Review on Deep Adversarial Learning of Entity Resolution for Cross-Modal Data

Abstract

With the repaid development of the Internet, multimedia data such as image, text, video, audio is increasing, which brings opportunities and challenges to the development of the economy and science. Cross-modal data entity resolution aims to find different objective descriptions of the semantically similar items from objects in different modalities. However, different modality data have the features with underlying heterogeneity and high-level semantic related. Starting from the problem of modality gap between cross-modal data, this paper introduces how to use the idea of confrontational learning to solve the cross-modal data entity resolution problem between images and text from the aspects of feature extraction and emotional state association.