Construction and longitudinal analysis of a learner corpus of Japanese on the elementary, intermediate and advanced level

Magistrsko delo opisuje izgradnjo korpusa usvajanja jezika študentov japonščine na osnovni, srednji in višji ravni in analizo slovničnih napak v njem. Namen dela je ustvariti javno dostopen korpus s spisi učencev japonščine na različnih nivojih jezikovne sposobnosti in s tem ustvariti orodje, ki uporabnikom omogoča enostavni dostop do podatkov o napakah v spisih slovenskih učencev japonščine. Drugi namen je s pomočjo analize napak na posamezni ravni identificirati slovnične strukture, pri katerih najpogosteje prihaja do napak in s primerjavo rezultatov na posamezni ravni analizirati spremembe v trendih pojavitve teh napak. Nastali korpus vsebuje 237 besedil, katerih je bilo označenih in kategoriziranih 1778 slovničnih napak. Na posamezni ravni so bile določene kategorije z največjim številom napak, ker pa korpus vsebuje tudi longitudinalne podatke, so bili tudi ti podatki analizirani in rezultati obeh analiz primerjani. Strukture, pri katerih najpogosteje prihaja do napak, so nato bile še podrobneje analizirane. Rezultati dela so lahko koristni za izboljšanje učnega procesa, nastali korpus pa predstavlja prvi javno dostopni specializirani korpus usvajanja japonščine s podatki slovenskih učencev in s tem dragocen jezikovni vir, tako za učitelje, učence kot tudi raziskovalce na področju usvajanja jezikov in korpusnega jezikoslovja.This Master\u27s thesis describes the construction of a learner corpus of Japanese on the elementary, intermediate and advanced levels and the analysis of grammatical errors in it. The purpose of the master\u27s thesis is to create a publicly accessible corpus containing composition data of Japanese learners at different levels of language ability and thereby create a tool that enables users to easily access data about errors in the writings of Slovenian learners of Japanese. Another purpose is to use the constructed corpus to determine which grammatical structures are most common on each level and, by comparing the results between levels, to analyze changes in the trends in the occurrence of various types of errors. As a result, a corpus containing 237 texts, and 1778 marked and categorized grammatical errors was created. For each level of proficiency, the categories with the highest number of errors were determined. Because the corpus also contains longitudinal data, the data was also analyzed and the results of both analyses then compared. Finally, the structures with the most frequent errors were analyzed in more detail. The results of the analysis can be useful to Japanese teachers in the execution of the learning process, while the resulting corpus represents the first publicly available specialized corpus of Japanese acquisition with data from Slovenian students and thus a valuable linguistic resource, both for teachers, students, and researchers in the field of language acquisition and corpus linguistics