Datasets:
Size:
1M<n<10M
ArXiv:
Tags:
Document_Understanding
Document_Packet_Splitting
Document_Comprehension
Document_Classification
Document_Recognition
Document_Segmentation
DOI:
License:
File size: 4,960 Bytes
165da3c | 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 | # Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
# SPDX-License-Identifier: CC-BY-NC-4.0
"""
Create Assets from Raw PDFs
Processes raw PDFs and creates structured assets:
- Page images (PNG at 300 DPI)
- OCR text (AWS Textract)
Usage:
python main.py
python main.py --workers 20 --limit 100
"""
import os
import argparse
from loguru import logger
from services.pdf_loader import PdfLoader
from services.textract_ocr import TextractOcr
from services.deepseek_ocr import DeepSeekOcr
from services.asset_writer import AssetWriter
from services.asset_creator import AssetCreator
def main():
parser = argparse.ArgumentParser(description='Create assets from raw PDFs')
parser.add_argument('--raw-data-path', default='../raw_data')
parser.add_argument('--output-path', default='../processed_assets')
parser.add_argument('--metadata-path', default='../metadata')
parser.add_argument('--workers', type=int, default=10)
parser.add_argument('--limit', type=int, default=None)
parser.add_argument('--save-mapping', action='store_true')
parser.add_argument('--use-deepseek-for-language', action='store_true', help='Use DeepSeek OCR for language docs (default: Textract)')
parser.add_argument('--s3-bucket', default=os.getenv('DOCSPLIT_S3_BUCKET'), help='S3 bucket for Textract temporary uploads')
parser.add_argument('--s3-prefix', default='textract-temp', help='S3 prefix for uploads')
args = parser.parse_args()
logger.info("Creating assets from PDFs")
# Load all PDFs
loader = PdfLoader(raw_data_path=args.raw_data_path)
documents = loader.get_all_documents()
successful_docs = []
if args.use_deepseek_for_language:
# Separate language documents for DeepSeek processing
language_docs = [doc for doc in documents if doc.doc_type == 'language']
other_docs = [doc for doc in documents if doc.doc_type != 'language']
# Process non-language documents with Textract
if other_docs:
logger.info(f"Processing {len(other_docs)} non-language documents with Textract")
ocr = TextractOcr(s3_bucket=args.s3_bucket, s3_prefix=args.s3_prefix)
writer = AssetWriter(output_base_path=args.output_path)
creator = AssetCreator(writer, ocr)
results = creator.create_all(
documents=other_docs,
workers=args.workers,
limit=args.limit
)
logger.info(f"Textract completed: {results}")
if results['failed'] > 0:
logger.warning(f"Textract failed: {results['failed_docs']}")
failed_names = set(results['failed_docs'])
successful_docs.extend([doc for doc in other_docs if doc.doc_name not in failed_names])
# Process language documents with DeepSeek
if language_docs:
logger.info(f"Processing {len(language_docs)} language documents with DeepSeek OCR")
try:
deepseek_ocr = DeepSeekOcr()
writer = AssetWriter(output_base_path=args.output_path)
creator = AssetCreator(writer, deepseek_ocr)
results = creator.create_all(
documents=language_docs,
workers=args.workers,
limit=args.limit
)
logger.info(f"DeepSeek completed: {results}")
if results['failed'] > 0:
logger.warning(f"DeepSeek failed: {results['failed_docs']}")
failed_names = set(results['failed_docs'])
successful_docs.extend([doc for doc in language_docs if doc.doc_name not in failed_names])
except Exception as e:
logger.error(f"DeepSeek OCR initialization failed: {e}")
else:
# Process ALL documents with Textract
logger.info(f"Processing {len(documents)} documents with Textract")
ocr = TextractOcr(s3_bucket=args.s3_bucket, s3_prefix=args.s3_prefix)
writer = AssetWriter(output_base_path=args.output_path)
creator = AssetCreator(writer, ocr)
results = creator.create_all(
documents=documents,
workers=args.workers,
limit=args.limit
)
logger.info(f"Textract completed: {results}")
if results['failed'] > 0:
logger.warning(f"Textract failed: {results['failed_docs']}")
failed_names = set(results['failed_docs'])
successful_docs.extend([doc for doc in documents if doc.doc_name not in failed_names])
# Save mapping AFTER processing with only successful documents
if args.save_mapping:
mapping_path = f"{args.metadata_path}/document_mapping.csv"
loader.save_document_mapping(successful_docs, output_path=mapping_path)
logger.info(f"Saved mapping for {len(successful_docs)} successful documents")
if __name__ == '__main__':
main()
|