|
|
import json |
|
|
import re |
|
|
from typing import List, Dict, Tuple |
|
|
|
|
|
def parse_timestamp(timestamp: str) -> Tuple[int, int]: |
|
|
"""Convert timestamp string like '00:15' to seconds.""" |
|
|
minutes, seconds = map(int, timestamp.split(':')) |
|
|
return minutes * 60 + seconds |
|
|
|
|
|
def extract_time_range(entry: str) -> Tuple[int, int]: |
|
|
"""Extract start and end times from an entry like '[00:00 - 00:13]'.""" |
|
|
match = re.match(r'\[(\d{2}:\d{2}) - (\d{2}:\d{2})\]', entry) |
|
|
if not match: |
|
|
return None |
|
|
start_time = parse_timestamp(match.group(1)) |
|
|
end_time = parse_timestamp(match.group(2)) |
|
|
return (start_time, end_time) |
|
|
|
|
|
def has_overlap(range1: Tuple[int, int], range2: Tuple[int, int]) -> bool: |
|
|
"""Check if two time ranges overlap.""" |
|
|
start1, end1 = range1 |
|
|
start2, end2 = range2 |
|
|
return not (end1 <= start2 or end2 <= start1) |
|
|
|
|
|
def clean_transcript(transcript: str) -> str: |
|
|
"""Clean a single transcript by removing overlapping segments.""" |
|
|
lines = transcript.split('\n') |
|
|
cleaned_lines = [] |
|
|
time_ranges = [] |
|
|
|
|
|
for line in lines: |
|
|
if not line.strip(): |
|
|
continue |
|
|
|
|
|
time_range = extract_time_range(line) |
|
|
if time_range is None: |
|
|
continue |
|
|
|
|
|
|
|
|
has_conflict = False |
|
|
for existing_range in time_ranges: |
|
|
if has_overlap(time_range, existing_range): |
|
|
has_conflict = True |
|
|
break |
|
|
|
|
|
if not has_conflict: |
|
|
time_ranges.append(time_range) |
|
|
cleaned_lines.append(line) |
|
|
|
|
|
return '\n'.join(cleaned_lines) |
|
|
|
|
|
def process_file(input_file: str, output_file: str): |
|
|
"""Process the JSON file and clean overlapping transcriptions.""" |
|
|
with open(input_file, 'r', encoding='utf-8') as f: |
|
|
data = json.load(f) |
|
|
|
|
|
if isinstance(data, dict): |
|
|
data = [data] |
|
|
|
|
|
cleaned_data = [] |
|
|
for entry in data: |
|
|
if 'model_output' in entry: |
|
|
entry['model_output'] = clean_transcript(entry['model_output']) |
|
|
cleaned_data.append(entry) |
|
|
|
|
|
with open(output_file, 'w', encoding='utf-8') as f: |
|
|
json.dump(cleaned_data, f, ensure_ascii=False, indent=2) |
|
|
|
|
|
if __name__ == '__main__': |
|
|
input_file = 'silence_overlaps/overlap5s_transcriptions.json' |
|
|
output_file = 'silence_overlaps/cleaned_transcriptions.json' |
|
|
process_file(input_file, output_file) |
|
|
print(f"Cleaned transcriptions have been saved to {output_file}") |
