This dataset was derived from the exposure, outcomes, significance, direction, type and value of effect size in the titles and abstracts of cohort study publications. It was extracted by the large language model and was used to construct a cohort knowledge graph.