scitex 2.14.0__py3-none-any.whl

scitex/io/_load_modules/_pdf.py

@@ -0,0 +1,958 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+# Timestamp: "2025-10-06 10:27:52 (ywatanabe)"
+# File: /home/ywatanabe/proj/scitex_repo/src/scitex/io/_load_modules/_pdf.py
+# ----------------------------------------
+from __future__ import annotations
+import os
+
+__FILE__ = __file__
+__DIR__ = os.path.dirname(__FILE__)
+# ----------------------------------------
+
+"""
+Enhanced PDF loading module with comprehensive extraction capabilities.
+
+This module provides advanced PDF extraction for scientific papers, including:
+- Text extraction with formatting preservation
+- Table extraction as pandas DataFrames
+- Image extraction with metadata
+- Section-aware text parsing
+- Multiple extraction modes for different use cases
+"""
+
+import hashlib
+import re
+import tempfile
+from typing import Any, Dict, List
+
+from scitex import logging
+from scitex.dict import DotDict
+
+logger = logging.getLogger(__name__)
+
+# Try to import PDF libraries in order of preference
+try:
+    import fitz  # PyMuPDF - preferred for text and images
+
+    FITZ_AVAILABLE = True
+except ImportError:
+    FITZ_AVAILABLE = False
+
+try:
+    import pdfplumber  # Best for table extraction
+
+    PDFPLUMBER_AVAILABLE = True
+except ImportError:
+    PDFPLUMBER_AVAILABLE = False
+
+try:
+    import PyPDF2  # Fallback option
+
+    PYPDF2_AVAILABLE = True
+except ImportError:
+    PYPDF2_AVAILABLE = False
+
+try:
+    import pandas as pd
+
+    PANDAS_AVAILABLE = True
+except ImportError:
+    PANDAS_AVAILABLE = False
+
+
+def _load_pdf(lpath: str, mode: str = "full", metadata: bool = False, **kwargs) -> Any:
+    """
+    Load PDF file with comprehensive extraction capabilities.
+
+    Args:
+        lpath: Path to PDF file
+        mode: Extraction mode (default: 'full')
+            - 'full': Complete extraction including text, sections, metadata, pages, tables, and images
+            - 'scientific': Optimized for scientific papers (text + sections + tables + images + stats)
+            - 'text': Plain text extraction only
+            - 'sections': Section-aware text extraction
+            - 'tables': Extract tables as DataFrames
+            - 'images': Extract images with metadata
+            - 'metadata': PDF metadata only
+            - 'pages': Page-by-page extraction
+        metadata: If True, return (result, metadata_dict) tuple for API consistency with images.
+            If False (default), return result only. (default: False)
+        **kwargs: Additional arguments
+            - backend: 'auto' (default), 'fitz', 'pdfplumber', or 'pypdf2'
+            - clean_text: Clean extracted text (default: True)
+            - extract_images: Extract images to files (default: False for 'full' mode, True for 'scientific')
+            - output_dir: Directory for extracted images/tables (default: temp dir)
+            - save_as_jpg: Convert all extracted images to JPG format (default: True)
+            - table_settings: Dict of pdfplumber table extraction settings
+
+    Returns:
+        Extracted content based on mode and metadata parameter:
+
+        When metadata=False (default):
+            - 'text': str
+            - 'sections': Dict[str, str]
+            - 'tables': Dict[int, List[pd.DataFrame]]
+            - 'images': List[Dict] with image metadata
+            - 'metadata': Dict with PDF metadata
+            - 'pages': List[Dict] with page content
+            - 'full': Dict with comprehensive extraction
+            - 'scientific': Dict with scientific paper extraction
+
+        When metadata=True:
+            - Returns: (result, metadata_dict) tuple
+            - metadata_dict contains embedded scitex metadata from PDF Subject field
+
+    Examples:
+        >>> import scitex.io as stx
+
+        >>> # Full extraction (default) - everything included
+        >>> data = stx.load("paper.pdf")
+        >>> print(data['full_text'])      # Complete text
+        >>> print(data['metadata'])       # PDF metadata
+
+        >>> # With metadata tuple (consistent with images)
+        >>> data, meta = stx.load("paper.pdf", metadata=True)
+        >>> print(meta['scitex']['version'])  # Embedded scitex metadata
+
+        >>> # Scientific mode
