QSU: Quran Semantic Units Dataset

Abstract

The Quran Semantic Units (QSU) Dataset, a novel segmentation of the Quranic text that achieves fine-grained granularity by applying the classical rules of Waqf wa Ibtida' (stopping and resuming), which is considered an essential tool for understanding and correctly interpreting the Quranic text. Unlike standard verse-by-verse divisions, which often contain multiple ideas within a single ayah, our dataset partitions the text into self-contained, semantically complete units. Built to advance NLP applications such as semantic search, question-answering, contextual embeddings, and automated tafsir, our methodology synthesizes established stopping conventions with rigorous linguistic analysis to accurately delineate semantic boundaries.