This dataset can be used for fine-tuning embedding models using positive text pairs (question, context).