A Survey on State-of-the-art Techniques for Knowledge Graphs Construction and Challenges ahead

TL;DR

This survey reviews current automated methods for constructing knowledge graphs, emphasizing their importance in enabling intelligent applications and discussing challenges to achieve near-human quality in knowledge graph generation.

Contribution

It provides a comprehensive critique of state-of-the-art automated techniques for knowledge graph construction and highlights key research challenges ahead.

Findings

Automated schemes can reduce construction costs significantly.

Current techniques aim for near-human quality in knowledge graphs.

Several research issues remain to be addressed for high-quality knowledge graphs.

Abstract

Global datasphere is increasing fast, and it is expected to reach 175 Zettabytes by 20251 . However, most of the content is unstructured and is not understandable by machines. Structuring this data into a knowledge graph enables multitudes of intelligent applications such as deep question answering, recommendation systems, semantic search, etc. The knowledge graph is an emerging technology that allows logical reasoning and uncovers new insights using content along with the context. Thereby, it provides necessary syntax and reasoning semantics that enable machines to solve complex healthcare, security, financial institutions, economics, and business problems. As an outcome, enterprises are putting their effort into constructing and maintaining knowledge graphs to support various downstream applications. Manual approaches are too expensive. Automated schemes can reduce the cost of building knowledge graphs up to 15-250 times. This paper critiques state-of-the-art automated techniques to produce knowledge graphs of near-human quality autonomously. Additionally, it highlights different research issues that need to be addressed to deliver high-quality knowledge graphs