Add PDF table extraction tools and requirements (#485)

* Add PDF table extraction tools and requirements

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

Author: pesc101

tools/ocr/__init__.py

@@ -0,0 +1,187 @@
+"""Module for PDF table extraction tools.
+
+This module defines tools for extracting tables from PDFs, converting them to HTML.
+"""
+
+from abc import ABC, abstractmethod
+from typing import Any
+
+from docling.backend.docling_parse_v2_backend import DoclingParseV2DocumentBackend
+from docling.datamodel.base_models import InputFormat
+from docling.datamodel.pipeline_options import (
+    EasyOcrOptions,
+    PdfPipelineOptions,
+    TableFormerMode,
+)
+from docling.document_converter import DocumentConverter, PdfFormatOption
+from docling_core.types.doc.document import TableItem
+from gmft.auto import AutoFormatConfig, AutoTableFormatter
+from gmft.detectors.common import CroppedTable
+from gmft.detectors.tatr import TATRDetector
+from gmft_pymupdf import PyMuPDFDocument
+from img2table.document import PDF
+from img2table.ocr import TesseractOCR
+from unstructured.documents.elements import Table
+from unstructured.partition.pdf import partition_pdf
+
+
+class ToolBase(ABC):
+    """Abstract base class for PDF table extraction tools.
+
+    Defines the interface that all tools must implement.
+    """
+
+    @abstractmethod
+    def extract_tables(self, pdf_file: str) -> Any:
+        """Extract tables from a PDF file and return them in a raw, tool-specific format."""
+        pass
+
+    @abstractmethod
+    def convert_to_html(self, tables: Any) -> list[str]:
+        """Convert the extracted table data to HTML format."""
+        pass
+
+
+class UnstructuredTool(ToolBase):
+    """Tool for extracting tables from PDFs using the Unstructured library."""
+
+    def __init__(self) -> None:
+        pass
+
+    def extract_tables(self, pdf_file: str) -> list[Table]:
+        """Uses Unstructured Tool to extract tables from a PDF file."""
+        elements = partition_pdf(
+            filename=pdf_file,
+            infer_table_structure=True,
+            strategy="hi_res",
+            languages=["eng"],
+            model_name="yolox",
+        )
+        tables: list[Table] = []
+        for el in elements:
+            if isinstance(el, Table):
+                tables.append(el)
+        return tables
+
+    def convert_to_html(self, tables: Table) -> Any:
+        """Convert extracted table data using Unstructured Tool to HTML format."""
+        try:
+            tables_html = tables.metadata.text_as_html
+        except Exception as e:
+            print(f"Error processing table: {e}")
+            tables_html = None
+        return tables_html
+
+
+class GMFTTool(ToolBase):
+    """Tool for extracting tables using the GMFT library."""
+
+    def __init__(self) -> None:
+        self.detector = TATRDetector()
+        config = AutoFormatConfig()
+        config.semantic_spanning_cells = True
+        config.enable_multi_header = True
+        self.formatter = AutoTableFormatter(config)
+
+    def ingest_pdf(self, pdf_path: str) -> list[CroppedTable]:
+        """Opens the PDF with PyMuPDFDocument, iterates over each page.
+
+        Uses GMFT Tool and the TATRDetector to extract tables.
+        """
+        doc = PyMuPDFDocument(pdf_path)
+        tables = []
+        for page in doc:
+            tables.extend(self.detector.extract(page))
+        return tables
+
+    def extract_tables(self, pdf_file: str) -> list[CroppedTable]:
+        """Extract tables from a PDF file using GMFT."""
+        return self.ingest_pdf(pdf_file)
+
+    def convert_to_html(self, tables: list[CroppedTable]) -> Any:
+        """Convert extracted table data using GMFT Tool to HTML format."""
+        ft = self.formatter.extract(tables)
+        try:
+            tables_html = ft.df().fillna("").to_html()
+        except Exception as e:
+            print(f"Error processing table: {e}")
+            tables_html = None
+        return tables_html
+
+
+class Img2TableTool(ToolBase):
+    """Tool for extracting tables from PDFs using the Img2Table library."""
+
+    def __init__(self) -> None:
+        self.ocr = TesseractOCR(n_threads=1, lang="eng")
+
+    def extract_tables(self, pdf_file: str) -> Any:
+        """Uses Img2Table Tool to extract tables from a PDF file."""
+        pdf = PDF(pdf_file, detect_rotation=False, pdf_text_extraction=True)
+        extracted_tables = pdf.extract_tables(
+            ocr=self.ocr, implicit_rows=True, borderless_tables=True, min_confidence=50
+        )
+        return extracted_tables[0]
+
+    def convert_to_html(self, tables: Any) -> Any:
+        """Convert extracted table data using Img2Table Tool to HTML format."""
+        try:
+            tables_html = tables.html_repr()
+        except Exception as e:
+            print(f"Error processing table: {e}")
+            tables_html = None
+        return tables_html
+
+
+class DoclingTool(ToolBase):
+    """Tool for extracting tables from PDFs using the Docling library."""
+
+    def __init__(self) -> None:
+        self.pipeline_options = PdfPipelineOptions(
+            do_table_structure=True, ocr_options=EasyOcrOptions()
+        )
+        self.pipeline_options.table_structure_options.mode = TableFormerMode.ACCURATE
+
+        self.doc_converter = DocumentConverter(
+            format_options={
+                InputFormat.PDF: PdfFormatOption(
+                    pipeline_options=self.pipeline_options,
+                    backend=DoclingParseV2DocumentBackend,
+                )
+            }
+        )
+
+    def extract_tables(self, pdf_file: str) -> list[TableItem]:
+        """Uses Docling tool to extract tables from the given PDF file.
+
+        Returns a list of table objects (Docling-specific).
+        """
+        result = self.doc_converter.convert(pdf_file)
+
+        tables = result.document.tables
+
+        return tables
+
+    def convert_to_html(self, tables: Any) -> Any:
+        """Convert the Docling table objects to HTML strings."""
+        try:
+            tables_html = tables.export_to_html()
+        except Exception as e:
+            print(f"Error processing tables: {e}")
+            return []
+        return tables_html
+
+
+def initialize_tools(tools: list[str] = ["all"]) -> dict:
+    """Initialize and return selected table extraction tools. Default is all tools."""
+    available_tools = {
+        "unstructured": UnstructuredTool(),
+        "gmft": GMFTTool(),
+        "img2table": Img2TableTool(),
+        "docling": DoclingTool(),
+    }
+
+    if tools == ["all"]:
+        return available_tools
+
+    return {name: tool for name, tool in available_tools.items() if name in tools}

tools/ocr/pdf_tools.py

@@ -0,0 +1,4 @@
+docling>=2.15.1
+gmft>=0.4.0
+img2table>=1.4.0
+unstructured[all-docs]>=0.14.10