Spaces:
Sleeping
Sleeping
File size: 40,618 Bytes
7d5083d 68171c1 7d5083d 68171c1 7d5083d 68171c1 7d5083d 68171c1 7d5083d |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 |
"""
Core topic extraction logic using Anthropic models.
This module implements the main topic extraction functionality,
processing transcript chunks and extracting business-relevant topics
with confidence scoring and categorization.
"""
import json
import asyncio
import uuid
from typing import List, Dict, Any, Optional, Tuple
from datetime import datetime
from dataclasses import dataclass
from models.input import TranscriptSentence, TranscriptRequest, PromptConfiguration, LanguageCode
from models.output import (
TopicDetail, TopicCategory, SpeakerInsight, ProcessingMetadata,
SegmentationResult, ProcessingStatus
)
from utils.chunking import TranscriptChunker, ChunkingResult, create_chunker, ChunkingStrategy
from core.model_manager import ModelManager
from core.anthropic_client import AnthropicResponse, AnthropicError
from config.logging import get_logger
logger = get_logger(__name__)
@dataclass
class ExtractionContext:
"""Context information for topic extraction."""
request_id: str
transcript_id: Optional[str]
language: LanguageCode
business_domain: Optional[str]
total_sentences: int
total_duration: float
unique_speakers: List[str]
prompt_config: PromptConfiguration
@dataclass
class ChunkExtractionResult:
"""Result of topic extraction from a single chunk."""
chunk_id: int
topics: List[Dict[str, Any]]
processing_time: float
tokens_used: Dict[str, int]
model_used: str
confidence_scores: List[float]
warnings: List[str]
class TopicExtractor:
"""
Core topic extraction engine using Anthropic models.
Handles the complete pipeline from transcript chunking to topic extraction,
including business categorization, confidence scoring, and result aggregation.
"""
# Business category mapping for topic classification
CATEGORY_KEYWORDS = {
TopicCategory.CLIENT_NEEDS_B2B: [
"business needs", "requirements", "challenges", "pain points", "workflow",
"process", "efficiency", "productivity", "enterprise", "organization"
],
TopicCategory.CLIENT_NEEDS_B2C: [
"customer needs", "user experience", "personal", "individual", "consumer",
"lifestyle", "convenience", "satisfaction", "preferences"
],
TopicCategory.CUSTOMER_FEEDBACK: [
"feedback", "opinion", "review", "rating", "satisfaction", "experience",
"suggestion", "improvement", "complaint", "praise"
],
TopicCategory.EMPLOYEE_FEEDBACK: [
"employee", "staff", "team", "worker", "internal", "workplace",
"culture", "management", "training", "development"
],
TopicCategory.SOLUTION_BARRIERS: [
"barrier", "obstacle", "challenge", "difficulty", "problem", "issue",
"limitation", "constraint", "blocker", "resistance"
],
TopicCategory.SOLUTION_BENEFITS: [
"benefit", "advantage", "value", "improvement", "gain", "positive",
"success", "achievement", "outcome", "result"
],
TopicCategory.AHA_MOMENTS: [
"aha", "realization", "insight", "breakthrough", "understanding",
"clarity", "revelation", "discovery", "lightbulb", "eureka"
],
TopicCategory.COMPANY_INFO: [
"company", "organization", "business", "revenue", "size", "industry",
"market", "competition", "growth", "strategy"
],
TopicCategory.TECHNICAL_REQUIREMENTS: [
"technical", "technology", "system", "integration", "API", "platform",
"infrastructure", "security", "performance", "scalability"
],
TopicCategory.ADDITIONAL_COMMENTS: [
"additional", "other", "miscellaneous", "general", "comment",
"note", "remark", "observation", "thought"
]
}
def __init__(self, model_manager: ModelManager):
"""
Initialize the topic extractor.
