pdfQA: Diverse, Challenging, and Realistic Question Answering over PDFs
This provides a basis for evaluating end-to-end QA pipelines on PDFs, addressing a gap for researchers and developers working with this common document type.
The authors tackled the lack of diverse and realistic question-answering datasets for PDFs by creating pdfQA, a multi-domain dataset with 2K human-annotated and 2K synthetic QA pairs across ten complexity dimensions, and used open-source LLMs to reveal challenges correlated with these dimensions.
PDFs are the second-most used document type on the internet (after HTML). Yet, existing QA datasets commonly start from text sources or only address specific domains. In this paper, we present pdfQA, a multi-domain 2K human-annotated (real-pdfQA) and 2K synthetic dataset (syn-pdfQA) differentiating QA pairs in ten complexity dimensions (e.g., file type, source modality, source position, answer type). We apply and evaluate quality and difficulty filters on both datasets, obtaining valid and challenging QA pairs. We answer the questions with open-source LLMs, revealing existing challenges that correlate with our complexity dimensions. pdfQA presents a basis for end-to-end QA pipeline evaluation, testing diverse skill sets and local optimizations (e.g., in information retrieval or parsing).