Comprehensive assessment of the significance of informative features to assess the influence on the society based on the analysis of the published text information

Abstract

In the course of the research on assessment of open information sources influence on the
society, the relevant informative features of this influence were proposed. For the purpose of testing and
evaluation of their significance, the experts were provided with the news publications. The text corpus with
meta-annotation was formed on the basis of Tengrinews and Central Asia Monitor news portals. According to
the results of the expert annotation of the text corpus, the analysis and the comprehensive assessment of the
significance of the developed informative features was carried out.
The procedure of comprehensive assessment includes construction of additive indices used in the
statistical model of significance set. The article describes the approach to data collection, also the mathematical
model was developed and tested on the formed text corpus.
According to the results of a multiple regression model for significance evaluation, a statistically
significant effect was established between the influence of the reliability, tone, and unbiasedness indices on the
“total quantity of “views” variable. Moreover, the model’s R-square showed a moderately good determination
coefficient (0.44) which indicates that the model reflected the phenomenon under study relatively well. The
results of statistical modeling can be used in future for the text mining procedure and ultimately for creation of
automated machine learning tool. In addition, these results will be used in the development of a methodology
for assessing the influence of publications on the society