Knowledge Portability with Semantic Expansion of Ontology Labels

Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing, pages 708–718, Beijing, China, July 26-31, 2015. c⃝2015 Association for Computational Linguistics Knowledge Portability with Semantic Expansion of Ontology Labels Mihael Arcan1 Marco Turchi2 Paul Buitelaar1 1 Insight Centre for Data Analytics, National University of Ireland, Galway firstname.lastname@insight-centre.org 2 FBK- Fondazione Bruno Kessler, Via Sommarive 18, 38123 Trento, Italy turchi@fbk.eu Abstract Our research focuses on the multilin- gual enhancement of ontologies that, of- ten represented only in English, need to be translated in different languages to en- able knowledge access across languages. Ontology translation is a rather different task then the classic document translation, because ontologies contain highly specific vocabulary and they lack contextual in- formation. For these reasons, to improve automatic ontology translations, we first focus on identifying relevant unambigu- ous and domain-specific sentences from a large set of generic parallel corpora. Then, we leverage Linked Open Data resources, such as DBPedia, to isolate ontology- specific bilingual lexical knowledge. In both cases, we take advantage of the se- mantic information of the labels to se- lect relevant bilingual data with the aim of building an ontology-specific statistical machine translation system. We evaluate our approach on the translation of a medi- cal ontology, translating from English into German. Our experiment shows a sig- nificant improvement of around 3 BLEU points compared to a generic as well as a domain-specific translation approach. 1 Introduction Currently, most of the semantically structured data, i.e. ontologies or taxonomies, has labels ex- pressed in English only.1 On the one hand, the increasing amount of ontologies offers an excel- lent opportunity to link this knowledge together (G´omez-P´erez et al., 2013). On the other hand, non-English users may encounter difficulties when 1Based on (Gracia et al., 2012), around 80% of ontology labels indexed in Watson are English. using the ontological knowledge represented only in English. Furthermore, applications in informa- tion retrieval, question answering or knowledge management, that use monolingual ontologies are therefore limited to the language in which the on- tology labels are stored. To make the ontologi- cal knowledge language-independent and accessi- ble beyond language borders, these monolingual resources need to be transformed into multilingual knowledge bases. This multilingual enhancement can enable queries on documents beyond English, e.g. for cross-lingual business intelligence in the financial domain (O’Riain et al., 2013), provid- ing information related to an ontology label, e.g. other intangible assets,2 in Spanish, German or Italian. The main challenge involved in build- ing multilingual knowledge bases is, however, to bridge the gap between language-specific informa- tion and the language-independent semantic con- tent of ontologies or taxonomies (Gracia et al., 2012). Since manual multilingual enhancement of on- tologies is a very time consuming and expensive process, we engage an ontology-specific statisti- cal machine translation (SMT) system to automat- ically translate the ontology labels. Due to the fact that ontology labels are usually highly domain- specific and stored only in knowledge represen- tations (Chandrasekaran et …

Content truncated to 500 words