Args:
model_manager: Model manager for Anthropic API access
"""
self.model_manager = model_manager
self.logger = get_logger(f"{__name__}.{self.__class__.__name__}")
async def extract_topics(
self,
request: TranscriptRequest,
context: Optional[ExtractionContext] = None
) -> SegmentationResult:
"""
Extract topics from a transcript request.
Args:
request: Transcript request with sentences and configuration
context: Optional extraction context
Returns:
Complete segmentation result with topics and metadata
"""
start_time = datetime.now()
request_id = context.request_id if context else str(uuid.uuid4())
self.logger.info(f"Starting topic extraction for request {request_id}")
try:
# Create extraction context if not provided
if not context:
context = self._create_context(request, request_id)
# Step 1: Chunk the transcript
chunking_result = await self._chunk_transcript(request, context)
# Step 2: Extract topics from each chunk
chunk_results = await self._extract_from_chunks(chunking_result, context)
# Check if all chunks failed
successful_chunks = [r for r in chunk_results if len(r.topics) > 0 or len(r.warnings) == 0]
if not successful_chunks and chunk_results:
# All chunks failed
processing_time = (datetime.now() - start_time).total_seconds()
error_metadata = ProcessingMetadata(
request_id=request_id,
timestamp=start_time,
model_used=self.model_manager.current_model.value,
processing_time=processing_time,
total_sentences=len(request.sentences),
total_duration=request.sentences[-1].end_time - request.sentences[0].start_time,
unique_speakers=len(set(s.speaker for s in request.sentences)),
topics_extracted=0,
average_confidence=0.0,
coverage_percentage=0.0,
warnings=[f"All {len(chunk_results)} chunks failed to process"]
)
return SegmentationResult(
status=ProcessingStatus.FAILED,
topics=[],
speaker_insights=[],
metadata=error_metadata
)
# Step 3: Aggregate and merge topics
aggregated_topics = await self._aggregate_topics(chunk_results, context, request)
# Step 4: Generate speaker insights
speaker_insights = await self._generate_speaker_insights(
request.sentences, aggregated_topics, context
)
# Step 5: Create processing metadata
processing_time = (datetime.now() - start_time).total_seconds()
metadata = self._create_metadata(
request, context, chunk_results, processing_time
)
# Step 6: Create final result
result = SegmentationResult(
status=ProcessingStatus.SUCCESS,
topics=aggregated_topics,
speaker_insights=speaker_insights,
metadata=metadata,
executive_summary=self._generate_executive_summary(aggregated_topics),
key_takeaways=self._extract_key_takeaways(aggregated_topics),
category_summary=self._create_category_summary(aggregated_topics)
)
self.logger.info(
f"Topic extraction completed for request {request_id}: "
f"{len(aggregated_topics)} topics in {processing_time:.2f}s"
)
return result
except Exception as e:
self.logger.error(f"Topic extraction failed for request {request_id}: {str(e)}")
# Return error result
processing_time = (datetime.now() - start_time).total_seconds()
error_metadata = ProcessingMetadata(
request_id=request_id,
timestamp=start_time,
model_used="unknown",
processing_time=processing_time,
total_sentences=len(request.sentences),
total_duration=request.sentences[-1].end_time - request.sentences[0].start_time,
unique_speakers=len(set(s.speaker for s in request.sentences)),
topics_extracted=0,
average_confidence=0.0,
coverage_percentage=0.0,
warnings=[f"Extraction failed: {str(e)}"]
)
return SegmentationResult(
status=ProcessingStatus.FAILED,
topics=[],
speaker_insights=[],
metadata=error_metadata
)
def _create_context(self, request: TranscriptRequest, request_id: str) -> ExtractionContext:
"""Create extraction context from request."""
return ExtractionContext(
request_id=request_id,
transcript_id=request.transcript_id,
language=request.prompt_config.language if request.prompt_config else LanguageCode.AUTO_DETECT,
business_domain=request.prompt_config.business_domain if request.prompt_config else None,
total_sentences=len(request.sentences),
total_duration=request.sentences[-1].end_time - request.sentences[0].start_time,
unique_speakers=list(set(s.speaker for s in request.sentences)),
prompt_config=request.prompt_config or PromptConfiguration()
)
async def _chunk_transcript(
self,
request: TranscriptRequest,
context: ExtractionContext
) -> ChunkingResult:
"""Chunk the transcript for processing."""
