A Proposal for a Part-of-Speech Tagset for the Albanian Language

A Proposal for a Part-of-Speech Tagset for the Albanian Language Besim Kabashi, Thomas Proisl Friedrich-Alexander-Universität Erlangen-Nürnberg Bismarckstraße 6, 91054 Erlangen, Germany besim.kabashi@fau.de, thomas.proisl@fau.de Abstract Part-of-speech tagging is a basic step in Natural Language Processing that is often essential. Labeling the word forms of a text with fine-grained word-class information adds new value to it and can be a prerequisite for downstream processes like a dependency parser. Corpus linguists and lexicographers also benefit greatly from the improved search options that are available with tagged data. The Albanian language has some properties that pose difficulties for the creation of a part-of-speech tagset. In this paper, we discuss those difficulties and present a proposal for a part-of-speech tagset that can adequately represent the underlying linguistic phenomena. Keywords: Albanian; Tagset; Morphosyntactic annotation 1. Introduction While there is no specification from EAGLE for Albanian, there are some works about morphological annotation and tagging for the Albanian language, cf. Trommer and Kallulli (2004), Kabashi (2004), Piton et al. (2007), Piton and Lagji (2008), Hasanaj (2012), Kadriu (2013) and Kabashi (2015). In the field of computational morphology, Trommer and Kallulli’s (2004) analyzer, which can be used as a kind of tagger, seems to cover the main Albanian inflection types. Its output format follows the EAGLE guide- lines standard (Leech and Wilson, 1999), i. e. tags con- sist of sets of attribute-value pairs, and looks either like this: [ cat:n case:nom num:sg def:+ gen: fem ] or, more compact, like this: [ n nom sg +def fem ]. Alter- natively, tags can be collapsed, e. g. the two tags tags [ n nom sg -def fem ] and [ n acc sg -def fem ] can be merged into [ n nom,acc sg -def fem ]. The tagset distinguishes between 17 broad word-class labels, seven of which are reserved for subtypes of pronouns, one for the preposed article, one for sentence equivalents, one for the participle form of the verb, and the rest for the traditional parts of speech, namely nouns, verbs, adjectives, adverbs, particles, prepositions, and conjunctions. The goal of the tool by Piton, Lagji and Përnaska (2007; 2008), is to cover the inflection of Albanian. Their analysis also allows for preposed articles, if the word forms can have them, cf. the analysis of the adjective i artë, engl. golden: i artë,A+FLX=Adj2+ei+m+s. It is not clear how many tags the tagset has or on how many major word- classes it is based, though it seems to cover at least the ten traditional parts of speech. The part-of-speech tagging modell for Albanian presented by Hasanaj (2012) consists of a basic tagset with 16 tags and a large tagset with 326 tags. In the basic tagset, there are three tags for delimiters, two for special cases (short forms of pronouns), one for articles, and ten for the traditional parts of speech. The large tagset encodes the word-classes, e. g. JJ [=Adj.], NN [=Noun], VB [=Verb] or PR [=Pronoun], and additional features, e. g. Number (Sg. and Pl.). A tag …

Content truncated to 500 words