Challenges of Adjective Mapping between plWordNet and Princeton WordNet

Challenges of Adjective Mapping between plWordNet and Princeton WordNet 1Ewa Rudnicka, 2Wojciech Witkowski, 3Katarzyna Podlaska 1,3Wrocław University of Technology, 2University of Wrocław Poland email: ewa.rudnicka@pwr.edu.pl, wojciech.witkowski@uwr.edu.pl, k.podlaska@hotmail.com Abstract The paper presents the strategy and results of mapping adjective synsets between plWordNet (the wordnet of Polish, cf. Piasecki et al. 2009, Maziarz et al. 2013) and Princeton WordNet (cf. Fellbaum 1998). The main challenge of this enterprise has been very different synset relation structures in the two networks: horizontal, dumbbell-model based in PWN and vertical, hyponymy-based in plWN. Moreover, the two wordnets display differences in the grouping of adjectives into semantic domains and in the size of the adjective category. The handle the above contrasts, a series of automatic prompt algorithms and a manual mapping procedure relying on corresponding synset and lexical unit relations as well as on inter-lingual relations between noun synsets were proposed in the pilot stage of mapping (Rudnicka et al. 2015). In the paper we discuss the final results of the mapping process as well as explain example mapping choices. Suggestions for further development of mapping are also given. Keywords: wordnets, inter-lingual mapping, adjective relation structure 1. Introduction The goal of this paper is to present solutions developed for the purposes of mapping two different relation structures describing adjectives in Princeton WordNet (henceforth, PWN, cf. Fellbaum 1998) and in plWordNet (Polish wordnet, henceforth, plWN, cf. Maziarz et al. 2013). plWordNet is one of the few world wordnets built fairly independently of PWN with the help of a unique method of lexico-semantic relations extraction from large text corpora (Piasecki et al. 2009). Nevertheless, the actual construction process is manual - a supervised team of lexicographers verifies automatic hints in lexicographic resources and only then introduces them into a database. Thus, it belongs in with the so called merge approach (cf. Vossen et al. 2002). It allows for a more truthful language description, but leads to differences in lexical coverage and relational structures between wordnets. This is clearly the case of plWN and PWN adjective domain, which has vertical, hyponymy-based structure in plWN (akin to that of nouns and verbs) (cf. Maziarz et al. 2012), and a horizontal, dumbbell model-based structure in PWN (cf. Miller 1998, Sheinman et al. 2013). Moreover, plWN has a slightly more fine-grained set of semantic domains comprising qualitative, relational and material adjectives, while PWN distinguishes only relational adjectives from the general adjective category. Another key issue in the process of adjective mapping is wordnet size. At the beginning the sizes of plWN and PWN adjective domains were comparable1. However, the process of mapping has been carried out parallel to the process of the extension of adjective category in plWN, and at the final stage of 1 This is based on the data from plWN 2.1 version, downloadable from http://nlp.pwr.wroc.pl/plwordnet/download/?lang=pl mapping the number of adjective synsets in plWN outgrew that of PWN twice2. In view of the above mentioned contrasts, the design of the mapping strategy for plWN and PWN adjectives had been a real …

Content truncated to 500 words