Large Language Models (LLMs) have shown remarkable progress in various natural language processing (NLP) tasks, yet they still struggle with role adaptation and emotional nuance in human-machine interactions, often producing formal and rigid outputs that deviate from natural human communication styles. To address these limitations and enhance LLMs’ interactive capabilities, we present Emotion-Comment, a novel dataset derived from social media platforms, and introduce Qwen-7B-Emotion, a model fine-tuned from Qwen-7B using Low-Rank Adaptation (LoRA) techniques. Our research culminates in a comparative analysis of the anthropomorphic qualities in the outputs of Qwen-7B-Emotion versus its base model, contributing to the advancement of more natural and emotionally intelligent language models. This study aims to bridge the gap between artificial and human-like communication, potentially improving user experiences across various applications of large language models.