PolVerbNet: an experimental database for Polish verbs

PolVerbNet: an experimental database for Polish verbs Barbara Gawronska University of Skövde Dept. of Languages Box 408, S-541 28 Skövde, Sweden barbara.gawronska@isp.his.se Abstract The semantics of verbs implies, as is known, a great number of difficulties, when it is to be represented in a computational lexicon. The Slavic languages are especially challenging in respect of this task because of the huge complexity of verbs, where the stems are combined with prefixes indicating aspect and Aktionsart. The current paper describes an approach to build PolVerbNet, a database for Polish verbs, considering the internal structure of the aspect-Aktionsart system. PolVerbNet is thus implemented in a larger English- Polish MT-system, which incorporates WordNet. We report our translation procedure and the system’s performance is evaluated and discussed. Keywords verbs, WordNet, aspect, Aktionsart, Polish Introduction: Aspect and Aktionsart in Machine Translation Commercial MT-systems of today are mostly quite good at handling word order, morphological agreement within noun phrases and the choice of grammatical case forms. Aspect and Aktionsart remain a problem. When testing an on-line English-Polish translation system (www.poltran.com) we achieved very poor results with regard to the choice of Aktionsart (only 21% recall and ca 28% precision for an input consisting of 80 sentences). Most of the output was incomprehensible, although other features of POLTRAN indicate that the system includes a great deal of linguistic knowledge. There is a gap between practical MT-applications for Slavic languages and the theoretical work on aspect and Aktionsart (enumerating all relevant references would fill several pages; thus, we only mention some work related to computational linguistics, like Reichenbach, 1947; Johnson, 1981; Maegaard, 1982; van Eynde et al, 1985; Dowty, 1986; Santos, 1992). In implementation-oriented computational research, most effort has been made with tense and aspect systems in Romance and Asian languages (Han et al., 2000; Murata et al., 1999; Palmer & Wu, 1995; Dorr et al., 2000; Santos,1992). The notions “aspect” and “Aktionsart” have been defined in many different ways. Although there is no general agreement as to the details in the proposed definitions, most authors characterize the distinction between the perfective and the imperfective aspect as the difference between events that are “countable”, “bounded”, “unique”, “constant” (perfective), and those that are “unbounded”, with no clear temporal limits (imperfective). The term Aktionsart is most frequently used when referring to the internal spatiotemporal structure of the event. Certain authors (e.g. Comrie 1976, Langacker 1982) include the notion of Aktionsart into the one of aspect. Furthermore, there is no agreement as to the question of universality of the two categories. The most plausible solution to the universality problem seems to be that all languages are able to express “an aspectual perspective” (Nakhimovsky 1987), and that all languages make use of different Aktionsarts, but the overt markers of these categories are different for different language groups. Even languages belonging to the same family display certain differences as to the internal structure of the aspect/Aktionsart systems (see figures 1a-1c in the next section). Thus, an analysis of an aspectual system in Korean …

Content truncated to 500 words