PDF Table Extraction & Multi-Extractor Fusion

This project focuses on extracting and merging table data from PDF documents using multiple OCR and parsing frameworks. Since the results of the extraction processes often differ in format and structure and frequently contain errors, the project addresses the challenges of data normalization and quality improvement. It explores ideas for transforming (possibly invalid) LaTeX into Markdown, leveraging AI to normalize heterogeneous Markdown table structures, and interpreting and merging the extracted data through voting. The overall goal is to explore how AI can be used to reliably handle complex table structures, including suitable training data, machine learning strategies, and robust normalization approaches.