BibleSTS: Bible Verse Dataset Pairs for Semantic Textual Similarity

This dataset is a collection of 6 million Bible verse pairs for training and evaluating language models to perform semantic textual similarity. Verse pairs extend across eighteen different translations of the Bible: Bible in Basic English Version, Brenton Septaugint Translation, Darby Translation, Douay-Rheims 1899 Translation, English Majority Text Version, Free Bible Version, Geneva Bible, George Noyes Bible, King James Version, Literal Standard Version, New English Translation (NET) Bible, Noah Webster Bible, Nyangumarta English Bible, Open English Bible, Unlocked Literal Bible, World English Bible, Wycliffe’s Bible (with common spelling), and Young’s Literal Translation. Each pair is categorized as either an exact match, a biblical cross-reference, or neither.