Spaces:
Running
Running
File size: 12,290 Bytes
4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 214a4d6 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 214a4d6 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 214a4d6 fb1eceb 4322c44 fb1eceb 214a4d6 fb1eceb 4322c44 fb1eceb 4322c44 214a4d6 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 fb1eceb 4322c44 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 |
import gradio as gr
import re
import difflib
import os
from typing import List, Dict, Tuple, Optional
from dataclasses import dataclass
import numpy as np
@dataclass
class Segment:
"""A segment of a transcript with a speaker and text"""
speaker: str
timestamp: str
text: str
original_text: str # The text as it appears in the original transcript
index: int # Position in the original transcript
def clean_text_for_matching(text: str) -> str:
"""Clean text for matching purposes (remove formatting, punctuation, etc.)"""
# Remove markdown links and formatting
text = re.sub(r'\[([^\]]+)\]\([^)]+\)', r'\1', text) # Replace markdown links with just the text
text = re.sub(r'\*\*|\*', '', text) # Remove bold and italic formatting
# Remove common filler words and punctuation for better matching
text = re.sub(r'[,.;:!?]', ' ', text)
text = re.sub(r'\s+', ' ', text)
return text.lower().strip()
def load_transcript_file(file_path: str) -> str:
"""Load transcript from a file"""
with open(file_path, 'r', encoding='utf-8') as f:
return f.read()
def parse_transcript(transcript: str) -> List[Segment]:
"""
Parse transcript into segments.
Works with both formats:
- Speaker LastName 00:00:00
- **Speaker LastName** *00:00:00*
"""
# Match both markdown and plain formats
pattern = r"(?:\*\*)?(?:Speaker\s+)?([A-Za-z]+)(?:\*\*)?\s+(?:\*)?([0-9:]+)(?:\*)?\s*\n\n(.*?)(?=\n\n(?:\*\*)?(?:Speaker\s+)?[A-Za-z]+|\Z)"
segments = []
for i, match in enumerate(re.finditer(pattern, transcript, re.DOTALL)):
speaker, timestamp, text = match.groups()
original_text = text.strip()
cleaned_text = clean_text_for_matching(original_text)
segments.append(Segment(speaker, timestamp, cleaned_text, original_text, i))
return segments
def align_segments(auto_segments: List[Segment], human_segments: List[Segment]) -> Dict[int, int]:
"""
Align segments from human-edited transcript to auto-generated transcript.
Returns a dictionary mapping human segment indices to auto segment indices.
"""
alignments = {}
# Create text similarity matrix
similarity_matrix = np.zeros((len(human_segments), len(auto_segments)))
for h_idx, h_segment in enumerate(human_segments):
for a_idx, a_segment in enumerate(auto_segments):
similarity = difflib.SequenceMatcher(None, h_segment.text, a_segment.text).ratio()
similarity_matrix[h_idx, a_idx] = similarity
# Find best matches while maintaining order
remaining_auto_indices = set(range(len(auto_segments)))
for h_idx, h_segment in enumerate(human_segments):
# Find the best matching auto segment that hasn't been assigned yet
best_match = -1
best_similarity = 0.5 # Threshold for considering a match
for a_idx in remaining_auto_indices:
similarity = similarity_matrix[h_idx, a_idx]
if similarity > best_similarity:
# Check if this would violate sequence ordering
if all(aligned_a_idx < a_idx for aligned_h_idx, aligned_a_idx in alignments.items() if aligned_h_idx < h_idx):
best_match = a_idx
best_similarity = similarity
if best_match >= 0:
alignments[h_idx] = best_match
remaining_auto_indices.remove(best_match)
return alignments
def update_transcript(human_segments: List[Segment], auto_segments: List[Segment],
alignments: Dict[int, int], is_markdown: bool) -> str:
"""
Create updated transcript by transferring timestamps from auto segments to human segments.
Preserves all human edits, formatting, links, etc.