+        >>> paper = stx.load("paper.pdf", mode="scientific")
+        >>> print(paper['sections'])
+
+        >>> # Simple text extraction
+        >>> text = stx.load("paper.pdf", mode="text")
+    """
+    mode = kwargs.get("mode", mode)
+    backend = kwargs.get("backend", "auto")
+    clean_text = kwargs.get("clean_text", True)
+    extract_images = kwargs.get("extract_images", False)
+    output_dir = kwargs.get("output_dir", None)
+    table_settings = kwargs.get("table_settings", {})
+
+    # Validate file exists
+    if not os.path.exists(lpath):
+        raise FileNotFoundError(f"PDF file not found: {lpath}")
+
+    # Extension validation removed - handled by load() function
+    # This allows loading files without extensions when ext='pdf' is specified
+
+    # Select backend based on mode and availability
+    backend = _select_backend(mode, backend)
+
+    # Create output directory if needed
+    if output_dir is None and (
+        extract_images or mode in ["images", "scientific", "full"]
+    ):
+        output_dir = tempfile.mkdtemp(prefix="pdf_extract_")
+        logger.debug(f"Using temporary directory: {output_dir}")
+
+    # Extract based on mode
+    if mode == "text":
+        result = _extract_text(lpath, backend, clean_text)
+    elif mode == "sections":
+        result = _extract_sections(lpath, backend, clean_text)
+    elif mode == "tables":
+        result = _extract_tables(lpath, table_settings)
+    elif mode == "images":
+        save_as_jpg = kwargs.get("save_as_jpg", True)
+        result = _extract_images(lpath, output_dir, save_as_jpg)
+    elif mode == "metadata":
+        result = _extract_metadata(lpath, backend)
+    elif mode == "pages":
+        result = _extract_pages(lpath, backend, clean_text)
+    elif mode == "scientific":
+        save_as_jpg = kwargs.get("save_as_jpg", True)
+        result = _extract_scientific(
+            lpath, clean_text, output_dir, table_settings, save_as_jpg
+        )
+    elif mode == "full":
+        save_as_jpg = kwargs.get("save_as_jpg", True)
+        result = _extract_full(
+            lpath,
+            backend,
+            clean_text,
+            extract_images,
+            output_dir,
+            table_settings,
+            save_as_jpg,
+        )
+    else:
+        raise ValueError(f"Unknown extraction mode: {mode}")
+
+    # If metadata parameter is True, return tuple (result, metadata_dict)
+    # This provides API consistency with image loading
+    if metadata:
+        try:
+            from .._metadata import read_metadata
+
+            metadata_dict = read_metadata(lpath)
+            return result, metadata_dict
+        except Exception:
+            # If metadata extraction fails, return with None
+            return result, None
+
+    return result
+
+
+def _select_backend(mode: str, requested: str) -> str:
+    """Select appropriate backend based on mode and availability."""
+    if requested != "auto":
+        return requested
+
+    # Mode-specific backend selection
+    if mode in ["tables"]:
+        if PDFPLUMBER_AVAILABLE:
+            return "pdfplumber"
+        else:
+            logger.warning(
+                "pdfplumber not available for table extraction. Install with: pip install pdfplumber"
+            )
+            return "fitz" if FITZ_AVAILABLE else "pypdf2"
+
+    elif mode in ["images", "scientific", "full"]:
+        if FITZ_AVAILABLE:
+            return "fitz"
+        else:
+            logger.warning(
+                "PyMuPDF (fitz) recommended for image extraction. Install with: pip install PyMuPDF"
+            )
+            return "pdfplumber" if PDFPLUMBER_AVAILABLE else "pypdf2"
+
+    else:  # text, sections, metadata, pages
+        if FITZ_AVAILABLE:
+            return "fitz"
+        elif PDFPLUMBER_AVAILABLE:
+            return "pdfplumber"
+        elif PYPDF2_AVAILABLE:
+            return "pypdf2"
+        else:
+            raise ImportError(
+                "No PDF library available. Install one of:\n"
+                "  pip install PyMuPDF     # Recommended\n"
+                "  pip install pdfplumber  # Best for tables\n"
+                "  pip install PyPDF2      # Basic fallback"
+            )
+
+
+def _extract_text(lpath: str, backend: str, clean: bool) -> str:
+    """Extract plain text from PDF."""