# Determine optimal chunking strategy
chunker = create_chunker(
strategy="token_estimate", # Use token-based chunking for Anthropic
chunk_size=6000, # Conservative chunk size
overlap_sentences=5, # Increased overlap for better context
model_name=self.model_manager.current_model.value
)
# Get optimal strategy based on transcript characteristics
optimal_strategy = chunker.get_optimal_strategy(request.sentences)
if optimal_strategy != chunker.strategy:
self.logger.info(f"Switching to optimal chunking strategy: {optimal_strategy}")
chunker.strategy = optimal_strategy
# Update chunk size for the new strategy
chunker.chunk_size = chunker.DEFAULT_CHUNK_SIZES[optimal_strategy]
# For large transcripts, force smaller chunks to get more granular topics
total_sentences = len(request.sentences)
if total_sentences > 100:
if chunker.strategy == ChunkingStrategy.SENTENCE_COUNT:
# Use smaller chunks for better topic granularity
chunker.chunk_size = min(40, chunker.chunk_size)
self.logger.info(f"Large transcript detected ({total_sentences} sentences), using smaller chunks: {chunker.chunk_size}")
elif chunker.strategy == ChunkingStrategy.TOKEN_ESTIMATE:
# Use smaller token chunks for better topic extraction
chunker.chunk_size = min(4000, chunker.chunk_size)
self.logger.info(f"Large transcript detected ({total_sentences} sentences), using smaller token chunks: {chunker.chunk_size}")
# Chunk the transcript
result = chunker.chunk_transcript(request.sentences, enable_overlap=True)
# Validate chunking result
warnings = chunker.validate_chunks(result)
if warnings:
self.logger.warning(f"Chunking warnings: {warnings}")
self.logger.info(
f"Transcript chunked into {result.total_chunks} chunks "
f"using {result.strategy_used} strategy (chunk_size: {chunker.chunk_size})"
)
return result
async def _extract_from_chunks(
self,
chunking_result: ChunkingResult,
context: ExtractionContext
) -> List[ChunkExtractionResult]:
"""Extract topics from each chunk."""
chunk_results = []
# Process chunks in parallel (with concurrency limit)
semaphore = asyncio.Semaphore(3) # Limit concurrent requests
async def process_chunk(chunk_idx: int, chunk_sentences: List[TranscriptSentence]):
async with semaphore:
return await self._extract_from_single_chunk(
chunk_idx, chunk_sentences, context
)
# Create tasks for all chunks
tasks = [
process_chunk(i, chunk)
for i, chunk in enumerate(chunking_result.chunks)
]
# Execute tasks and collect results
chunk_results = await asyncio.gather(*tasks, return_exceptions=True)
# Handle any exceptions
valid_results = []
for i, result in enumerate(chunk_results):
if isinstance(result, Exception):
self.logger.error(f"Chunk {i} processing failed: {str(result)}")
# Create error result
error_result = ChunkExtractionResult(
chunk_id=i,
topics=[],
processing_time=0.0,
tokens_used={"input_tokens": 0, "output_tokens": 0},
model_used="unknown",
confidence_scores=[],
warnings=[f"Processing failed: {str(result)}"]
)
valid_results.append(error_result)
else:
valid_results.append(result)
return valid_results
async def _extract_from_single_chunk(
self,
chunk_idx: int,
chunk_sentences: List[TranscriptSentence],
context: ExtractionContext
) -> ChunkExtractionResult:
"""Extract topics from a single chunk."""
