The Use of Text Alignment in Semi-Automatic Error Analysis: Use Case in the Development of the Corpus of the Latvian Language Learners

The Use of Text Alignment in Semi-Automatic Error Analysis: Use Case in the Development of the Corpus of the Latvian Language Learners Roberts Darģis1, Ilze Auziņa2, Kristīne Levāne-Petrova3 Faculty of Computing, University of Latvia1, Institute of Mathematics and Computer Science, University of Latvia2,3 Raina bulvaris 19, Riga, LV-1459, Latvia1, Raina bulvaris 29, Riga, LV-1459, Latvia2,3 { roberts.dargis, ilze.auzina, kristine.levane-petrova }@lumii.lv Abstract This article presents a different method for creation of error annotated corpora. The approach suggested in this paper consists of multiple parts - text correction, automated morphological analysis, automated text alignment and error annotation. Error annotation can easily be semi-automated with a rule-based system, similar to the one used in this paper. The text correction can also be semi- automated using a rule-based system or even machine learning. The use of the text correction, word, and letter alignment enables more in-depth analysis of errors types, providing opportunities for quantitative research. The proposed method has been approbated in the development of the corpus of the Latvian language learners. Spelling, punctuation, grammatical, syntactic and lexical errors are annotated in the corpus. Text that is not understandable is marked as unclear for additional analysis. The method can easily be adapted for the development of error corpora in any other languages with relatively free word order. The highest gain from this method will be for highly inflected languages with rich morphology. Keywords: learner corpus, error annotation, word alignment 1. Introduction The purpose of this article is to describe the error- annotating methodology and the tool that is used to annotate The Corpus of the Latvian Language Learners (Latvian as L2 and foreign). As the Sylviane Granger admits, the learner corpora constitute a new resource for second language acquisition and foreign language teaching specialists, especially if they are error-tagged. (Granger, 2003). Appropriately designed learner corpus and consistently annotated errors can provide answers to global questions such as: what is the most frequent type of error, how the native language influence the error type. As the developed corpus includes the texts of different levels of language acquisition the corpus can provide an answer to very specific questions, for example, are mistakes related to noun endings more frequent for B2 or C1 level? Based on the data from the corpus, also different workbooks might be developed for people learning a second language. Latvian is a language with rich morphology and a relatively free word order. Latvian in general can be considered a phonetic language – a language with a relatively simple relationship between orthography and phonology. From the language acquisition perspective, Latvian has several specific properties: short and long vowels and diphthongs, highly inflected language, rather free word order. These properties have to be taken into account in error-annotation. There are many learner corpora for English and last decades learner corpus have been created for other languages as well, for example, French, Swedish, Norwegian, Dutch, Spanish and German (Granger, 2008), and their range is expanding. Currently, The Corpus of the Latvian Language Learners is being created. …

Content truncated to 500 words