+    if backend == "fitz":
+        return _extract_text_fitz(lpath, clean)
+    elif backend == "pdfplumber":
+        return _extract_text_pdfplumber(lpath, clean)
+    else:
+        return _extract_text_pypdf2(lpath, clean)
+
+
+def _extract_text_fitz(lpath: str, clean: bool) -> str:
+    """Extract text using PyMuPDF."""
+    if not FITZ_AVAILABLE:
+        raise ImportError("PyMuPDF (fitz) not available")
+
+    try:
+        doc = fitz.open(lpath)
+        text_parts = []
+
+        for page_num, page in enumerate(doc):
+            text = page.get_text()
+            if text.strip():
+                text_parts.append(text)
+
+        doc.close()
+
+        full_text = "\n".join(text_parts)
+
+        if clean:
+            full_text = _clean_pdf_text(full_text)
+
+        return full_text
+
+    except Exception as e:
+        logger.error(f"Error extracting text with fitz from {lpath}: {e}")
+        raise
+
+
+def _extract_text_pdfplumber(lpath: str, clean: bool) -> str:
+    """Extract text using pdfplumber."""
+    if not PDFPLUMBER_AVAILABLE:
+        raise ImportError("pdfplumber not available")
+
+    try:
+        import pdfplumber
+
+        text_parts = []
+        with pdfplumber.open(lpath) as pdf:
+            for page in pdf.pages:
+                text = page.extract_text()
+                if text:
+                    text_parts.append(text)
+
+        full_text = "\n".join(text_parts)
+
+        if clean:
+            full_text = _clean_pdf_text(full_text)
+
+        return full_text
+
+    except Exception as e:
+        logger.error(f"Error extracting text with pdfplumber from {lpath}: {e}")
+        raise
+
+
+def _extract_text_pypdf2(lpath: str, clean: bool) -> str:
+    """Extract text using PyPDF2."""
+    if not PYPDF2_AVAILABLE:
+        raise ImportError("PyPDF2 not available")
+
+    try:
+        reader = PyPDF2.PdfReader(lpath)
+        text_parts = []
+
+        for page_num in range(len(reader.pages)):
+            page = reader.pages[page_num]
+            text = page.extract_text()
+            if text.strip():
+                text_parts.append(text)
+
+        full_text = "\n".join(text_parts)
+
+        if clean:
+            full_text = _clean_pdf_text(full_text)
+
+        return full_text
+
+    except Exception as e:
+        logger.error(f"Error extracting text with PyPDF2 from {lpath}: {e}")
+        raise
+
+
+def _extract_tables(
+    lpath: str, table_settings: Dict = None
+) -> Dict[int, List["pd.DataFrame"]]:
+    """
+    Extract tables from PDF as pandas DataFrames.
