Generating Partner-Adaptive Explanations: Leveraging Language Models for Dynamic Utterance Generation from Triples