Releasing the CRaQAn (Coreference Resolution in Question-Answering): An open-source dataset and dataset creation methodology using instruction-following models

Rob Grzywinski, Joshua D'Arcy, Robert Naidoff, Ashish Shukla, Alex Browne, Ren Gibbons, Brinnae Bent

Keywords: instruction following models, large language models (LLMs), natural language processing (NLP), automated dataset creation, coreference analysis, question answering, information retrieval

TL;DR: Using an instruction-following model (GPT-4) to generate an open-source dataset that focuses on coreference resolution in question answering within the context of information retrieval systems.

Abstract: Instruction-following language models demand robust methodologies for information retrieval to augment instructions for question-answering applications. A primary challenge is the resolution of coreferences in the context of chunking strategies for long documents. The critical barrier to experimentation of handling coreferences is a lack of open source datasets, specifically in question-answering tasks that require coreference resolution. In this work we present our Coreference Resolution in Question-Answering (CRaQAn) dataset, an open-source dataset that caters to the nuanced information retrieval requirements of coreference resolution in question-answering tasks by providing over 250 question-answer pairs containing coreferences. To develop this dataset, we developed a novel approach for creating high-quality datasets using an instruction-following model (GPT-4) and a Recursive Criticism and Improvement Loop.

Submission Number: 36