+
+    Returns:
+        Dict mapping page numbers to list of DataFrames
+    """
+    if not PDFPLUMBER_AVAILABLE:
+        raise ImportError(
+            "pdfplumber required for table extraction. Install with:\n"
+            "  pip install pdfplumber pandas"
+        )
+
+    if not PANDAS_AVAILABLE:
+        raise ImportError("pandas required for table extraction")
+
+    import pandas as pd
+    import pdfplumber
+
+    tables_dict = {}
+    table_settings = table_settings or {}
+
+    try:
+        with pdfplumber.open(lpath) as pdf:
+            for page_num, page in enumerate(pdf.pages):
+                # Extract tables from page
+                tables = page.extract_tables(**table_settings)
+
+                if tables:
+                    # Convert to DataFrames
+                    dfs = []
+                    for table in tables:
+                        if table and len(table) > 0:
+                            # First row as header if it looks like headers
+                            if len(table) > 1 and all(
+                                isinstance(cell, str) for cell in table[0] if cell
+                            ):
+                                df = pd.DataFrame(table[1:], columns=table[0])
+                            else:
+                                df = pd.DataFrame(table)
+
+                            # Clean up DataFrame
+                            df = (
+                                df.replace("", None)
+                                .dropna(how="all", axis=1)
+                                .dropna(how="all", axis=0)
+                            )
+
+                            if not df.empty:
+                                dfs.append(df)
+
+                    if dfs:
+                        tables_dict[page_num] = dfs
+
+        logger.info(f"Extracted tables from {len(tables_dict)} pages")
+        return tables_dict
+
+    except Exception as e:
+        logger.error(f"Error extracting tables: {e}")
+        raise
+
+
+def _extract_images(
+    lpath: str, output_dir: str = None, save_as_jpg: bool = True
+) -> List[Dict[str, Any]]:
+    """
+    Extract images from PDF with metadata.
+
+    Args:
+        lpath: Path to PDF file
+        output_dir: Directory to save images (optional)
+        save_as_jpg: If True, convert all images to JPG format (default: True)
+
+    Returns:
+        List of dicts containing image metadata and paths
+    """
+    if not FITZ_AVAILABLE:
+        raise ImportError(
+            "PyMuPDF (fitz) required for image extraction. Install with:\n"
+            "  pip install PyMuPDF"
+        )
+
+    images_info = []
+
+    try:
+        doc = fitz.open(lpath)
+
+        for page_num, page in enumerate(doc):
+            image_list = page.get_images()
+
+            for img_index, img in enumerate(image_list):
+                xref = img[0]
+
+                # Extract image data
+                base_image = doc.extract_image(xref)
+                image_bytes = base_image["image"]
+                original_ext = base_image["ext"]
+
+                image_info = {
+                    "page": page_num + 1,
+                    "index": img_index,
+                    "width": base_image["width"],
+                    "height": base_image["height"],
+                    "colorspace": base_image["colorspace"],
+                    "bpc": base_image["bpc"],  # bits per component
+                    "original_ext": original_ext,
+                    "size_bytes": len(image_bytes),
+                }
+
+                # Save image if output directory provided
+                if output_dir:
+                    os.makedirs(output_dir, exist_ok=True)
+
+                    if save_as_jpg and original_ext not in ["jpg", "jpeg"]:
+                        # Convert to JPG using PIL
+                        try:
+                            from PIL import Image
+                            import io
+
+                            # Open image from bytes
+                            img_pil = Image.open(io.BytesIO(image_bytes))
+
+                            # Convert RGBA to RGB if necessary
+                            if img_pil.mode in ("RGBA", "LA", "P"):
+                                # Create a white background
+                                background = Image.new(
+                                    "RGB", img_pil.size, (255, 255, 255)
+                                )
+                                if img_pil.mode == "P":
+                                    img_pil = img_pil.convert("RGBA")
+                                background.paste(
+                                    img_pil,
+                                    mask=img_pil.split()[-1]
+                                    if img_pil.mode == "RGBA"
+                                    else None,
+                                )
+                                img_pil = background
+                            elif img_pil.mode != "RGB":
+                                img_pil = img_pil.convert("RGB")
+
+                            # Save as JPG
+                            filename = f"page_{page_num + 1}_img_{img_index}.jpg"
+                            filepath = os.path.join(output_dir, filename)
+                            img_pil.save(filepath, "JPEG", quality=95)
+
+                            image_info["ext"] = "jpg"
+                        except ImportError:
+                            logger.warning(
+                                "PIL not available for image conversion. Install with: pip install Pillow"
+                            )
+                            # Fall back to original format
+                            filename = (
+                                f"page_{page_num + 1}_img_{img_index}.{original_ext}"
+                            )
+                            filepath = os.path.join(output_dir, filename)
+                            with open(filepath, "wb") as img_file:
+                                img_file.write(image_bytes)
+                            image_info["ext"] = original_ext
+                    else:
+                        # Save with original format
+                        ext = "jpg" if original_ext == "jpeg" else original_ext
+                        filename = f"page_{page_num + 1}_img_{img_index}.{ext}"
+                        filepath = os.path.join(output_dir, filename)
+                        with open(filepath, "wb") as img_file:
+                            img_file.write(image_bytes)
+                        image_info["ext"] = ext
+
+                    image_info["filepath"] = filepath
+                    image_info["filename"] = filename
+
+                images_info.append(image_info)
+
+        doc.close()
+
+        logger.info(f"Extracted {len(images_info)} images from PDF")
+        return images_info
+
+    except Exception as e:
+        logger.error(f"Error extracting images: {e}")
+        raise
+
+
+def _extract_sections(lpath: str, backend: str, clean: bool) -> Dict[str, str]:
+    """Extract text organized by sections."""
