Slash / api /summarizer.py
ND06-25's picture
first commit to AI repo
6880cd9
raw
history blame
8.27 kB
from transformers import pipeline, AutoTokenizer, AutoModelForSeq2SeqLM
from typing import List, Dict, Any, Optional, Union
import torch
import logging
from .utils import chunk_text
logger = logging.getLogger(__name__)
class BookSummarizer:
"""
Handles AI-powered text summarization using transformer models.
"""
def __init__(self, model_name: str = "facebook/bart-large-cnn"):
"""
Initialize the summarizer with a specific model.
Args:
model_name: Hugging Face model name for summarization
"""
self.model_name = model_name
self.summarizer = None
self.tokenizer = None
self.model = None
self.device = "cuda" if torch.cuda.is_available() else "cpu"
logger.info(f"Initializing summarizer with model: {model_name}")
logger.info(f"Using device: {self.device}")
def load_model(self):
"""
Load the summarization model and tokenizer.
"""
try:
logger.info("Loading summarization model...")
# Load tokenizer and model
self.tokenizer = AutoTokenizer.from_pretrained(self.model_name)
self.model = AutoModelForSeq2SeqLM.from_pretrained(self.model_name)
# Move model to appropriate device
self.model.to(self.device)
# Create pipeline
self.summarizer = pipeline(
"summarization",
model=self.model,
tokenizer=self.tokenizer,
device=0 if self.device == "cuda" else -1
)
logger.info("Model loaded successfully")
except Exception as e:
logger.error(f"Error loading model: {str(e)}")
raise
def summarize_text(self, text: str, max_length: int = 150, min_length: int = 50,
do_sample: bool = False) -> Dict[str, Any]:
"""
Summarize a single text chunk.
Args:
text: Text to summarize
max_length: Maximum length of summary
min_length: Minimum length of summary
do_sample: Whether to use sampling for generation
Returns:
Dictionary containing summary and metadata
"""
try:
if not self.summarizer:
self.load_model()
# Check if text is too short
if len(text.split()) < 50:
return {
'success': True,
'summary': text,
'original_length': len(text.split()),
'summary_length': len(text.split()),
'compression_ratio': 1.0
}
# Generate summary
summary_result = self.summarizer(
text,
max_length=max_length,
min_length=min_length,
do_sample=do_sample,
truncation=True
)
summary = summary_result[0]['summary_text']
# Calculate compression ratio
original_words = len(text.split())
summary_words = len(summary.split())
compression_ratio = summary_words / original_words if original_words > 0 else 0
return {
'success': True,
'summary': summary,
'original_length': original_words,
'summary_length': summary_words,
'compression_ratio': compression_ratio
}
except Exception as e:
logger.error(f"Error summarizing text: {str(e)}")
return {
'success': False,
'summary': '',
'error': str(e)
}
def summarize_book(self, text: str, chunk_size: int = 1000, overlap: int = 100,
max_length: int = 150, min_length: int = 50) -> Dict[str, Any]:
"""
Summarize a complete book by processing it in chunks.
Args:
text: Complete book text
chunk_size: Size of each text chunk
overlap: Overlap between chunks
max_length: Maximum length of each summary
min_length: Minimum length of each summary
Returns:
Dictionary containing complete summary and metadata
"""
try:
logger.info("Starting book summarization...")
# Split text into chunks
chunks = chunk_text(text, chunk_size, overlap)
logger.info(f"Split text into {len(chunks)} chunks")
# Summarize each chunk
chunk_summaries = []
total_original_words = 0
total_summary_words = 0
for i, chunk in enumerate(chunks):
logger.info(f"Processing chunk {i+1}/{len(chunks)}")
result = self.summarize_text(chunk, max_length, min_length)
if result['success']:
chunk_summaries.append(result['summary'])
total_original_words += result['original_length']
total_summary_words += result['summary_length']
else:
logger.warning(f"Failed to summarize chunk {i+1}: {result.get('error', 'Unknown error')}")
# Include original chunk if summarization fails
chunk_summaries.append(chunk[:200] + "...")
# Combine all summaries
combined_summary = " ".join(chunk_summaries)
# Create final summary if the combined summary is still too long
if len(combined_summary.split()) > 500:
logger.info("Creating final summary from combined summaries...")
final_result = self.summarize_text(combined_summary, max_length=300, min_length=100)
if final_result['success']:
combined_summary = final_result['summary']
# Calculate overall statistics
overall_compression = total_summary_words / total_original_words if total_original_words > 0 else 0
return {
'success': True,
'summary': combined_summary,
'statistics': {
'total_chunks': len(chunks),
'total_original_words': total_original_words,
'total_summary_words': total_summary_words,
'overall_compression_ratio': overall_compression,
'final_summary_length': len(combined_summary.split())
},
'chunk_summaries': chunk_summaries
}
except Exception as e:
logger.error(f"Error in book summarization: {str(e)}")
return {
'success': False,
'summary': '',
'error': str(e)
}
def get_available_models(self) -> List[Dict[str, Union[str, int]]]:
"""
Get list of available summarization models.
"""
return [
{
'name': 'facebook/bart-large-cnn',
'description': 'BART model fine-tuned on CNN news articles (recommended)',
'max_length': 1024
},
{
'name': 't5-small',
'description': 'Small T5 model, faster but less accurate',
'max_length': 512
},
{
'name': 'facebook/bart-base',
'description': 'Base BART model, balanced performance',
'max_length': 1024
}
]
def change_model(self, model_name: str):
"""
Change the summarization model.
Args:
model_name: New model name to use
"""
self.model_name = model_name
self.summarizer = None
self.tokenizer = None
self.model = None
logger.info(f"Model changed to: {model_name}")