End-user engineering of ontology-based knowledge bases

ABSTRACT

Knowledge bases store information on certain topics. Applying a well-structured and machine-readable format for a knowledge base is a prerequisite for any AI-based processing or reasoning. Semantic technologies (e.g. RDF) offer such a format and have the advantages that they make it possible to define the semantics of the information and support advanced querying. However, the disadvantage is that using such technologies is challenging for people not trained in IT, such as subject matter experts. This means that they need to rely on semantic technology experts to create, maintain, and query their knowledge bases. However, these experts are, in turn, not trained in the subject matter, while domain knowledge is essential for the construction of high-quality knowledge bases. In this paper, we present an end-user engineering approach for ontology-based knowledge bases. The goal is to allow subject matter experts to develop, maintain, and exploit the knowledge base themselves. We also present the supporting tools developed so far. The tools for the construction and the manual filling of the knowledge base are using the jigsaw metaphor to hide technicalities and guide the users. We also developed tools to automatically import data from spreadsheets into the knowledge base and to perform some type of quality control on the data. The end-user approach and the tools are demonstrated and evaluated for building a knowledge base in the toxicology domain.

KEYWORDS:

• Knowledge base engineering
• end-user development
• domain ontology development
• data lifting
• quality assurance
• jigsaw metaphor

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 https://www.openaire.eu/opscidia-ontology-generator

2 The process of designing, building, and using a system

3 Example Safety Evaluation Opinion: https://ec.europa.eu/health/scientific_committees/consumer_safety/docs/sccs_o_199.pdf

4 We refer to https://www.w3.org/wiki/ConverterToRdf for a non-exhaustive list.

5 While an oversimplification, it suffices to state, for this article, that in an OWA setting, information that is not known is not necessarily false.

6 Most often limited to inferring triples with a reasoning engine before applying SHACL.

7 https://github.com/TopQuadrant/shacl

8 https://github.com/RDFLib/pySHACL

9 In a RDF graph, entities refer to the nodes in the graph.

10 test is a domain concept

11 https://jena.apache.org/documentation/fuseki2/

12 http://vowl.visualdataweb.org/webvowl.html

Additional information

Funding

The research of Audrey Sanctorum has been funded by an FWO Postdoc Fellowship (1276721N) of the Research Foundation Flanders. Financially supported by Vrije Universiteit Brussel under Grant IRP19; and Cosmetics Europe and the European Chemical Industry Council (CEFIC).