start_time = datetime.now()
try:
# Build the prompt for this chunk
prompt = self._build_extraction_prompt(chunk_sentences, context)
# Make the API call
response = await self.model_manager.generate_completion(
messages=[{"role": "user", "content": prompt}],
system_prompt=self._get_system_prompt(context),
max_tokens=4000,
temperature=0.0 # Deterministic for consistency
)
# Parse the response
topics = self._parse_extraction_response(response.content, chunk_sentences)
# Calculate confidence scores
confidence_scores = [topic.get("confidence_score", 0.5) for topic in topics]
processing_time = (datetime.now() - start_time).total_seconds()
result = ChunkExtractionResult(
chunk_id=chunk_idx,
topics=topics,
processing_time=processing_time,
tokens_used=response.usage or {"input_tokens": 0, "output_tokens": 0},
model_used=response.model,
confidence_scores=confidence_scores,
warnings=[]
)
self.logger.debug(
f"Chunk {chunk_idx} processed: {len(topics)} topics, "
f"{processing_time:.2f}s, {sum(response.usage.values()) if response.usage else 0} tokens"
)
return result
except Exception as e:
self.logger.error(f"Error processing chunk {chunk_idx}: {str(e)}")
processing_time = (datetime.now() - start_time).total_seconds()
return ChunkExtractionResult(
chunk_id=chunk_idx,
topics=[],
processing_time=processing_time,
tokens_used={"input_tokens": 0, "output_tokens": 0},
model_used="unknown",
confidence_scores=[],
warnings=[f"Processing error: {str(e)}"]
)
def _build_extraction_prompt(
self,
chunk_sentences: List[TranscriptSentence],
context: ExtractionContext
) -> str:
"""Build the topic extraction prompt for a chunk using the prompt manager."""
from core.prompt_manager import get_prompt_manager
# Get the prompt manager
prompt_manager = get_prompt_manager()
# Process the prompt configuration
prompt_context = {
"transcript_text": self._format_transcript_chunk(chunk_sentences),
"speaker_count": len(context.unique_speakers),
"sentence_count": len(chunk_sentences),
"business_domain": context.business_domain
}
try:
processed_prompt = prompt_manager.process_prompt_configuration(
context.prompt_config,
prompt_context,
chunk_sentences
)
# Log validation results
if hasattr(processed_prompt.validation_result, 'warnings') and processed_prompt.validation_result.warnings:
self.logger.warning(
f"Prompt validation warnings: {len(processed_prompt.validation_result.warnings)}"
)
# Use the processed user prompt
self.logger.info(f"Generated prompt length: {len(processed_prompt.user_prompt)} chars")
self.logger.debug(f"Generated prompt (first 300 chars): {processed_prompt.user_prompt[:300]}")
return processed_prompt.user_prompt
except Exception as e:
self.logger.warning(f"Error processing prompt configuration: {str(e)}")
# Fallback to basic prompt
fallback_prompt = self._build_fallback_prompt(chunk_sentences, context)
self.logger.info(f"Using fallback prompt, length: {len(fallback_prompt)} chars")
self.logger.debug(f"Fallback prompt (first 300 chars): {fallback_prompt[:300]}")
return fallback_prompt
def _build_fallback_prompt(
self,
chunk_sentences: List[TranscriptSentence],
context: ExtractionContext
) -> str:
"""Build a fallback prompt when prompt manager fails."""
transcript_text = self._format_transcript_chunk(chunk_sentences)
return f"""Please analyze the following transcript segment and extract meaningful business topics.
TRANSCRIPT SEGMENT:
{transcript_text}
CONTEXT:
- Language: {context.language.value}
- Business Domain: {context.business_domain or "General"}
- Total Speakers: {len(context.unique_speakers)}
Please respond with a JSON array of topics with the following fields:
topic_name, topic_type, topic_detail, start_sentence_index, end_sentence_index,
primary_speaker, all_speakers, confidence_score, key_phrases, actionable_insights"""
def _format_transcript_chunk(self, chunk_sentences: List[TranscriptSentence]) -> str:
"""Format transcript chunk for prompt."""
