RAGulator: Effective RAG for Regulatory Question Answering
Abstract
Regulatory Natural Language Processing (Reg-NLP) is a multidisciplinary domain focusedon facilitating access to and comprehension ofregulatory documents and requirements. Thispaper outlines our strategy for creating a systemto address the Regulatory Information Retrievaland Answer Generation (RIRAG) challenge,which was conducted during the Reg-NLP 2025 Workshop. The objective of thiscompetition is to design a system capable ofefficiently extracting pertinent passages fromregulatory texts (ObliQA) and subsequentlygenerating accurate, cohesive responses to inquiriesrelated to compliance and obligations.Our proposed method employs a lightweightBM25 pre-filtering in retrieving relevant passages.This technique efficiently shortlistingcandidates for subsequent processing withTransformer-based embeddings, thereby optimizingthe use of resources.
Similar publications
partnership