Researchers have developed a framework called Cognitive Pairwise Comparison Classification Model Selection (CPC-CMS) for document-level sentiment analysis. This framework uses expert knowledge to weigh evaluation criteria such as accuracy and precision when selecting the best classification model. The authors tested CPC-CMS on three social media datasets, finding that ALBERT performed well in terms of accuracy, but no single model consistently outperformed others when considering time factors. The study also compared CPC-CMS with other aggregation methods, including AHP and TOPSIS, and found it to be robust. --- Why it matters: This work matters for AI researchers because it provides a framework for selecting the best classification model for document-level sentiment analysis tasks, which is a common problem in natural language processing. The authors' use of expert knowledge to weigh evaluation criteria can also inform other applications where multiple models are being compared. Source: https://arxiv.org/abs/2507.14022