📚 Research Support

This section lists key research papers supporting the technical and theoretical foundation of the AI-Powered Taxation Assistant project.

🧠 Core RAG & LLM Papers

#	Paper	Year	Source
1	Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks — Lewis et al., NeurIPS 2020	2020	NeurIPS
2	Retrieval-Augmented Generation for Large Language Models: A Survey — Gao et al., arXiv	2023	arXiv
3	CBR-RAG: Case-Based Reasoning for Retrieval-Augmented Generation — Wang et al., arXiv	2024	arXiv

📊 Financial Document Analysis

#	Paper	Year	Source
4	TableNet: Deep Learning Model for End-to-End Table Detection and Structure Recognition	2020	arXiv
5	Graph Neural Networks for Table Detection in Invoices — Riba et al., ICDAR	2019	arXiv
6	PdfTable: A Unified Toolkit for Table Extraction from Financial Documents	2024	arXiv

💼 Domain Models & Financial NLP

#	Paper	Year	Source
7	FinBERT: Financial Sentiment Analysis with Pre-trained Language Models — Araci	2020	arXiv
8	Survey of Deep Learning Methods for Document and Table Detection	2024	ACM Digital Library

⚖️ Explainability & Ethics

#	Paper	Year	Source
9	Explainable Artificial Intelligence (XAI) in Auditing — Parker, Cho & Vasarhelyi	2022	Elsevier
10	Exploring Explainable AI in the Tax Domain — Górski	2024	SSRN

⚙️ Applied RAG in Legal & Financial Contexts

#	Paper	Year	Source
11	Interpretable Long-Form Legal QA with Retrieval-Augmented Generation	2024	arXiv
12	RAG-Enhanced Evidence Recommendation in Financial Contexts	2025	ACM Digital Library

🧩 How to Use These Papers

RAG Pipeline: Follow Lewis (2020) and Gao (2023) for architecture design.
Extraction Models: TableNet and PdfTable support PDF parsing and OCR integration.
Domain Tuning: Fine-tune FinBERT for tax/financial contexts.
Explainability: Integrate XAI patterns from Parker (2022) and Górski (2024).
Evaluation: Use factuality, retrieval precision, and expert validation metrics.

🔗 Quick Access