Using the Swadesh list for creating a simple common taxonomy

Using the Swadesh list for creating a simple common taxonomy Laurent Prévot, Huang Chu-Ren, Su I-Li Institute of Linguistics, Academia Sinica Nankang, Taipei, Taiwan 115 {prevot,churen,isu}@gate.sinica.edu.tw Abstract. The long-term goal of the research described in this paper is to develop a multilingual language resource linked to the Princeton WordNet. The paper describes the experiments we are conducing for de- termining a basic vocabulary and for designing a language-independent core for the future resource. More precisely, in this paper we use the universality of the Swadesh list [15] for selecting it as a basic core vo- cabulary and we present several options for designing a minimal upper ontology underlying the list. Keywords: Language Resource, Lexicon, Multilinguality, Ontology, In- terlingual index, Swadesh list 1 Introduction One of the main goal of ’Developing International Standards of Language Re- sources for Semantic Web Applications’ [16], an international project sponsored by Japan’s NEDO foundation, is to implement standards of language resources that can be very robust when applied to different languages of the world. In ad- dition, the project concentrates on Asian language resources. Hence the project plans to construct lexica of Japanese, Mandarin, Thai and Italian; and inte- grate them as parts of a multilingual resource linked to the original Princeton WordNet (WN) [2]. It is therefore comparable with projects such as EuroWord- Net [17], in which the coherence and homogeneity across different languages remained the main issue that hampered the true inter-operability of the final resource. Since the project is exposed to these risks by its design, a priority for the project members is therefore to tackle this issue from the very beginning. For this purpose, one of the measures taken is to build jointly an upper-level ontology1 that will play the role of a structured interlingual index. The NEDO participants are currently exploring several ways for selecting a basic vocabulary that will serve as a starting point for designing this language-independent core of the resource. This paper describes some of the preliminary experiments we 1 In this paper, ontology has to be understood in a light-weight sense. At this de- velopment stage the ‘ontology’ is much more a simple taxonomy than a complete axiomatic theory. 324 are currently conducing. More precisely we are (i) using the Swadesh list [15] as a basic core vocabulary and (ii) exploring the possibilities offered by this list for creating a simple common ontology. These practical objectives confront us however with a complex discussion of contemporary linguistics, namely the rela- tivist/universalist debate [4], and more precisely its consequences for the lexical organization. In the context of this project the existence and the nature of a common "universal" structure is a background question that we would like to contribute to answer. 2 Approaches for designing a core lexicon Traditional approaches considered for establishing a compact list of basic terms (or core lexicon) can be divided into two categories according to their criteria for selecting the terms: semantic primacy and frequency. 2.1 Semantic primacy criterion This approach proposes to …

Content truncated to 500 words