Building Reliable And Robust Natural Language Processing Models: Enhancing Understanding Of Semantically Equivalent Texts