PeerJ Computer Science (Sep 2022)

Constructing marine expert management knowledge graph based on Trellisnet-CRF

  • Jiajing Wu,
  • Zhiqiang Wei,
  • Dongning Jia,
  • Xin Dou,
  • Huo Tang,
  • Nannan Li

DOI
https://doi.org/10.7717/peerj-cs.1083
Journal volume & issue
Vol. 8
p. e1083

Abstract

Read online Read online

Creating and maintaining a domain-specific database of research institutions, academic experts and scholarly literature is essential to expanding national marine science and technology. Knowledge graphs (KGs) have now been widely used in both industry and academia to address real-world problems. Despite the abundance of generic KGs, there is a vital need to build domain-specific knowledge graphs in the marine sciences domain. In addition, there is still not an effective method for named entity recognition when constructing a knowledge graph, especially when including data from both scientific and social media sources. This article presents a novel marine science domain-based knowledge graph framework. This framework involves capturing marine domain data into KG representations. The proposed approach utilizes various entity information based on marine domain experts to enrich the semantic content of the knowledge graph. To enhance named entity recognition accuracy, we propose a novel TrellisNet-CRF model. Our experiment results demonstrate that the TrellisNet-CRF model reached a 96.99% accuracy rate for marine domain named entity recognition, which outperforms the current state-of-the-art baseline. The effectiveness of the TrellisNet-CRF module was then further demonstrated and confirmed on entity recognition and visualization tasks.

Keywords