DOCMASTER: A Unified Platform for Annotation, Training, & Inference in Document Question-Answering
This addresses the problem of complex and privacy-lacking tools for businesses and researchers working with PDF documents, though it appears incremental as it integrates existing methods into a unified platform.
The paper tackles the challenge of training NLP models for PDF document question-answering by introducing DOCMASTER, a unified platform that enables on-device annotation, training, and inference, resulting in privacy-preserving applications like an AI assistant for UCSD's ISEO to process large volumes of PDFs.
The application of natural language processing models to PDF documents is pivotal for various business applications yet the challenge of training models for this purpose persists in businesses due to specific hurdles. These include the complexity of working with PDF formats that necessitate parsing text and layout information for curating training data and the lack of privacy-preserving annotation tools. This paper introduces DOCMASTER, a unified platform designed for annotating PDF documents, model training, and inference, tailored to document question-answering. The annotation interface enables users to input questions and highlight text spans within the PDF file as answers, saving layout information and text spans accordingly. Furthermore, DOCMASTER supports both state-of-the-art layout-aware and text models for comprehensive training purposes. Importantly, as annotations, training, and inference occur on-device, it also safeguards privacy. The platform has been instrumental in driving several research prototypes concerning document analysis such as the AI assistant utilized by University of California San Diego's (UCSD) International Services and Engagement Office (ISEO) for processing a substantial volume of PDF documents.