+    # Get full text first
+    text = _extract_text(lpath, backend, clean=False)
+
+    # Parse into sections
+    sections = _parse_sections(text)
+
+    # Clean section text if requested
+    if clean:
+        for section, content in sections.items():
+            sections[section] = _clean_pdf_text(content)
+
+    return sections
+
+
+def _parse_sections(text: str) -> Dict[str, str]:
+    """
+    Parse text into sections based on IMRaD structure.
+
+    Follows the standard scientific paper structure:
+    - frontpage: Title, authors, affiliations, keywords
+    - abstract: Paper summary
+    - introduction: Background and motivation
+    - methods: Methodology (materials and methods, experimental design)
+    - results: Findings
+    - discussion: Interpretation and implications
+    - references: Citations
+    """
+    sections = {}
+    current_section = "frontpage"
+    current_text = []
+
+    # Simplified section patterns - IMRaD + frontpage only
+    # Only match standalone section headers (exact matches)
+    section_patterns = [
+        r"^abstract\s*$",
+        r"^summary\s*$",
+        r"^introduction\s*$",
+        r"^background\s*$",
+        r"^methods?\s*$",
+        r"^materials?\s+and\s+methods?\s*$",
+        r"^methodology\s*$",
+        r"^results?\s*$",
+        r"^discussion\s*$",
+        r"^references?\s*$",
+    ]
+
+    lines = text.split("\n")
+
+    for line in lines:
+        line_lower = line.lower().strip()
+        line_stripped = line.strip()
+
+        # Check if this line is a section header
+        is_header = False
+        for pattern in section_patterns:
+            if re.match(pattern, line_lower):
+                # Additional validation: header lines should be short (< 50 chars)
+                # and not contain numbers/punctuation (except spaces)
+                if len(line_stripped) < 50:
+                    # Save previous section
+                    if current_text:
+                        sections[current_section] = "\n".join(current_text).strip()
+
+                    # Start new section
+                    current_section = line_lower.strip()
+                    current_text = []
+                    is_header = True
+                    break
+
+        if not is_header:
+            current_text.append(line)
+
+    # Save last section
+    if current_text:
+        sections[current_section] = "\n".join(current_text).strip()
+
+    return sections
+
+
+def _extract_metadata(lpath: str, backend: str) -> Dict[str, Any]:
+    """Extract PDF metadata."""
