Aspect-based sentiment analysis (ABSA) is a critical task in fine-grained sentiment analysis, focusing on determining the sentiment polarity of specific aspect terms. While recent transformer-based models have shown effectiveness in this domain, they often lack efficiency and have not been extensively applied within conversational interactions, such as in conversational recommender systems. This paper proposes an efficient approach to ABSA for conversational recommendation, leveraging a Small BERT model. Our method utilizes knowledge distillation from a pre-trained BERT model to a smaller TinyBERT model, enhancing the performance of TinyBERT while maintaining its compact size. Experimental results on a publicly available dataset demonstrate that our approach is both effective and efficient, making it suitable for real-world applications where computational resources are limited.