Image-Attention based User Entity Recognition of Device

In this study, we investigate a method for identifying the attribution of domain user based on web page information. In the process of using domains, users or network administrators usually use the method of querying the domain registry or open portal website when detecting the real organization of the domain. However, because many organizations rely on the services provided by cloud service providers and domain service providers, The user obtained by the traditional query method is often the registration user of the domain, not the user of the domain. Named entity recognition, as an important branch of natural language processing, usually only processes and recognizes information through text. However, the information provided by the webpage contains both text and images, and many images also contain the name of the user to which the domain belongs. In the research of this paper, using the text information in the website and the user information in the image, a multi-modal fusion method is proposed to identify the user to which the domain belongs.