Multimodal Visual Communication Effect Evaluation Method for University New Media Matrix

With the rapid rise of short video platforms and social media graphic content, the new media matrix of colleges and universities has become an important carrier for promoting educational brand communication and public interaction. However, the current influence of multimodal content in actual communication still lacks a systematic quantitative evaluation method, especially in the scenario where images, texts and user behavior information are highly integrated, and existing research is difficult to fully reflect the essential differences in communication effects. To solve this problem, this paper constructs a multimodal visual communication effect evaluation framework for the new media matrix of colleges and universities. This method combines image feature extraction (ResNet), text semantic modeling (BERT) and user behavior data, and uses a multi-layer perceptron to perform fusion modeling and communication effect scoring on multimodal information. Experimental results show that the proposed method is significantly better than the traditional single-modal model in prediction accuracy, and has good platform generalization ability and content type adaptability. Through a systematic comparative analysis of graphic and text fusion, visual-dominant and textdominant content, the key role of behavioral characteristics in the evaluation model is verified. The research results provide scientific quantitative support for the digital content operation strategy of colleges and universities, and expand the theoretical boundaries of multimodal communication effect research.