Design of Scholar Portrait System Based on Network Representation Learning

In order to obtain more representative scholar vectors, this paper introduces a network representation learning algorithm to construct a scholar academic text information and co authorship network, with scholars as nodes. By using node text features and maximizing node adjacency structure and global mutual information of the network, higher-order node representations are obtained as scholar representation vectors. The system defines various dimensions of the portrait through a tag management module, performs data statistics and calculations, and sets up tag version recording and historical backtracking functions to better adapt tags to academic knowledge bases at different stages. The scholar portrait customization module supports querying different portraits of individual scholars, facilitating timely adjustment of portrait data and portrait construction methods.