Structured Table Extraction And Reconstruction With Llm Enhancement

PDF to Markdown converter with deep learning.

Unique: Combines heuristic cell alignment with optional LLM-based refinement — uses spatial analysis to reconstruct table structure, then optionally invokes LLMs to correct misaligned cells or infer missing content. Supports pluggable LLM services (OpenAI, Anthropic, local models) for accuracy tuning without rewriting extraction logic.

vs others: More accurate than regex-based table extraction; supports LLM refinement unlike pure heuristic tools; better handling of merged cells than simple grid-based approaches.