+    metadata = {
+        "file_path": lpath,
+        "file_name": os.path.basename(lpath),
+        "file_size": os.path.getsize(lpath),
+        "backend": backend,
+    }
+
+    if backend == "fitz" and FITZ_AVAILABLE:
+        try:
+            doc = fitz.open(lpath)
+            pdf_metadata = doc.metadata
+
+            metadata.update(
+                {
+                    "title": pdf_metadata.get("title", ""),
+                    "author": pdf_metadata.get("author", ""),
+                    "subject": pdf_metadata.get("subject", ""),
+                    "keywords": pdf_metadata.get("keywords", ""),
+                    "creator": pdf_metadata.get("creator", ""),
+                    "producer": pdf_metadata.get("producer", ""),
+                    "creation_date": str(pdf_metadata.get("creationDate", "")),
+                    "modification_date": str(pdf_metadata.get("modDate", "")),
+                    "pages": len(doc),
+                    "encrypted": doc.is_encrypted,
+                }
+            )
+
+            # Try to parse scitex metadata from subject field (for consistency with PNG)
+            subject = pdf_metadata.get("subject", "")
+            if subject:
+                try:
+                    import json
+
+                    # Check if subject is JSON (scitex metadata)
+                    parsed_subject = json.loads(subject)
+                    if isinstance(parsed_subject, dict):
+                        # Merge parsed scitex metadata with standard PDF metadata
+                        # This makes PDF metadata format consistent with PNG
+                        metadata.update(parsed_subject)
+                        # Remove the raw JSON string from subject to avoid duplication
+                        metadata.pop("subject", None)
+                except (json.JSONDecodeError, ValueError):
+                    # Not JSON, keep subject as string
+                    pass
+
+            doc.close()
+
+        except Exception as e:
+            logger.error(f"Error extracting metadata with fitz: {e}")
+
+    elif backend == "pdfplumber" and PDFPLUMBER_AVAILABLE:
+        try:
+            import pdfplumber
+
+            with pdfplumber.open(lpath) as pdf:
+                metadata["pages"] = len(pdf.pages)
+                if hasattr(pdf, "metadata"):
+                    metadata.update(pdf.metadata)
+
+                # Try to parse scitex metadata from subject field (for consistency with PNG)
+                if "Subject" in metadata or "subject" in metadata:
+                    subject = metadata.get("Subject") or metadata.get("subject", "")
+                    if subject:
+                        try:
+                            import json
+
+                            parsed_subject = json.loads(subject)
+                            if isinstance(parsed_subject, dict):
+                                # Merge parsed scitex metadata with standard PDF metadata
+                                metadata.update(parsed_subject)
+                                # Remove the raw JSON string from subject to avoid duplication
+                                metadata.pop("Subject", None)
+                                metadata.pop("subject", None)
+                        except (json.JSONDecodeError, ValueError):
+                            # Not JSON, keep subject as string
+                            pass
+        except Exception as e:
+            logger.error(f"Error extracting metadata with pdfplumber: {e}")
+
+    elif backend == "pypdf2" and PYPDF2_AVAILABLE:
+        try:
+            reader = PyPDF2.PdfReader(lpath)
+
+            if reader.metadata:
+                metadata.update(
+                    {
+                        "title": reader.metadata.get("/Title", ""),
+                        "author": reader.metadata.get("/Author", ""),
+                        "subject": reader.metadata.get("/Subject", ""),
+                        "creator": reader.metadata.get("/Creator", ""),
+                        "producer": reader.metadata.get("/Producer", ""),
+                        "creation_date": str(reader.metadata.get("/CreationDate", "")),
+                        "modification_date": str(reader.metadata.get("/ModDate", "")),
+                    }
+                )
+
+            metadata["pages"] = len(reader.pages)
+            metadata["encrypted"] = reader.is_encrypted
+
+            # Try to parse scitex metadata from subject field (for consistency with PNG)
+            subject = metadata.get("subject", "")
+            if subject:
+                try:
+                    import json
+
+                    parsed_subject = json.loads(subject)
+                    if isinstance(parsed_subject, dict):
+                        # Merge parsed scitex metadata with standard PDF metadata
+                        metadata.update(parsed_subject)
+                        # Remove the raw JSON string from subject to avoid duplication
+                        metadata.pop("subject", None)
+                except (json.JSONDecodeError, ValueError):
+                    # Not JSON, keep subject as string
+                    pass
+
+        except Exception as e:
+            logger.error(f"Error extracting metadata with PyPDF2: {e}")
+
+    # Generate file hash
+    metadata["md5_hash"] = _calculate_file_hash(lpath)
+
+    return metadata
+
+
+def _extract_pages(lpath: str, backend: str, clean: bool) -> List[Dict[str, Any]]:
+    """Extract content page by page."""
