Name Disambiguation Based on Entity Relationship Graph in Big Data

Abstract

Aiming at the problem of insufficient utilization of author information and low accuracy of the existing name disambiguation methods, a name disambiguation method based on commonly used author information entity relationship graph is proposed. The entity relationship graph is constructed according to the information of the co-authors and the authors’ affiliated institutions, years of birth, gender and degrees, and the edges in the graph is divided into two categories: the vertices in the first type edges are the authors, and the vertices in the second type edges must include any one of the affiliated institution, year of birth, gender and degree. The connection strength of two authors with the same name in the graph is calculated by following steps: first, the length of the paths is limited; second, the first type edges and the second type edges are searched respectively in the graph; then, the connection strengths of different types of paths are calculated and normalized according to the number and the length of paths, and weighted summed to obtain the connection strength between two authors; finally, the obtained connection strength is compared with the threshold to realize name disambiguation. The experimental results show that the proposed method has higher accuracy than baselines.