells

row_cells[0].text = str('50% - 59%')

row_cells[1].text = str(stats['5'])

row_cells[2].text = str(round(stats['5'] / stats['total'] * 100, 2)) + '%'

row_cells = table.add_row().cells

row_cells[0].text = str('40% - 49%')

row_cells[1].text = str(stats['4'])

row_cells[2].text = str(round(stats['4'] / stats['total'] * 100, 2)) + '%'

row_cells = table.add_row().cells

row_cells[0].text = str('30% - 39%')

row_cells[1].text = str(stats['3'])

row_cells[2].text = str(round(stats['3'] / stats['total'] * 100, 2)) + '%'

row_cells = table.add_row().cells

row_cells[0].text = str('20% - 29%')

row_cells[1].text = str(stats['2'])

row_cells[2].text = str(round(stats['2'] / stats['total'] * 100, 2)) + '%'

row_cells = table.add_row().cells

row_cells[0].text = str('10% - 19%')

row_cells[1].text = str(stats['1'])

row_cells[2].text = str(round(stats['1'] / stats['total'] * 100, 2)) + '%'

row_cells = table.add_row().cells

row_cells[0].text = str('0% - 9%')

row_cells[1].text = str(stats['0'])

row_cells[2].text = str(round(stats['0'] / stats['total'] * 100, 2)) + '%'

# Add paragraph for spacing

document.add_paragraph()

# Display scatter graph of confidence

# Confidence of each word as scatter graph

plt.scatter(stats['timestamps'], stats['accuracy'])

# Mean average as line across graph

plt.plot([stats['timestamps'][0], stats['timestamps'][-1]],

[statistics.mean(stats['accuracy']), statistics.mean(stats['accuracy'])], 'r')

# Formatting

plt.xlabel('Time (seconds)')

# plt.xticks(range(0, int(stats['timestamps'][-1]), 60))

plt.ylabel('Accuracy (percent)')

plt.yticks(range(0, 101, 10))

plt.title('Accuracy during video')

plt.legend(['Accuracy average (mean)', 'Individual words'], loc='lower center')

plt.savefig('chart.png')

document.add_picture('chart.png', width=Cm(14.64))

document.paragraphs[-1].alignment = WD_ALIGN_PARAGRAPH.CENTER

document.add_page_break()

# Process and display transcript by speaker segments

click.echo('Writing transcript...')

table = document.add_table(rows=1, cols=3)

table.style = document.styles['Light List Accent 1']

hdr_cells = table.rows[0].cells

hdr_cells[0].text = 'Time'

hdr_cells[1].text = 'Speaker'

hdr_cells[2].text = 'Content'

with click.progressbar(data['results']['speaker_labels']['segments']) as bar:

for segment in bar:

# If there is content in the segment

if len(segment['items']) > 0:

# Add a row, write the time and speaker

row_cells = table.add_row().cells

row_cells[0].text = convert_time_stamp(segment['start_time'])

row_cells[1].text = str(segment['speaker_label'])

# Segments group individual word results by speaker. They are cross-referenced by time.

# For each word in the segment...

for word in segment['items']:

# Run through the word results and get the corresponding result

for result in data['results']['items']:

if result['type'] == 'pronunciation':

if result['start_time'] == word['start_time']:

# Get the word with the highest confidence

if len(result['alternatives']) > 0:

current_word = dict()

confidence_scores = []

for score in result['alternatives']:

confidence_scores.append(score['confidence'])

for alternative in result['alternatives']:

if alternative['confidence'] == max(confidence_scores):

current_word = alternative.copy()

# Write and format the word

run = row_cells[2].paragraphs[0].add_run(' ' + current_word['content'])

if float(current_word['confidence']) < threshold_for_grey:

font = run.font

font.color.rgb = RGBColor(204, 204, 204)

# If the next item is punctuation, add it

try:

if data['results']['items'][data['results']['items'].index(result) + 1]['type'] == 'punctuation':

run = row_cells[2].paragraphs[0].add_run(data['results']['items'][data['results']['items'].index(result) + 1]['alternatives'][0]['content'])

# Occasional IndexErrors encountered

except:

pass

# Formatting transcript table widthds

click.echo('Formatting...')

widths = (Inches(0.6), Inches(1), Inches(4.5))

with click.progressbar(table.rows) as bar:

for row in bar:

for idx, width in enumerate(widths):

row.cells[idx].width = width

# Save the file

document_title = f"{data['jobName']}.docx"

document.save(document_title)

click.echo(f"{document_title} saved.")

if log:

# Logging

logs = boto3.client('logs') # this might throw exception if we don't have `aws configure`

def write_log(log_text):

log_info = logs.describe_log_streams(

logGroupName=log[0],

logStreamNamePrefix=log[1])

log_time = int(datetime.datetime.now().timestamp() * 1000)

response = logs.put_log_events(

logGroupName=log[0],

logStreamName=log[1],

logEvents=[

{

'timestamp': log_time,

'message': log_text

},

],

sequenceToken=log_info['logStreams'][0]['uploadSequenceToken']

)

job_finish = datetime.datetime.now()

job_duration = job_finish - job_start

write_log(f"Job name: {data['jobName']}, Word count: {stats['total']}, Accuracy average: {round(statistics.mean(stats['accuracy']), 2)}, Job duration: {job_duration.seconds}")