A Holistic Natural Language Generation Framework for the Semantic Web

Proceedings of Recent Advances in Natural Language Processing, pages 819–828, Varna, Bulgaria, Sep 2–4, 2019. https://doi.org/10.26615/978-954-452-056-4_095 819 A Holistic Natural Language Generation Framework for the Semantic Web Axel-Cyrille Ngonga Ngomo1 Diego Moussallem1 Lorenz B¨uhmann2 1Data Science Group, University of Paderborn, Germany 2AKSW Research Group, University of Leipzig, Germany {first.lastname}@upb.de lastname@informatik.uni-leipzig.de Abstract With the ever-growing generation of data for the Semantic Web comes an increas- ing demand for this data to be made avail- able to non-semantic Web experts. One way of achieving this goal is to translate the languages of the Semantic Web into natural language. We present LD2NL, a framework for verbalizing the three key languages of the Semantic Web, i.e., RDF, OWL, and SPARQL. Our framework is based on a bottom-up approach to verbal- ization. We evaluated LD2NL in an open survey with 86 persons. Our results suggest that our framework can generate verbaliza- tions that are close to natural languages and that can be easily understood by non- experts. Therewith, it enables non-domain experts to interpret Semantic Web data with more than 91% of the accuracy of domain experts. 1 Introduction Natural Language Generation (NLG) is the process of automatically generating coherent Natural Lan- guage (NL) text from non-linguistic data (Reiter and Dale, 2000a). Recently, the field has seen an increased interest in the development of NLG systems focusing on verbalizing resources from Semantic Web (SW) data (Gardent et al., 2017). The SW aims to make information available on the Web easier to process for machines and humans. However, the languages underlying this vision, i.e., Resource Description Framework (RDF), SPARQL Query Language (SPARQL) and Web Ontology Language (OWL), are rather difficult to understand for non-expert users. For example, while the meaning of the OWL class expres- sion Class: Professor SubClassOf: worksAt SOME University is obvious to every SW expert, this expression (“Every professor works at a university”) is rather difficult to fathom for lay persons. Previous works such as SPARQL2NL (Ngonga Ngomo et al., 2013) and SPARTIQULATION (Ell et al., 2012) have already shown the usefulness of the verbalization of SPARQL 1 and RDF in ar- eas such as question answering (Lehmann et al., 2012) and the explanation of the output of systems based onSW technologies (Ngonga Ngomo et al., 2013). However, other SW languages are rarely investigated, such as OWL. In this paper, we present an open-source holis- tic NLG framework for the SW, named LD2NL, which facilitates the verbalization of the three key languages of the SW, i.e., RDF, OWL, and SPARQL into NL. Our framework is based on a bottom-up paradigm for verbalizing SW data. Ad- ditionally, LD2NL builds upon SPARQL2NL as it is open-source and the paradigm it follows can be reused and ported to RDF and OWL. Thus, LD2NL is capable of generating either a single sentence or a summary of a given resource, rule, or query. To val- idate our framework, we evaluated LD2NL using experts 66 in Natural Language Processing (NLP) and SW as well as 20 non-experts who were …

Content truncated to 500 words