Towards Data-driven Ontologies: a Filtering Approach using Keywords and Natural Language Constructs

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 2285–2292 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 2285 Towards Data-driven Ontologies: a Filtering Approach using Keywords and Natural Language Constructs Maaike H. T. de Boer and Jack P. C. Verhoosel TNO Anna van Buerenplein 1, 2595 DA, The Hague, The Netherlands maaike.deboer@tno.nl, jack.verhoosel@tno.nl Abstract Creating ontologies is an expensive task. Our vision is that we can automatically generate ontologies based on a set of relevant documents to create a kick-start in ontology creating sessions. In this paper, we focus on enhancing two often used methods, OpenIE and co- occurrences. We evaluate the methods on two document sets, one about pizza and one about the agriculture domain. The methods are evaluated using two types of F1-score (objective, quantitative) and through a human assessment (subjective, qualitative). The results show that 1) Cooc performs both objectively and subjectively better than OpenIE; 2) the filtering methods based on keywords and on Word2vec perform similarly; 3) the filtering methods both perform better compared to OpenIE and similar to Cooc; 4) Cooc-NVP performs best, especially considering the subjective evaluation. Although, the investigated methods provide a good start for extracting an ontology out of a set of domain documents, various improvements are still possible, especially in the natural language based methods. Keywords: knowledge representation, ontologies, text mining 1. Introduction In communities it is important to have a common language. Often information standards are created to capture the com- mon language. The creation of such standards can be sup- ported by an ontology that defines the common concepts in the domain. In order to create such an ontology, a common approach is for experts to come together in working ses- sions and discuss about the main domain concepts and their relations. These sessions need the dedicated effort from people to construct the model which takes a lot of time and effort. In this paper, we aim to generate a quick start with an ontology as a starting point for further improvement by experts in the domain. The field in which ontologies are learned from available knowledge using data is named on- tology learning. This paper is an extension of our previous papers on this topic (de Boer and Verhoosel, 2019a) (de Boer and Verhoosel, 2019b). In the first paper (de Boer and Verhoosel, 2019b), eight different methods were evaluated and in the second paper (de Boer and Verhoosel, 2019a) this evaluation was enhanced with two additional document sets and new evaluation metrics. In this paper, we focus on two of the methods and further enhance them. We also added a qualitative evaluation by human expert assessors. In the next section, the related work is described focusing on ontology learning, including open information extrac- tion, and the evaluation of ontologies. In section 3, the ex- perimental set-up with the document sets, characteristics of the resulting ontologies and our evaluation methodology is explained. Section 4 contains the results of the …

Content truncated to 500 words