<?xml version="1.0" standalone="yes"?>
<Paper uid="A00-2006">
  <Title>Encoding information on adjectives in a lexical-semantic net for computational applications</Title>
  <Section position="1" start_page="0" end_page="0" type="abstr">
    <SectionTitle>
Abstract
</SectionTitle>
    <Paragraph position="0"> The goal of this paper is to describe how the EuroWordNet framework for representing lexical meaning is being modified within an Italian National Project in order to include information on adjectives. The focus is on the 'new' semantic relations being encoded and on the revisions we have made to the EuroWordNet Top Ontology structure. We also briefly discuss the utility of the information which is being encoded for computational applications.</Paragraph>
    <Paragraph position="1"> Introduction The Princeton WordNet (henceforth WN) is a lexical semantic network in which the meanings of words are represented in terms of their conceptual and lexical relations to other words. The basic notion around which it is developed is that of a synset (synonyms set), i.e. a set of words with the same Part-of-Speech (PoS) that can be interchanged in a certain context. Various conceptual and lexical relations are then encoded between synsets of the same PoS: e.g., hyponymy, antonymy, meronymy, etc. (Miller et al. 1990; Fellbaum 1998b).</Paragraph>
    <Paragraph position="2"> Within the EuroWordNet (henceforth EWN) project I a similar (multilingual) lexical resource was developed, retaining the basic underlying design of WN, but enriching the set of lexical-semantic relations to be encoded for nouns and verbs in various ways 2, in order to obtain a maximally re-usable resource for computational applications. Thus, a) cross-PoS (xPos) relations were added so that different surface realizations of similar concepts within and across languages could be matched (e.g., the noun research and the verb to research could be linked as 1 EWN was a project in the EC Language Engineering (LE-4003 and LE-8328) programme. In a first phase, the partners involved were the University of Amsterdam (coordinator); the Istituto di Linguistica Computazionale, CNR, Pisa; the Fundacion Universidad Empresa (a cooperation of UNED, Madrid, Politecnica de Catalunya, Barcelona, and the University of Barcelona); the University of Sheffield; and Novell Linguistic Development (Antwerp), changed to Lemout &amp; Hauspie during the project. In a further phase, the database was extended with German, French, Estonian and Czech. Complete information on EWN can be found at its web site: http://www.hum.uva.nl/~ewn/.</Paragraph>
    <Paragraph position="3"> 2 Adjectives and adverbs were encoded in EWN only as targets of relations from nouns and verbs.</Paragraph>
    <Paragraph position="4">  'XPOS NEAR__SYNONYMS' in EWN); b) some relations were identified which provide detailed information on semantic components lexicalized within word roots (e.g., to hammer could be linked to the noun hammer by means of an 'INVOLVED_INSTRUMENT' relation); c) some labels were distinguished which could be added to relations to make their semantic entailments more explicit and precise (cf. Alonge et al.</Paragraph>
    <Paragraph position="5"> 1998). The links among the wordnets of different languages were realized by means of an Interlingual-Index (ILI), constituted by an unstructured list of the Princeton WN (version 1.5) synsets. In addition, a hierarchy of language-independent concepts, reflecting fundamental semantic distinctions (e.g., Object and Substance, Dynamic and Static, Cause, Manner, etc.), was built: the Top Ontology (TO). The TO consists of language-independent features which may (or may not) be lexicalized in various ways, or according to different patterns, in different languages (Rodriguez et al. 1998). Via the ILI, all the concepts in the monolingual wordnets are directly or indirectly linked to the TO. The following picture shows the EWN data structure (see also Vossen 1999): While in WN all PoSs are represented, in EWN detailed information was encoded only for nouns and verbs and no analysis was carried out with respect to lexical-semantic relations which could be used to describe the semantics of adjectives and adverbs. In an Italian National Project we are building a large wordnet, ltalWordNet (henceforth IWN) 3, by extending the network built for Italian in EWN. Thus, we are both increasing the coverage for nouns and verbs and adding adjectives and some adverbs 4. To be able to encode information on adjectives we have enriched the set of the EWN lexical-semantic relations, aiming at encoding data which can be useful for computational applications. Moreover, we have revised the TO in order to account for the semantics of the new lexical categories being encoded.</Paragraph>
    <Paragraph position="6"> In this paper we describe the main changes made to the EWN framework in order to encode information on adjectives in IWN. Firstly, we provide a brief overview of the WN treatment of adjectives. Then, we discuss the set of relations being encoded for this category in IWN. Finally, we show the integration made to the EWN TO.</Paragraph>
    <Paragraph position="7"> We then conclude the paper by adding some remarks on the utility of the data being encoded for computational applications.</Paragraph>
  </Section>
class="xml-element"></Paper>