Microsyntactic Unit Detection Using Word Embedding Models: Experiments on Slavic Languages

Proceedings of Recent Advances in Natural Language Processing, pages 1265–1273 Varna, Sep 4–6, 2023 https://doi.org/10.26615/978-954-452-092-2_134 1265 Microsyntactic Unit Detection using Word Embedding Models: Experiments on Slavic Languages Iuliia Zaitova, Irina Stenger, Tania Avgustinova Department of Language Science and Technology, Saarland University, Germany izaitova@lsv.uni-saarland.de ira.stenger@mx.uni-saarland.de avgustinova@coli.uni-saarland.de Abstract Microsyntactic units have been defined as language-specific transitional entities between lexicon and grammar, whose idiomatic prop- erties are closely tied to syntax. These units are typically described based on individual constructions, making it difficult to under- stand them comprehensively as a class. This study proposes a novel approach to detect mi- crosyntactic units using Word Embedding Mod- els (WEMs) trained on six Slavic languages, namely Belarusian, Bulgarian, Czech, Polish, Russian, and Ukrainian, and evaluates how well these models capture the nuances of syntactic non-compositionality. To evaluate the models, we develop a cross- lingual inventory of microsyntactic units using the lists of microsyntantic units available at the Russian National Corpus. Our results demon- strate the effectiveness of WEMs in capturing microsyntactic units across all six Slavic lan- guages under analysis. Additionally, we find that WEMs tailored for syntax-based tasks con- sistently outperform other WEMs at the task. Our findings contribute to the theory of mi- crosyntax by providing insights into the de- tection of microsyntactic units and their cross- linguistic properties. 1 Introduction Microsyntactic units, which include syntactic id- ioms and non-standard syntactic constructions, have been defined as language-specific transitional entities between the lexicon and the grammar, id- iomatic properties of which are closely tied to syn- tax (Iomdin, 2017). These units include all the syntactic units that have very specific and even syn- tactic properties and do not fit into the standard syntax (Iomdin, 2015). Recent research efforts have resulted in the development of several linguis- tic resources for microsyntactic analysis, such as a microsyntactic dictionary of Russian, a microsyn- tactically annotated corpus of Russian texts, and a typology of relevant phenomena (Marakasova and Iomdin, 2016; Iomdin, 2016, 2017; Avgustinova and Iomdin, 2019). Given the vast number and diverse nature of mi- crosyntactic phenomena, it is not surprising that they are often described on the basis of individual constructions or small classes of syntactic phrases. In order to gain a more comprehensive and sys- tematic understanding of these phenomena, it is crucial to attempt an analysis of microsyntactic phenomena at scale, rather than in isolation. In this study, we add to the line of research on microsyntax by adapting quantitative and computational meth- ods used in idiom recognition for identification of microsyntactic units in large corpora of texts and across different languages. We apply different types of Word Embedding Models (WEMs) to the task of microsyntactic unit detection, and test their performance on five functional categories of mi- crosyntactic unit (prepositions, adverbials and pred- icatives, parenthetical expressions, conjunctions, and particles) in six Slavic languages (Belarusian, Bulgarian, Czech, Polish, Russian, Ukrainian). Concretely, the contributions of this paper are as follows: 1. We demonstrate that the methods used for id- iom recognition can be applied for …

Content truncated to 500 words