Longitudinal studies of language errors based on a German-language learner corpus

In the age of digitalization and the active spread of corpus technologies in linguistic education, linguodidactics specialists are constantly discovering new opportunities in working with big data. One relatively new phenomenon in Russian education is the collection of corpora of student texts in a foreign language. It’s possibilities for linguodidactical research depend primarily on the duration of the data collection and on the markup that corpus contains. The article focuses on the corpus of German-language student texts PACT (Petrozavodsk annotated corpus of texts) and longitudinal research of types of linguistic mistakes made by students throughout the study of the German language for 5 years. The result of the research is statistics for 90 classes of errors, divided into 7 major groups – grammar, vocabulary, orthography, punctuation, discourse, omissions and superfluous elements – and the dynamics of these statistics over the 5 years of German language study. Comparison of the most frequent errors made by 1st and 5th year students respectively shows that subjects causing the most problems for students during all years of study are lexeme selection, orthography, omissions in text, punctuation and reverse word order. At the end of study problems with indefinite articles, adjective and noun declension, formation of plural form and gender of nouns are giving way to other issues such as superfluous elements in text, logic, word order in subordinate sentences and stylistic errors.

learner corpus, German as a foreign language, language errors, educational data mining