In the last seven years the IXA research group has been working on the Basque WordNet (Agirre et al., 2006). The focus have been on the representation of common vocabulary (we have incorporated around 27,000 words), but now, we have turned our attention to specialized language.

Due to the unstoppable development of specialized language and terminology, it is becoming increasingly difficult to capture and organize terminological information. Domain ontologies are helpful to handle this kind of information, and have been shown to be useful in knowledge representation, management and exchange. For instance, there are several representative ontologies in the domains of e-commerce (UNSPSC1,NAICS2), medicine (GALEN3, UMLS), engineering -EngMath (Gruber & Olsen, 1994), PhysSys (Borst, 1997) -, enterprise -Enterprise Ontology (Uschold et al., 1998) -, and knowledge management -KA (Decker et al. 1999) . Moreover, in the NLP community domain ontologies are being used to develop and evaluate different computational systems and applications (Navigli et al. 2003; Sagri et al., 2003; Stamou et al., 2002; Roventini & Marinelli, 2003).

The aim of WNTERM is to create a light-weight ontology belonging to the science and technology domains for Basque and English. WNTERM (i) stores the domain terminology, (ii) fixes relations among all the domain terms, and (iii) is connected to the Basque and English wordnets through the Multilingual Central Repository (MCR) (Atserias et al., 2004).

In order to have WNTERM linked to the wordnets currently available, we decided to structure it following the MCR framework, which we also used to build the Basque WordNet). The MCR model provides English-Basque equivalence links across the English and Basque wordnets, and it also offers the possibility of enriching both the Basque WordNet and WNTERM simultaneously.

Domain ontology terms are imported into WNTERM from the MCR itself (more precisely, from the Basque and English wordnets) and the in-house Basic Encyclopaedic Dictionary of Science and Technology (BDST) data-base4. Taking this sources into account, there are additional objectives which we want to address with WNTERM: (i) to extend the Basque WordNet with terminology, (ii) to organize hierarchically all the terms of the BDST according to the MCR architecture, in order to provide to lexicographers more specific information about a term in their navigation through the data-base, and (iii) to link all the sources in the project: the MCR, the BDST and the domain ontology (WNTERM).

The final result of this project will be a domain ontology on science and technology, which will show the taxonomic and semantic relations among all its domain concepts, and which is also linked to the MCR and the BDST. In this paper we focus on the overall methodology for the construction of WNTERM, which will be illustrated on the ecology domain, a subset of science and technology.

This paper is organized as follows. The resources used in this project are introduced in Section 2 (MCR and the Basque WordNet) and Section 3 (BDST). In Section 4, we present the methodology followed to create a domain ontology, illustrated on the ecology domain. Finally, future work is presented in Section 5.

The Multilingual Central Repository (MCR)5follows the model proposed by the EuroWordNet project6.EuroWordNet (Vossen, 1998) is a multilingual semantic lexicon with wordnets for several European languages, which are structured as the Princeton WordNet (Fellbaum, 1998).

It groups each languages' words into sets of synonyms called synsets, and records various semantic relations (such as hypernymy, hyponymy, meronymy, holonymy) between these synonym sets forming a hierarchy. Each of these synsets corresponds to a lexical concept and many have a textual gloss which often provides an explanation of what this concept represents. The MCR (Atserias et al., 2004) is a result of the 5th Framework Meaning project7(Rigau et al., 2003). The MCR integrates in the same EuroWordNet framework wordnets from five different languages, including Spanish, Italian, Catalan and Basque (together with six English WordNet versions).

The wordnets are currently linked via an Inter-Lingual-Index (ILI) allowing the connection from words in one language to translation equivalent words in any of the other languages. In that way, the MCR constitutes a natural multilingual large-scale linguistic resource for a number of semantic processes that need large amount of multilingual knowledge to be effective tools. For instance, the English synset {party, political_party} is linked through the ILI to the Basque synset {partidu, partidu_politiko, alderdi_politiko, alderdi}. The MCR also integrates the latest version of the WordNet Domains (Magnini & Cavagliä, 2000), new versions of the Base Concepts and the Top Concept Ontology (Älvez et al., 2008), and the SUMO ontology (Niles & Pease, 2001). The current version of the MCR contains 934,771 semantic relations between synsets, most of them acquired by automatic means. This represents almost four times larger than the Princeton WordNet (235,402 unique semantic relations in WordNet 3.0).

Although these resources have been derived using different WordNet versions, using the technology for the automatic alignment of wordnets (Daude et al., 2001), most of these resources have been integrated in the MCR maintaining the compatibility among all the knowledge resources which use a particular WordNet version as a sense repository.

MCR was developed based on WordNet 1.6 version. However, for this project we have moved the MCR to the latest WordNet version (3.0), because WordNet 3.0 has a higher amount of terminology (than the 1.6 version). The result of the automatic mapping was manually corrected. We will now present WordNet domains and the Basque WordNet.

In this section, we will describe the steps for the construction of WNTERM. We first compare the two resources, with special attention to the relation between the domain labels used in each resource, and an automatic analysis of the different cases found. In the following steps we select the concepts to be added to WNTERM and, the terms to be added. We finally structure the concepts in a hierarchical structure. We will illustrate the steps with data from the ecology domain.

In this paper we have presented the construction methodology of WNTERM, a science and technology light-weight ontology based on the MCR architecture which has been built with the combination of the Basque and English wordnets and the BDST. The construction methodology has been illustrated on the ecology domain. The final result is a multilingual (Basque and English) light-weight domain ontology, including taxonomic and other semantic relations among its concepts, which is tightly connected to the English WordNet 3.0 and other wordnets. The study also reveals that both wordnets and the BDST are mutually enriched in the process. In the future, we would like to reorganize the hierarchy coming from the MCR, that is, to design the terminological hierarchical organization according to the criteria of domain experts (lexicographers working on the BDST). We also want to develop a semi-automatic framework based on domain corpora (Vossen, 2001), in order to speed up the construction of other domain ontologies. WNTERM will be used as the evaluation benchmark for the automatic procedures.

We would like to thank Mari Susperregi and Pili Lizaso who have collaborated in this research. Eli Pociello has a grant from the Provincial Council of Gipuzkoa. Part of this work has been funded by the European Commission (KYOTO, ICT-2007-211423) and the Education Ministry (KNOW, TIN2006-15049-C03-01).