Document Classification Pattern Recognition via Information Fusion: A Systematic Review of Multimodal and Multiview Representation Approaches
A systematic review of 139 studies on information fusion for document classification reveals that integrating multiple data sources or representations significantly boosts accuracy. Multimodal fusion, in particular, shows a mean accuracy gain of 5.28 percentage points. However, the review highlights a critical lack of reproducibility, with a small percentage of studies employing statistical tests to validate their findings, undermining the reliability of many results. AI
IMPACT Provides a quantitative synthesis of information fusion techniques for document classification, highlighting accuracy gains and reproducibility challenges.