lines = []
for sentence in chunk_sentences:
timestamp = f"[{sentence.start_time:.1f}s-{sentence.end_time:.1f}s]"
lines.append(f"{sentence.sentence_index}. {timestamp} {sentence.speaker}: {sentence.text}")
return "\n".join(lines)
def _get_system_prompt(self, context: ExtractionContext) -> str:
"""Get the system prompt for topic extraction."""
from core.prompt_manager import get_prompt_manager
try:
prompt_manager = get_prompt_manager()
processed_prompt = prompt_manager.process_prompt_configuration(
context.prompt_config,
{"business_domain": context.business_domain},
None
)
return processed_prompt.system_prompt
except Exception as e:
self.logger.warning(f"Error getting system prompt: {str(e)}")
# Fallback system prompt
return f"""You are an expert business analyst specializing in topic extraction from business conversations.
Language: {context.language.value}
Business domain: {context.business_domain or "General business"}
Always respond with valid JSON format and ensure all required fields are included."""
def _parse_extraction_response(
self,
response_content: str,
chunk_sentences: List[TranscriptSentence]
) -> List[Dict[str, Any]]:
"""Parse the Anthropic response into topic dictionaries."""
try:
content = response_content.strip()
topics_data = []
# Method 1: Try to parse as direct JSON array
if content.startswith('['):
topics_data = json.loads(content)
# Method 2: Try to parse as JSON object with "topics" key
elif content.startswith('{'):
parsed_obj = json.loads(content)
if isinstance(parsed_obj, dict) and "topics" in parsed_obj:
topics_data = parsed_obj["topics"]
else:
raise ValueError("JSON object does not contain 'topics' key")
# Method 3: Extract JSON from markdown code blocks
elif "```json" in content:
start_marker = "```json"
end_marker = "```"
start_idx = content.find(start_marker) + len(start_marker)
end_idx = content.find(end_marker, start_idx)
if start_idx > len(start_marker) - 1 and end_idx > start_idx:
json_str = content[start_idx:end_idx].strip()
parsed_obj = json.loads(json_str)
if isinstance(parsed_obj, list):
topics_data = parsed_obj
elif isinstance(parsed_obj, dict) and "topics" in parsed_obj:
topics_data = parsed_obj["topics"]
else:
raise ValueError("Extracted JSON is not valid format")
else:
raise ValueError("Could not extract JSON from markdown")
# Method 4: Search for JSON array or object in text
else:
# Look for JSON array first
array_start = content.find('[')
array_end = content.rfind(']') + 1
# Look for JSON object
obj_start = content.find('{')
obj_end = content.rfind('}') + 1
json_str = None
if array_start >= 0 and array_end > array_start:
json_str = content[array_start:array_end]
topics_data = json.loads(json_str)
elif obj_start >= 0 and obj_end > obj_start:
json_str = content[obj_start:obj_end]
parsed_obj = json.loads(json_str)
if isinstance(parsed_obj, dict) and "topics" in parsed_obj:
topics_data = parsed_obj["topics"]
else:
raise ValueError("JSON object does not contain 'topics' key")
else:
raise ValueError("No valid JSON found in response")
# Ensure we have a list
if not isinstance(topics_data, list):
raise ValueError(f"Expected list of topics, got {type(topics_data)}")
# Validate and enhance topics
validated_topics = []
for topic_data in topics_data:
validated_topic = self._validate_and_enhance_topic(topic_data, chunk_sentences)
if validated_topic:
validated_topics.append(validated_topic)
self.logger.info(f"Successfully parsed {len(validated_topics)} topics from response")
return validated_topics
except (json.JSONDecodeError, ValueError) as e:
self.logger.warning(f"Failed to parse extraction response: {str(e)}")
self.logger.warning(f"Response content (first 500 chars): {response_content[:500]}")
self.logger.warning(f"Response content (last 200 chars): {response_content[-200:]}")
# Return empty list if parsing fails
return []
def _validate_and_enhance_topic(
self,
topic_data: Dict[str, Any],
chunk_sentences: List[TranscriptSentence]
) -> Optional[Dict[str, Any]]:
"""Validate and enhance a topic from the response."""
