Identifying Errors in Russian Web Corpora

Abstract The explosion of the Web leads to production large amounts texts and inevitably influences their quality. Errors that tend occur more often can distort results, especially when are used for scientific purposes, in language teaching or learning. Hence, there is a need examine existing corpora based on web clean up data, which may contain such “noisy” fragments. In our study, we deal with problem errors analyze Aranea Russicum Maximum corpus. Among errors, name, above all, encoding incorrect font types, as well segments written other languages. These phenomena result morphological analysis lemmatization, frequency distortion, fact lexical units cannot be found therefore displayed corpus users. paper focuses describes types outlines possible ways eliminate them.