Aspect-oriented sentiment analysis is a fine-grained sentiment analysis aimed at identifying the sentiment polarity of specific aspects within a sentence. The existing suffer from sentence and aspect semantic loss, inadequate semantic extraction, and ignore contextual information and positional relationships of aspect terms in sentences, leading to a decrease in the accuracy of sentiment polarity recognition. Therefore, a multi-attention aspect-level sentiment analysis model based on RoBERTa and positional features was proposed for aspect-level sentiment analysis. To verify the effectiveness of the model, a large number of experiments were conducted on SemEval2014 Restaurant 14, SemEval2014 Laptop 14, SemEval2015 Restaurant 15, SemEval2016 Restaurant 16 datasets. The experimental results demonstrate that the proposed model achieves higher accuracy and F₁ compared to the benchmark models.