Abstract: The project focuses on enhancing Document Question Answering (DocQA) systems in the financial sector through the integration of the fine-tuned DoNUT model. This model enables swift and accurate extraction of crucial information from various financial documents, facilitating faster decision-making and regulatory compliance. Through training and fine-tuning on diverse synthetic financial datasets, the proposed system aims for high precision and recall in extracting key financial information from the forms. Empirical evaluations and case studies within financial institutions aim to quantify the time savings and efficiency gains achieved by AI-driven DocQA systems, highlighting the tangible benefits of the DoNUT-enhanced model. Ultimately, this research underscores the transformative potential of AI-driven document understanding in the financial sector, emphasizing the importance of sophisticated AI models for improving operational efficiency, regulatory compliance.

Keywords: Document Question Answering (DocQA), DoNUT, artificial intelligence (AI), key-value pair extraction.

| DOI: 10.17148/IARJSET.2024.11515