Уравнение регрессии для прогноза удобочитаемости шрифта на основе его характеристик

Abstract

The development of font readability prediction is a relevant tool for gaining new knowledge about font quality as a means of conveying information. Using different methods helps to formulate the holistic content of the concept of quality for a given object.The article describes the experience of building an equation based on the regression-correlation analysis linking a number of concepts reflecting the characteristics of a font, for example, ‘elegant’, ‘dense’, ‘easy’, ‘expressive’, ‘comfortable’, ‘readable’, ‘beautiful’, and a high or low readability, which can be used to predict the quality of the font.As a result of the experience of regression analysis of data obtained by interviewing users of the font by the proposed method, it was revealed that:1) A statistically significant regression equation can be obtained that connects a number of concepts that reflect the characteristics of a font and a group of high and low readability defined separately by the method of pairwise comparisons.2) According to the algorithm, the following independent variables that have little effect on the final result can be excluded from the equation based on the correlation analysis: ‘stylish’, ‘capacious’, ‘slender’ and ‘attractive’.3) It was revealed that the factors ‘convenient’, ‘readable’, ‘beautiful’ have a directly proportional connection with the dependent variable, and the weight of the coefficient within this model is average for them.4) The equation can be used to predict if there are data obtained from a survey using the methodology described in the article.This work expands the methodological field of quantitative assessment of font quality, proves the possibility of constructing regression equations that predict readability not only by geometric parameters of the font, but also on the basis of data from the font perception area by users.