INTELLIGENT SYSTEMS AND TECHNOLOGIES

Entropy information approach for evaluating the quality of machine translation texts

G. N. Akhobadze, E. Yu. Rusyaeva, A. V. Poltavsky

V. A. Trapeznikov Institute of Control Sciences of Russian Academy of Sciences, Moscow, Russia

Abstract: Based on the entropy-information approach, a method for assessing the quality of machine translation texts is proposed. An analysis is given of the dispersion and entropy coefficients of concordance used to assess the consistency of expert opinions with close rankings of various objects. The prospects of using the entropy coefficient of concordance, which makes it possible to fix the fact of division of opinions into two opposing groups, are substantiated. This provision is important for the study, since in this method of expert evaluation of text translations, it is important to take into account the different opinions of several experts involved in the examination. Examples of calculating the entropy coefficient of concordance with a changing rank system, the number of experts and ranked evaluation objects are given.

Keywords: concrodation coefficient, text ranking, expert opinion, information entropy, rank, expert opinion matrix, probabilistic parameters.

DOI: 10.14357/20718632230402

Bibliographic databases:

•