From research articles to knowledge graphs: Methods for ontology-driven knowledge base creation from text

Abstract

Understanding and extracting knowledge contained in text and encoding it as linked data for the WEB is a highly complex task that poses several challenges, requiring expertise from different fields such as conceptual modeling, natural language processing and web technologies including web mining, linked data generation and publishing, etc. When it comes to the scholarly domain, the transformation of human readable research articles into machine comprehensible knowledge bases is considered of high importance and necessity today due to the explosion of scientific publications in every major discipline, that makes it increasingly difficult for experts to maintain an overview of their domain or relate ideas from different domains. This situation could be significantly alleviated by knowledge bases capable of supporting queries such as: find all papers that address a given problem; how was the problem solved; which methods are employed by whom in addressing particular tasks; etc. that currently cannot be addressed by commonly used search engines such as Google Scholar or Semantic Scholar. This tutorial addresses the above challenge by introducing the participants to methods required in order to model knowledge regarding a given domain, extract information from available texts using advanced machine learning techniques, associate it with other information mined from the web in order to infer new knowledge and republish everything as linked open data on the Web. To this end, we will use a specific use case - that of the scholarly domain, and will show how to model research processes, extract them from research articles, associate them with contextual information from article metadata and other linked repositories and create knowledge bases available as linked data. Our aim is to show how methodologies from different computer science fields, namely natural language processing, machine learning and conceptual modeling, can be combined with Web technologies in a single meaningful workflow.