Research of pupils' Linguistic errors on the base of learner corpus

The paper presents an on-going research project to develop the specialized learner corpus that consists of essays written by young (8-10 years old) language learners at schools with a profound teaching of the foreign language in Ulyanovsk, Russia. The learner corpus contains about 500 works, produced on the most widespread topics of the standard educational materials used by English teachers of the 2nd and 3d grades at Russian schools. The main research areas of the project are the identification and analysis of lexical, spelling and grammatical errors in the essays. We have studied the methods of classification of vocabulary and grammar-related mistakes as well as techniques of their coding and tagging in the electronic corpus. We also provide the errors statistics, which allows us to carry out the comprehensive linguistic analysis of each mistake type for the applied-linguistic purposes. The developed meta-language was used to tag the mistakes found in the corpus and to obtain the statistical data with the help of the open concordance programs. The total number of the analyzed and tagged mistakes in the learner corpus is more than 3,500, among which we identify 1,500 spelling and lexical errors. Spelling errors make the majority and are 56% of the total of all the lexical mistakes in comparison with the quantitative indices of other mistakes. By the computer analysis of more than 2,000 grammatical errors in the learner corpus we found out that the most frequent mistakes in written works by the elementary graders are the wrong usage of articles and punctuation marks. In our opinion, the prevalence and frequency of the tagged error types of both classes in the essays of the focus group of schoolchildren have the certain psycholinguistic reasons. The analysis results are useful in applied linguistics and didactic domains, and prove the potential of using corpus techniques with ESL learners and teachers.