try:
# Required fields
required_fields = [
"topic_name", "topic_type", "topic_detail",
"start_sentence_index", "end_sentence_index",
"primary_speaker", "all_speakers", "confidence_score"
]
for field in required_fields:
if field not in topic_data:
self.logger.warning(f"Missing required field '{field}' in topic")
return None
# Validate sentence indices
start_idx = topic_data["start_sentence_index"]
end_idx = topic_data["end_sentence_index"]
if start_idx < 1 or end_idx < start_idx:
self.logger.warning(f"Invalid sentence indices: {start_idx}-{end_idx}")
return None
# Find corresponding sentences
matching_sentences = [
s for s in chunk_sentences
if start_idx <= s.sentence_index <= end_idx
]
if not matching_sentences:
self.logger.warning(f"No matching sentences for indices {start_idx}-{end_idx}")
return None
# Enhance with timing information
topic_data["start_time"] = matching_sentences[0].start_time
topic_data["end_time"] = matching_sentences[-1].end_time
# Validate topic category
topic_type = topic_data["topic_type"]
try:
TopicCategory(topic_type)
except ValueError:
self.logger.warning(f"Invalid topic category: {topic_type}, using 'general'")
topic_data["topic_type"] = TopicCategory.GENERAL.value
# Ensure confidence score is valid
confidence = topic_data["confidence_score"]
if not isinstance(confidence, (int, float)) or not (0.0 <= confidence <= 1.0):
self.logger.warning(f"Invalid confidence score: {confidence}, using 0.5")
topic_data["confidence_score"] = 0.5
# Ensure lists are properly formatted
if not isinstance(topic_data.get("all_speakers", []), list):
topic_data["all_speakers"] = [topic_data["primary_speaker"]]
if not isinstance(topic_data.get("key_phrases", []), list):
topic_data["key_phrases"] = []
# Handle actionable_insights - can be string or list
insights = topic_data.get("actionable_insights", [])
if isinstance(insights, str):
# Split string into list if it's a single string
topic_data["actionable_insights"] = [insights] if insights.strip() else []
elif not isinstance(insights, list):
topic_data["actionable_insights"] = []
return topic_data
except Exception as e:
self.logger.warning(f"Error validating topic: {str(e)}")
return None
async def _aggregate_topics(
self,
chunk_results: List[ChunkExtractionResult],
context: ExtractionContext,
request: TranscriptRequest
) -> List[TopicDetail]:
"""Aggregate topics from all chunks and remove duplicates."""
all_topics = []
# Collect all topics from chunks
for chunk_result in chunk_results:
for topic_data in chunk_result.topics:
try:
topic = TopicDetail(
topic_name=topic_data["topic_name"],
topic_type=TopicCategory(topic_data["topic_type"]),
topic_detail=topic_data["topic_detail"],
start_time=topic_data["start_time"],
end_time=topic_data["end_time"],
start_sentence_index=topic_data["start_sentence_index"],
end_sentence_index=topic_data["end_sentence_index"],
primary_speaker=topic_data["primary_speaker"],
all_speakers=topic_data["all_speakers"],
confidence_score=topic_data["confidence_score"],
key_phrases=topic_data.get("key_phrases", []),
actionable_insights=topic_data.get("actionable_insights", [])
)
all_topics.append(topic)
except Exception as e:
self.logger.warning(f"Error creating TopicDetail: {str(e)}")
# Remove duplicates and merge similar topics
if request.merge_similar_topics:
all_topics = self._merge_similar_topics(all_topics)
# Sort by start time
all_topics.sort(key=lambda t: t.start_time)
return all_topics
def _merge_similar_topics(self, topics: List[TopicDetail]) -> List[TopicDetail]:
"""Merge similar or duplicate topics."""