"""
updated_segments = []
for h_idx, h_segment in enumerate(human_segments):
if h_idx in alignments:
# Segment was matched, use timestamp from auto segment
a_idx = alignments[h_idx]
if is_markdown:
updated_segments.append(f"**{h_segment.speaker}** *{auto_segments[a_idx].timestamp}*\n\n{h_segment.original_text}")
else:
updated_segments.append(f"Speaker {h_segment.speaker} {auto_segments[a_idx].timestamp}\n\n{h_segment.original_text}")
else:
# No match found, keep original timestamp but mark it
if is_markdown:
updated_segments.append(f"**{h_segment.speaker}** *{h_segment.timestamp} [NO MATCH]*\n\n{h_segment.original_text}")
else:
updated_segments.append(f"Speaker {h_segment.speaker} {h_segment.timestamp} [NO MATCH]\n\n{h_segment.original_text}")
return "\n\n".join(updated_segments)
def generate_match_report(human_segments: List[Segment], auto_segments: List[Segment],
alignments: Dict[int, int]) -> str:
"""Generate a report about the matching process"""
total_human = len(human_segments)
total_auto = len(auto_segments)
total_matched = len(alignments)
report = f"### Matching Report\n\n"
report += f"- Human segments: {total_human}\n"
report += f"- Auto segments: {total_auto}\n"
report += f"- Matched segments: {total_matched} ({total_matched/total_human*100:.1f}%)\n"
if total_matched < total_human:
report += f"\n### Unmatched Segments ({total_human - total_matched})\n\n"
for h_idx, h_segment in enumerate(human_segments):
if h_idx not in alignments:
report += f"- Speaker {h_segment.speaker} at {h_segment.timestamp}: '{h_segment.text[:50]}...'\n"
# Calculate average similarity of matches
if alignments:
similarities = [
difflib.SequenceMatcher(None,
human_segments[h_idx].text,
auto_segments[a_idx].text).ratio()
for h_idx, a_idx in alignments.items()
]
avg_similarity = sum(similarities) / len(similarities)
report += f"\n### Match Quality\n\n"
report += f"- Average similarity: {avg_similarity:.2f}\n"
return report
def process_transcripts(auto_transcript, human_transcript):
"""Process the auto and human transcripts to update timestamps"""
try:
# Load transcripts
auto_content = auto_transcript.decode('utf-8') if isinstance(auto_transcript, bytes) else auto_transcript
human_content = human_transcript.decode('utf-8') if isinstance(human_transcript, bytes) else human_transcript
# Check if transcripts use markdown formatting
is_markdown = "**" in human_content
# Parse transcripts
auto_segments = parse_transcript(auto_content)
human_segments = parse_transcript(human_content)
if not auto_segments or not human_segments:
return "Error: Could not parse transcripts. Please check the format.", ""
# Align segments
alignments = align_segments(auto_segments, human_segments)
# Update transcript
updated_transcript = update_transcript(human_segments, auto_segments, alignments, is_markdown)
# Generate report
report = generate_match_report(human_segments, auto_segments, alignments)
return updated_transcript, report
except Exception as e:
return f"Error processing transcripts: {str(e)}", ""
def save_transcript(transcript: str) -> str:
"""Save transcript to a temporary file and return the path"""
output_dir = "output"
if not os.path.exists(output_dir):
os.makedirs(output_dir)
output_path = os.path.join(output_dir, "updated_transcript.md")
with open(output_path, 'w', encoding='utf-8') as f:
f.write(transcript)
return output_path
# Create Gradio interface
with gr.Blocks(title="Transcript Timestamp Synchronizer") as demo:
gr.Markdown("""
# ๐๏ธ Transcript Timestamp Synchronizer
This tool updates timestamps in human-edited transcripts based on new auto-generated transcripts.
## Instructions:
1. Upload or paste your new auto-generated transcript (with updated timestamps)
2. Upload or paste your human-edited transcript (with old timestamps)
3. Click "Synchronize Timestamps" to generate an updated transcript
The tool will match segments between the transcripts and update the timestamps while preserving all human edits.
""")
with gr.Row():
with gr.Column():
auto_source = gr.Radio(
["Upload File", "Paste Text"],
label="Auto-generated Transcript Source",
value="Paste Text"
)
auto_file = gr.File(
label="Upload Auto-generated Transcript",
file_types=[".md", ".txt"],
visible=False
)
auto_text = gr.TextArea(
label="Auto-generated Transcript (with new timestamps)",
placeholder="Paste the auto-generated transcript here...",
lines=15,
visible=True
)
with gr.Column():
human_source = gr.Radio(
["Upload File", "Paste Text"],
label="Human-edited Transcript Source",
value="Paste Text"
)
human_file = gr.File(
label="Upload Human-edited Transcript",
file_types=[".md", ".txt"],
visible=False
)
human_text = gr.TextArea(
label="Human-edited Transcript (with old timestamps)",
placeholder="Paste the human-edited transcript here...",
lines=15,
visible=True
)
update_btn = gr.Button("Synchronize Timestamps")
with gr.Tabs():
with gr.TabItem("Updated Transcript"):
updated_transcript = gr.TextArea(
label="Updated Transcript",
placeholder="The updated transcript will appear here...",
lines=20
)
download_btn = gr.Button("Download Updated Transcript")
download_path = gr.File(label="Download", visible=False)
with gr.TabItem("Matching Report"):
matching_report = gr.Markdown(
label="Matching Report",
value="The matching report will appear here..."
)
# Handle visibility of upload/paste options
def update_auto_visibility(choice):
return gr.update(visible=choice=="Upload File"), gr.update(visible=choice=="Paste Text")
def update_human_visibility(choice):
return gr.update(visible=choice=="Upload File"), gr.update(visible=choice=="Paste Text")
auto_source.change(update_auto_visibility, auto_source, [auto_file, auto_text])
human_source.change(update_human_visibility, human_source, [human_file, human_text])
# Load file content if uploaded
def load_auto_file(file):
if file is None:
return ""
with open(file.name, "r", encoding="utf-8") as f:
return f.read()
def load_human_file(file):
if file is None:
return ""
with open(file.name, "r", encoding="utf-8") as f:
return f.read()
auto_file.change(load_auto_file, auto_file, auto_text)
human_file.change(load_human_file, human_file, human_text)
# Process transcripts
def handle_process(auto_content, human_content):
return process_transcripts(auto_content, human_content)
update_btn.click(
fn=handle_process,
inputs=[auto_text, human_text],
outputs=[updated_transcript, matching_report]
)
# Handle download
def prepare_download(transcript):
if not transcript:
return None
return save_transcript(transcript)
download_btn.click(
fn=prepare_download,
inputs=[updated_transcript],
outputs=[download_path]
)
# For local testing
if __name__ == "__main__":
demo.launch() |