Enhancing the Reasoning Capabilities of Natural Language Inference Models with Attention Mechanisms and External Knowledge