Visual Modeling of Turkish Morphology

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 3984–3990 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 3984 Visual Modeling of Turkish Morphology Berke ¨Ozenc¸, Ercan Solak Department of Computer Engineering Is¸ık University, ˙Istanbul, Turkey {berke.ozenc,ercan.solak}@isikun.edu.tr Abstract In this paper, we describe the steps in a visual modeling of Turkish morphology using diagramming tools. We aimed to make modeling easier and more maintainable while automating much of the code generation. We released the resulting analyzer, MorTur, and the diagram conversion tool, DiaMor as free, open-source utilities. MorTur analyzer is also publicly available on its web page as a web service. MorTur and DiaMor are part of our ongoing efforts in building a set of natural language processing tools for Turkic languages under a consistent framework. Keywords: Turkish, morphology, morphological analyzer 1. Introduction Morphological analysis is among the first steps in the natu- ral processing pipeline of morphologically rich languages. Analysis is often preceded by the relatively simpler steps of sentence segmentation and tokenization. Morphological analysis of a token usually yields more than one analysis, in which case a disambiguation step is performed before any further processing. Disambiguation is closely related to syntactic parsing as the distinct analyses yield the first level nodes in distinct parses for a given sentence. The accuracy of an analyzer depends on both the ex- tent of its base lexicon and the fidelity of its morphologi- cal/phonological model of the target language. Construct- ing a base lexicon and a model requires linguistic expertise in the target language and is a labor intensive tasks. In this paper, we demonstrate how the modeling stage can be made easier using visual tools and automated code generation. We used the diagramming tool draw.io (draw.io, 2019) to completely model the morphotactics of Turkish in a dia- gram and we automatically converted the diagram into a finite-state transducer (FST) model using DiaMor ( ¨Ozenc¸ and Solak, 2019c). We provide in the following sections, the details of the model, its linguistic base and its tag set. The resulting analyzer, MorTur, is freely available together with its source. Currently, there are three publicly accessible morphological analyzers for Turkish, (Akın and Akın, 2018), (S¸ahin et al., 2013), (C¸ ¨oltekin, 2014). Although the first computational description of Turkish morphology was given in (Oflazer, 1994), Oflazer’s xfst based implementation is not pub- licly accessible as of December 2019. ITU analyzer described in (S¸ahin et al., 2013) is based on Oflazer’s description in (Oflazer, 1994) and closely follows its conventions and tags. The analyzer’s source code is not publicly available. Still, it can be used through its web page for small amounts of text. It is also possible to use the ana- lyzer as a web service upon permission from its creators. The Zemberek analyzer is an open source java tool, (Akın and Akın, 2018). The implementation is done on a custom implementation of a finite transducer engine. The analyzer can be embedded as part of a larger …

Content truncated to 500 words