Innovation of Digital Design of Intangible Cultural Heritage Based on Knowledge Graph and Multimodal Integration

This article conducted data application and task performance optimization analysis of multiple distribution categories on a multimodal dataset.The impact construction of the model in different periods was achieved by implementing a long short-term classification plan for the training data step size.On the basis of ensuring the efficient operation of the classification model, this article compares and analyzes the performance of other machine models.This article compares the accuracy of the LSTM model and support vector machine with the random forest model and finds that the performance of the LSTM model is more accurate than the other two models.The multimodal model features have shown an improvement in accuracy analysis after emphasizing the fusion of different states.With the assistance of computer verification of the recall rate, a critical investigation and verification were conducted on the fusion state data distribution of multiple models.Improving the classification scheme of the model greatly enhances the accuracy of related technologies in practical applications.