Using a Lexical Semantic Network for the Ontology Building

Proceedings of Recent Advances in Natural Language Processing, pages 92–101, Varna, Bulgaria, Sep 2–4, 2019. https://doi.org/10.26615/978-954-452-056-4_012 92 Using a Lexical Semantic Network for the Ontology Building Nadia Bebeshina-Clairet LIRMM / LIMICS Sylvie Despres LIMICS 74 rue Marcel Cachin 93017 Bobigny Mathieu Lafourcade LIRMM 860 rue de St Priest 34095 Montpellier Abstract Building multilingual ontologies is a hard task as on- tologies are often data-rich resources. We introduce an approach which allows exploiting structured lex- ical semantic knowledge for the ontology building. Given a multilingual lexical semantic (non ontolog- ical) resource and an ontology model, it allows min- ing relevant semantic knowledge and make the on- tology building and enhancement process faster. 1 Introduction Nowadays, termino-ontological resources are in- creasingly rich in terms of data they rely upon. The scientific community works intensively on data ac- quisition for the ontology building. In particular, the NeOn project1 has been set up to provide a method- ology for the ontology engineering by integrating preexisting knowledge resources into an ontology building process. The NeOn methodology contains a consistent framework for modular ontology build- ing as well as for setting up ontology networks. Here we focus on exploiting Lexical Semantic Networks (LSNs) to enrich an ontology or accompany the ontology building process. We assume that LSNs represent knowledge as it is expressed through the human language whereas ontologies provide a for- mal description (specification) of a conceptualiza- tion (concepts and relationships between those con- cepts) shared by a community of agents. A con- 1http://neon-project.org/nw/About_NeOn.html cept corresponds to a set of individuals sharing sim- ilar characteristics and may or may not be lexical- ized. Thus, the ontology labels cannot be polyse- mous. The strength of ontologies is in their for- mal consistency. The weaknesses are linked to their coverage, size (as stated in (Raad and Cruz, 2015), ”large ontologies usually cause serious scalability problems”), and human effort needed for their build- ing. The potential of the LSNs is linked to the large amount of explicit semantic information they con- tain. However, a filtering process is needed to dis- criminate irrelevant information (polysemy, noise). 2 State of the Art The opportunity of ontology construction empow- ered by the use of Natural Language Processing (NLP) techniques and tools has been explored for more than 20 years. Among the achievements, one can distinguish the tools which take into account the difference between the lexical term and the ontol- ogy concept (differentiated tools) and those that do not make such distinction. Differentiated tools and methods suggest extracting the terminological units from texts and organizing them as a network using a set of hierarchical and equivalence relation types. Such network guides the ontology expert through the conceptualisation and ontology building pro- cess. Such process relies on an intermediary struc- ture, a termino-ontology (for instance, the Termi- nae suite, (Szulman, 2012)). Undifferentiated tools use some statistical information to suggest the can- 93 didate concepts. They exploit such methods as for- mal concept analysis (Mondary, 2011) or knowledge based methods (for instance, TextToOnto2). …

Content truncated to 500 words