Expert recommendation method based on multi-modal information learning

Abstract

The invention discloses an expert recommendation method based on multi-modal information learning, and the method comprises the steps: crawling network expert information data and project information data, and arranging the expert information data in an existing expert database; constructing an expert review heterogeneous graph and an expert attribute heterogeneous graph; the text information of the expert is learned by the bert model, the attribute heterogeneous graph of the expert is learned by the graph neural network, the historical review sequence of the expert is learned by building a self-attention recommendation model, the learned embedding is input into the fusion layer to obtain expert information embedding, and a pre-training model is obtained; and extracting expert information embedding by using a pre-training model, encoding project information to obtain project information embedding, and inputting the obtained expert attribute representation and project information representation into a multi-layer perceptron to train the model to obtain recommendation scores of experts and projects. According to the method, rich semantics and attribute information are fully fused into embedding and model parameters of experts, and the accuracy of expert recommendation is improved.

Description

technical field [0001] The invention belongs to the technical field of data processing, in particular to an expert recommendation method based on multimodal information learning. Background technique [0002] With the development of science and technology and other aspects to vigorously promote theoretical innovation, the number of applications for various innovative projects has increased significantly, which in turn leads to an increasing number of applications for science and technology projects. Among them, there is a key step in the stage of scientific research project initiation and conclusion—recommendation by review experts. Review expert recommendation refers to recommending experts in related fields to review the project according to the scientific research project documents, so as to evaluate the practical significance, feasibility and completion quality of the project. This requires that the skills and fields of expertise mastered by the review experts match the...

AI Technical Summary

Problems solved by technology

At the same time, when a new expert appears, there is often a problem of incomplete expert information, the model cannot learn the effective representation of the expert, and it is difficult to recommend the expert when a project that needs to be reviewed appears, resulting in the problem of low recommendation accuracy

Images