if len(topics) <= 1:
return topics
merged_topics = []
used_indices = set()
for i, topic in enumerate(topics):
if i in used_indices:
continue
# Find similar topics
similar_topics = [topic]
for j, other_topic in enumerate(topics[i + 1:], i + 1):
if j in used_indices:
continue
if self._are_topics_similar(topic, other_topic):
similar_topics.append(other_topic)
used_indices.add(j)
# Merge similar topics
if len(similar_topics) > 1:
merged_topic = self._merge_topic_group(similar_topics)
merged_topics.append(merged_topic)
else:
merged_topics.append(topic)
used_indices.add(i)
return merged_topics
def _are_topics_similar(self, topic1: TopicDetail, topic2: TopicDetail) -> bool:
"""Check if two topics are similar enough to merge."""
# Same category and similar names
if topic1.topic_type == topic2.topic_type:
name_similarity = self._calculate_text_similarity(
topic1.topic_name.lower(), topic2.topic_name.lower()
)
if name_similarity > 0.7:
return True
# Overlapping time ranges
time_overlap = min(topic1.end_time, topic2.end_time) - max(topic1.start_time, topic2.start_time)
if time_overlap > 0:
overlap_ratio = time_overlap / min(topic1.duration, topic2.duration)
if overlap_ratio > 0.5:
return True
return False
def _calculate_text_similarity(self, text1: str, text2: str) -> float:
"""Calculate simple text similarity (Jaccard similarity)."""
words1 = set(text1.split())
words2 = set(text2.split())
if not words1 and not words2:
return 1.0
intersection = words1.intersection(words2)
union = words1.union(words2)
return len(intersection) / len(union) if union else 0.0
def _merge_topic_group(self, topics: List[TopicDetail]) -> TopicDetail:
"""Merge a group of similar topics into one."""
# Use the topic with highest confidence as base
base_topic = max(topics, key=lambda t: t.confidence_score)
# Merge time ranges
start_time = min(t.start_time for t in topics)
end_time = max(t.end_time for t in topics)
start_sentence_index = min(t.start_sentence_index for t in topics)
end_sentence_index = max(t.end_sentence_index for t in topics)
# Merge speakers
all_speakers = list(set(speaker for t in topics for speaker in t.all_speakers))
# Merge key phrases and insights
key_phrases = list(set(phrase for t in topics for phrase in t.key_phrases))
actionable_insights = list(set(insight for t in topics for insight in t.actionable_insights))
# Average confidence score
avg_confidence = sum(t.confidence_score for t in topics) / len(topics)
# Create merged topic
return TopicDetail(
topic_name=base_topic.topic_name,
topic_type=base_topic.topic_type,
topic_detail=f"Merged topic: {base_topic.topic_detail}",
start_time=start_time,
end_time=end_time,
start_sentence_index=start_sentence_index,
end_sentence_index=end_sentence_index,
primary_speaker=base_topic.primary_speaker,
all_speakers=all_speakers,
confidence_score=avg_confidence,
key_phrases=key_phrases,
actionable_insights=actionable_insights
)
async def _generate_speaker_insights(
self,
sentences: List[TranscriptSentence],
topics: List[TopicDetail],
context: ExtractionContext
) -> List[SpeakerInsight]:
"""Generate insights for each speaker."""
speaker_insights = []
for speaker in context.unique_speakers:
speaker_sentences = [s for s in sentences if s.speaker == speaker]
if not speaker_sentences:
continue
# Calculate speaker statistics
total_sentences = len(speaker_sentences)
total_duration = sum(s.end_time - s.start_time for s in speaker_sentences)
# Find topics this speaker contributed to
topics_mentioned = []
for topic in topics:
if speaker in topic.all_speakers:
topics_mentioned.append(topic.topic_name)
# Generate key insights (simplified)
key_insights = []
if total_sentences > 0:
avg_sentence_length = sum(len(s.text.split()) for s in speaker_sentences) / total_sentences
key_insights.append(f"Average sentence length: {avg_sentence_length:.1f} words")
if total_duration > 0:
speaking_rate = total_sentences / (total_duration / 60) # sentences per minute
key_insights.append(f"Speaking rate: {speaking_rate:.1f} sentences/minute")
insight = SpeakerInsight(
speaker=speaker,
speaker_role=speaker_sentences[0].speaker_role if speaker_sentences else None,
total_sentences=total_sentences,
total_duration=total_duration,
topics_mentioned=topics_mentioned,
key_insights=key_insights
)
speaker_insights.append(insight)
return speaker_insights
def _create_metadata(
self,
request: TranscriptRequest,
context: ExtractionContext,
chunk_results: List[ChunkExtractionResult],
processing_time: float
) -> ProcessingMetadata:
"""Create processing metadata."""
