Explainable identification of similarities between entities for discovery in large text