A name disambiguation module for intelligent robotic consultant in industrial internet of things. (February 2020)

• Record Type: Journal Article
• Authors: Ma, Xiao; Wang, Ranran; Zhang, Yin; Jiang, Chi; Abbas, Haider
• Is Part Of: Mechanical systems and signal processing. Volume 136(2020)
• Publication Date: 2020-02
• Subjects: Intelligent robotics -- Name disambiguation -- Industrial internet of things -- Heterogeneous information networks -- Representation learning; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2019.106413
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: A heterogeneous network representation learning method. Refining the initialization of the representation vectors of papers. Two meta-path-based proximity measures. Abstract: Nowadays, the industrial internet of things (IIoT) is becoming more and more popular and necessary in our daily life. Intelligent robotic consultant which is a very important application of IIoT has attracted the attention of researchers. Robotic literature consultant which aims at helping users find related news or articles are becoming more and more important. As far as we know that the name disambiguation problem of robotic literature consultant still remain unsolved. networks (e.g., Utilizing simple topological structure in academic network (e.g., the relationships of co-authors) and feature engineering are the main method. However, it is generally difficult to generate the features considering the information's privacy and availability. In addition, the semantics underlying the real-world academic data cannot be easily captured by the simple relationship data. Therefore, in this paper, a graph embedding based name disambiguation module named Mech-RL is introduced for the robotic literature consultant and a novel meta-path channel based heterogeneous network representation learning method named Mech-RL is proposed. Firstly, given some meaningful meta-paths, the heterogeneous information network and meta-path channels are extracted. Then, by proposing two meta-path based proximity measures, Mech-RL obtains the node embeddings by sampling and jointly updating according to each meta-path channel. Finally, we apply the clustering algorithm on the generated paper embeddings to complete the disambiguation task. The real-world dataset based experimental results demonstrate our approach is effective compared to the related author name disambiguation approaches.