Error Types in the Learner Corpus of the Second Baltic Language

Abstract

Errors in language learning are seen as normal and even necessary. However, researching them is often undermined by the need for a clear definition what is or should be considered an error and by the lack of an error taxonomy. This paper shortly discusses the notion of error in various contexts, especially in learner corpora research. Then it offers an error taxonomy that was created for error-tagging a learner corpus of Baltic languages. The aim of the study is to create a taxonomy that is suitable for annotating beginner texts of Latvian and Lithuanian, and efficient in use. The taxonomy is based on the previous work of S. Granger who identified error types for a learner corpus of French. These error types are reviewed, modified and/or replaced where necessary in order to match the structure of Latvian and Lithuanian languages. 5 error types (form; morphology and word-formation; syntax; vocabulary; punctuation) with 29 subtypes are distinguished. Those are described in the article along with examples from the corpus. The taxonomy is now being used for annotation the learner corpus of the second Baltic language which provides researchers with valuable material on language learning outcomes.