+    pages = []
+
+    if backend == "fitz" and FITZ_AVAILABLE:
+        doc = fitz.open(lpath)
+
+        for page_num, page in enumerate(doc):
+            text = page.get_text()
+            if clean:
+                text = _clean_pdf_text(text)
+
+            pages.append(
+                {
+                    "page_number": page_num + 1,
+                    "text": text,
+                    "char_count": len(text),
+                    "word_count": len(text.split()),
+                }
+            )
+
+        doc.close()
+
+    elif backend == "pdfplumber" and PDFPLUMBER_AVAILABLE:
+        import pdfplumber
+
+        with pdfplumber.open(lpath) as pdf:
+            for page_num, page in enumerate(pdf.pages):
+                text = page.extract_text() or ""
+                if clean:
+                    text = _clean_pdf_text(text)
+
+                pages.append(
+                    {
+                        "page_number": page_num + 1,
+                        "text": text,
+                        "char_count": len(text),
+                        "word_count": len(text.split()),
+                    }
+                )
+
+    elif backend == "pypdf2" and PYPDF2_AVAILABLE:
+        reader = PyPDF2.PdfReader(lpath)
+
+        for page_num in range(len(reader.pages)):
+            page = reader.pages[page_num]
+            text = page.extract_text()
+            if clean:
+                text = _clean_pdf_text(text)
+
+            pages.append(
+                {
+                    "page_number": page_num + 1,
+                    "text": text,
+                    "char_count": len(text),
+                    "word_count": len(text.split()),
+                }
+            )
+
+    return pages
+
+
+def _extract_scientific(
+    lpath: str,
+    clean_text: bool,
+    output_dir: str,
+    table_settings: Dict,
+    save_as_jpg: bool = True,
+) -> DotDict:
+    """
+    Optimized extraction for scientific papers.
+    Extracts text, tables, images, and sections in a structured format.
+    """
+    result = {
+        "pdf_path": lpath,
+        "filename": os.path.basename(lpath),
+        "extraction_mode": "scientific",
+    }
+
+    try:
+        # Extract text and sections
+        backend = _select_backend("text", "auto")
+        result["text"] = _extract_text(lpath, backend, clean_text)
+        result["sections"] = _extract_sections(lpath, backend, clean_text)
+
+        # Extract metadata
+        result["metadata"] = _extract_metadata(lpath, backend)
+
+        # Extract tables if pdfplumber available
+        if PDFPLUMBER_AVAILABLE and PANDAS_AVAILABLE:
+            try:
+                result["tables"] = _extract_tables(lpath, table_settings)
+            except Exception as e:
+                logger.warning(f"Could not extract tables: {e}")
+                result["tables"] = {}
+        else:
+            result["tables"] = {}
+            logger.info("Table extraction requires pdfplumber and pandas")
+
+        # Extract images if fitz available
+        if FITZ_AVAILABLE:
+            try:
+                result["images"] = _extract_images(lpath, output_dir, save_as_jpg)
+            except Exception as e:
+                logger.warning(f"Could not extract images: {e}")
+                result["images"] = []
+        else:
+            result["images"] = []
+            logger.info("Image extraction requires PyMuPDF (fitz)")
+
+        # Calculate statistics
+        result["stats"] = {
+            "total_chars": len(result["text"]),
+            "total_words": len(result["text"].split()),
+            "total_pages": result["metadata"].get("pages", 0),
+            "num_sections": len(result["sections"]),
+            "num_tables": sum(len(tables) for tables in result["tables"].values()),
+            "num_images": len(result["images"]),
+        }
+
+        logger.info(
+            f"Scientific extraction complete: "
+            f"{result['stats']['total_pages']} pages, "
+            f"{result['stats']['num_sections']} sections, "
+            f"{result['stats']['num_tables']} tables, "
+            f"{result['stats']['num_images']} images"
+        )
+
+    except Exception as e:
+        logger.error(f"Error in scientific extraction: {e}")
+        result["error"] = str(e)
+
+    return DotDict(result)
+
+
+def _extract_full(
+    lpath: str,
+    backend: str,
+    clean: bool,
+    extract_images: bool,
+    output_dir: str,
+    table_settings: Dict,
+    save_as_jpg: bool = True,
+) -> DotDict:
+    """Extract comprehensive data from PDF."""
