Exploiting survey and online patient experience comments in general practice: validating a fine-tuned language model for automatic sentiment analysis

The authors aimed to evaluate the effectiveness of a fine-tuned masked language model (NorBERT3large) for automated sentiment analysis of patient experience comments in general practice, comparing its performance to traditional methods and a large language model. They found that the fine-tuned model achieved high classification accuracy and demonstrated construct validity, suggesting it could serve as a reliable measure of patient experience and assist in identifying areas for quality improvement.