Expert information complementation updating method for twin map neural network

Abstract

The invention discloses an expert information complementation updating method for a twin map neural network, and the method comprises the steps: extracting expert data information for multi-source heterogeneous expert information data, and constructing an expert information map; obtaining sub-graphs in the expert information graph to construct an expert attribute graph; an expert attribute graph is learned by adopting a twin graph neural network, meanwhile, a multi-dimensional attention mechanism is introduced into the graph neural network for training, a loss function is calculated by using a comparative learning method, and the similarity between two expert entities is evaluated through calculation of the loss function. Therefore, whether the two expert entities are the same entity or not is judged. And after the expert information graph is judged to be the same entity, fusing and complementing the expert information graph by adopting a cyclic adaptation domain method based on an EMD distance according to the time information to obtain complete expert entity information. According to the method, rich semantic information and structure information in expert data are combined, and the similarity between two expert entities is judged through the twin map neural network. And the same expert entity reference time information is fused, so that the purpose of complementing the expert information is achieved, and the expert information is updated regularly.

Description

technical field [0001] The invention belongs to the technical field of information completion and update, in particular to an expert information completion and update method of a twin graph neural network. Background technique [0002] At present, the relevant data resources of scientific research experts are huge in number, scattered and not processed in depth, which makes it difficult for users to quickly and accurately obtain academic information that is of practical value to them. For example, if the project review department wants to find suitable reviewers, scientific research users seek collaborators, and professional problems need to be consulted by experts in the field, etc., they need to manually search through massive information resources, which will take a lot of time and effort, and the query results will be inconsistent. must be valuable. It is particularly important to extract and integrate complete expert information from massive resources and update it. A...

AI Technical Summary

Problems solved by technology

However, the traditional algorithm based on entity attribute similarity matching based on entity attribute similarity matching is not suitable for entity matching with many fields and there may be a large number of empty fields. If there is almost no intersection between empty fields, it is likely to be judged as two different entities when performing similarity matching. At the same time, the traditional expert information completion method does not take into account the impact of time information on expert information completion.

Images