Towards Building a Knowledge Graph with Open Data – A Roadmap

Abstract

With the increasing interest in knowledge graph over the years, several approaches have been proposed for building knowledge graphs. Most of the recent approaches involve using semi-structured sources such as Wikipedia or information crawled from the web using a combination of extraction methods and Natural Language Processing (NLP) techniques. In most cases, these approaches tend to make a compromise between accuracy and completeness. In our ongoing work, we examine a technique for building a knowledge graph over the increasing volume of open data published on the web. The rationale for this is two-fold. First, we intend to provide a foundation for making existing open datasets searchable through keywords similar to how information is sought on the web. The second reason is to generate logically consistent facts from usually inaccurate and inconsistent open datasets. Our approach to knowledge graph development will compute the confidence score of every relationship elicited from underpinning open data in the knowledge graph. Our method will also provide a scheme for extending coverage of a knowledge graph by predicting new relationships that are not in the knowledge graph. In our opinion, our work has major implications for truly opening up access to the hitherto untapped value in open datasets not directly accessible on the World Wide Web today.