The text preprocessing influence analyze for author identification problem by bigram method
Abstract:
On the example of sufficiently representative number of authors and texts, a comparative analysis of the impact of text preprocessing programs on the possibility of identifying authors is carried out. The question of the sensitivity of the identification error by the proportion of changes in the source text is investigated. It is shown that the author's originality is preserved after preprocessing almost at the level of the original text.
Keywords:
machine classification, text preprocessing, bigram distribution, author identification
Publication language:russian, pages:18
Research direction:
Mathematical modelling in actual problems of science and technics