+    result = {
+        "pdf_path": lpath,
+        "filename": os.path.basename(lpath),
+        "backend": backend,
+        "extraction_params": {
+            "clean_text": clean,
+            "extract_images": extract_images,
+        },
+    }
+
+    # Extract all components
+    try:
+        result["full_text"] = _extract_text(lpath, backend, clean)
+        result["sections"] = _extract_sections(lpath, backend, clean)
+        result["metadata"] = _extract_metadata(lpath, backend)
+        result["pages"] = _extract_pages(lpath, backend, clean)
+
+        # Extract tables if available
+        if PDFPLUMBER_AVAILABLE and PANDAS_AVAILABLE:
+            try:
+                result["tables"] = _extract_tables(lpath, table_settings)
+            except Exception as e:
+                logger.warning(f"Could not extract tables: {e}")
+                result["tables"] = {}
+
+        # Extract images if requested and available
+        if extract_images and FITZ_AVAILABLE:
+            try:
+                result["images"] = _extract_images(lpath, output_dir, save_as_jpg)
+            except Exception as e:
+                logger.warning(f"Could not extract images: {e}")
+                result["images"] = []
+
+        # Calculate statistics
+        result["stats"] = {
+            "total_chars": len(result["full_text"]),
+            "total_words": len(result["full_text"].split()),
+            "total_pages": len(result["pages"]),
+            "num_sections": len(result["sections"]),
+            "num_tables": sum(
+                len(tables) for tables in result.get("tables", {}).values()
+            ),
+            "num_images": len(result.get("images", [])),
+            "avg_words_per_page": (
+                len(result["full_text"].split()) / len(result["pages"])
+                if result["pages"]
+                else 0
+            ),
+        }
+
+    except Exception as e:
+        logger.error(f"Error in full extraction: {e}")
+        result["error"] = str(e)
+
+    return DotDict(result)
+
+
+def _clean_pdf_text(text: str) -> str:
+    """Clean extracted PDF text."""
+    # Remove excessive whitespace
+    text = re.sub(r"\s+", " ", text)
+
+    # Fix hyphenated words at line breaks
+    text = re.sub(r"(\w+)-\s*\n\s*(\w+)", r"\1\2", text)
+
+    # Remove page numbers (common patterns)
+    text = re.sub(r"\n\s*\d+\s*\n", "\n", text)
+    text = re.sub(r"Page\s+\d+\s+of\s+\d+", "", text, flags=re.IGNORECASE)
+
+    # Clean up common PDF artifacts
+    text = text.replace("\x00", "")  # Null bytes
+    text = re.sub(r"[\x01-\x1f\x7f-\x9f]", "", text)  # Control characters
+
+    # Normalize quotes and dashes
+    text = text.replace('"', '"').replace('"', '"')
+    text = text.replace(""", "'").replace(""", "'")
+    text = text.replace("–", "-").replace("—", "-")
+
+    # Remove multiple consecutive newlines
+    text = re.sub(r"\n{3,}", "\n\n", text)
+
+    return text.strip()
+
+
+def _calculate_file_hash(lpath: str) -> str:
+    """Calculate MD5 hash of file."""
+    hash_md5 = hashlib.md5()
+    with open(lpath, "rb") as f:
+        for chunk in iter(lambda: f.read(4096), b""):
+            hash_md5.update(chunk)
+    return hash_md5.hexdigest()
+
+
+# EOF