Google books ngram: Problems of representativeness and data reliability

Abstract

© Springer Nature Switzerland AG 2020. The article discusses representativeness of Google Books Ngram as a multi-purpose corpus. Criticism of the corpus is analysed and discussed. A comparative study of the GBN data and the data obtained using the Russian National Corpus and the General Internet Corpus of Russian is performed to show that the Google Books Ngram corpus can be successfully used for corpus-based studies. A new concept “diachronically balanced corpus” is introduced. Besides, the article describes the problems of word spelling and metadata errors presented in the GBN corpus and proposes possible ways of improving quality of the GBN data.