# Aggregate token usage
total_tokens = {"input_tokens": 0, "output_tokens": 0}
for result in chunk_results:
for key in total_tokens:
total_tokens[key] += result.tokens_used.get(key, 0)
# Collect warnings
warnings = []
for result in chunk_results:
warnings.extend(result.warnings)
# Calculate average confidence
all_confidences = []
for result in chunk_results:
all_confidences.extend(result.confidence_scores)
avg_confidence = sum(all_confidences) / len(all_confidences) if all_confidences else 0.0
return ProcessingMetadata(
request_id=context.request_id,
timestamp=datetime.now(),
model_used=self.model_manager.current_model.value,
processing_time=processing_time,
total_sentences=context.total_sentences,
total_duration=context.total_duration,
unique_speakers=len(context.unique_speakers),
topics_extracted=sum(len(result.topics) for result in chunk_results),
average_confidence=avg_confidence,
coverage_percentage=100.0, # Simplified
tokens_used=total_tokens,
detected_language=context.language,
warnings=warnings
)
def _generate_executive_summary(self, topics: List[TopicDetail]) -> str:
"""Generate an executive summary of the topics."""
if not topics:
return "No topics were extracted from the transcript."
# Count topics by category
category_counts = {}
for topic in topics:
category_counts[topic.topic_type] = category_counts.get(topic.topic_type, 0) + 1
# Generate summary
summary_parts = [
f"Analysis identified {len(topics)} key topics from the transcript."
]
if category_counts:
top_categories = sorted(category_counts.items(), key=lambda x: x[1], reverse=True)[:3]
category_text = ", ".join([f"{count} {cat.value.replace('_', ' ')}" for cat, count in top_categories])
summary_parts.append(f"Primary focus areas include: {category_text}.")
# Add confidence information
avg_confidence = sum(t.confidence_score for t in topics) / len(topics)
confidence_text = "high" if avg_confidence > 0.8 else "moderate" if avg_confidence > 0.6 else "low"
summary_parts.append(f"Overall extraction confidence is {confidence_text} ({avg_confidence:.2f}).")
return " ".join(summary_parts)
def _extract_key_takeaways(self, topics: List[TopicDetail]) -> List[str]:
"""Extract key takeaways from topics."""
takeaways = []
# Collect actionable insights
for topic in topics:
takeaways.extend(topic.actionable_insights)
# Add category-based insights
category_counts = {}
for topic in topics:
category_counts[topic.topic_type] = category_counts.get(topic.topic_type, 0) + 1
if TopicCategory.CLIENT_NEEDS_B2B in category_counts or TopicCategory.CLIENT_NEEDS_B2C in category_counts:
takeaways.append("Customer needs and requirements were clearly identified")
if TopicCategory.SOLUTION_BARRIERS in category_counts:
takeaways.append("Implementation barriers and challenges were discussed")
if TopicCategory.SOLUTION_BENEFITS in category_counts:
takeaways.append("Solution benefits and value propositions were highlighted")
# Limit to top takeaways
return takeaways[:10]
def _create_category_summary(self, topics: List[TopicDetail]) -> Dict[TopicCategory, int]:
"""Create a summary of topics by category."""
category_summary = {}
for topic in topics:
category_summary[topic.topic_type] = category_summary.get(topic.topic_type, 0) + 1
return category_summary |