A Multilingual Linguistic Domain Ontology

A Multilingual Linguistic Domain Ontology Mariem Neji MIRACL Laboratory Sfax, Tunisia mariem.neji @yahoo.fr Fatma Ghorbel CEDRIC Laboratory Paris, France fatmaghorbel @gmail.com Bilel Gragouri MIRACL Laboratory Sfax, Tunisia bilel.gargouri @fsegs.rnu.tn Nada Mimouni CEDRIC Laboratory Paris, France nada.mimouni @lecnam.net Elisabeth M´etais CEDRIC Laboratory Paris, France elisabeth .metais@cnam.fr Abstract Natural Language Processing provides a very significant contribution to various research ar- eas such as the e-health, e-business, educa- tion and antiterrorism. However, understand- ing the meaning, scope and usage of linguis- tic knowledge is a tedious task for a hetero- geneity of potential users. Several approaches have been proposed to represent heteroge- neous linguistic knowledge covering specific features of some languages. However, these approaches focused only on the data aspect of linguistic knowledge and neglected the pro- cessing one. Moreover, most of them do not support multilingualism and lack of powerful semantic representation and reasoning abili- ties. In this paper, we propose a multilingual linguistic domain ontology, called LingOnto, that represents and reasons about (1) linguis- tic data, (2) linguistic processing functionali- ties and (3) linguistic processing features. Our ontology supports English, French and Arabic languages and can be used by linguistically under-skilled users. In order to evaluate Lin- gOnto and measure its efficiency, we applied it to a framework of identifying valid composi- tion workflows of linguistic web services. Fi- nally, we give the results of the carried-out ex- periments. 1 Introduction Natural Language Processing (NLP) provides a very significant contribution to various research areas such as the e-health, e-business, education and an- titerrorism. It has witnessed over the last years an acceleration in progress on a wide range of different applications such as sentiment analysis, knowledge mining and reasoning and search engine (Zhou et al., 2020). However, understanding the meaning, scope and usage of linguistic knowledge is a tedious task. This complexity is mainly due to three reasons. First, NLP domain’s potential users are heterogeneous and a considerable number of them are linguisti- cally under-skilled. Second, the language is always changing, evolving, and adapting to its user’s needs. For instance, words can acquire new meanings over time (e.g., the meaning of ”apple” is a fruit but the meaning of ”Apple” is a company). Finally, every language has its own specificities. Indeed, each lan- guage has its own structures and ways of interpret- ing. For example, Arabic has verbal and nominal sentences; but English has only verbal sentences. In NLP, two types of approaches have been pro- posed to represent linguistic knowledge : (1) online registries-based approach such as the ISOcat reg- istry1, the SIL Glossary of linguistic terms2 and the CLARIN Concept Registry (Schuurman et al., 2016) and (2) ontologies-based approach such as the Gen- eral Ontology for Linguistic Description (GOLD) (Farrar and Langendoen, 2010), OntoTag ontologies (De Cea et al., 2004) and WordNet (Gangemi et al., 2002). However, these approaches present only lin- guistic data (e.g., word, noun, verb and adjective) and do not focus on modeling linguistic process- ing functionalities (e.g., tokenization, stemming and part of speech tagging) and …

Content truncated to 500 words