Enhanced Contextual Feature Extraction using Convolution and Interaction Layers

By applying two convolution layers and interaction layers, we can aggregate the features of the two subtasks and obtain a more comprehensive context feature FM. This approach allows for richer contextual feature extraction by leveraging the relationships between the features of the two subtasks, leading to improved performance in various applications.