Multivariate analysis of the importance of semantic scales
Abstract
Multivariate analysis of the importance of semantic scales
Incoming article date: 30.09.2016The study was conducted on more practical material for monitoring the emotional perception of such social phenomena as local and Central political orders, their own level of relative deprivation, cultural infrastructure and cultural life of the city as a whole, their own emotional state. Since 2009, performed a survey of several thousand respondents on the subject with the use of technology of semantic differential. The aim of this work is the ranking used semantic scales in terms of relevance to respondents and their selection of the optimal set. It is necessary to increase the adequacy of the measurement and reduce errors. For calibrating semantic scales in the questionnaire, respondents were asked to evaluate not only the real object, but two perfect (positive completely satisfied respondents, and negative - totally unacceptable) of the same class. The distance between these images allowed us to assess the importance of each scale. For all the above objects of study, and each used a semantic scale was obtained of the distribution function of the significance level for the respondents. Obtained that this function is strongly smeared, across the range of values (in our normalization is from zero to one). Thus the studied scales are divided into three sets. A group of scales with good significance (for example, the scale of smart-stupid), for which zero significance show about 15% of the respondents, and the maximum is about two times more. Scales of moderate importance (for example, scale fast-slow), for which the significance in the whole range from 0 to 1 more or less same. And a group of poor significance (sharp, rounded etc.), for which the majority of the respondents poorly distinguish the corresponding characteristic. In the method of semantic differential emotional perception of the object is determined not by individual semantic scales, and the set of indicators. And for selection of optimal scales analysis of their individual significances insufficient. Need more total study of the significance of sets of scales. For these purposes, on the basis of a package of multidimensional data mining, Cognos (IBM) has developed a multidimensional model of the importance of semantic scales. It allows you to explore the sections with a fixed value for different sets of scales. On the basis of the analysis of the 20 initial semantic features were selected 8. More compared to the minimum (four) number provided the overlap does not distinguish between scales and the increase in the total significance of the whole set. For all the above-mentioned objects of study (in addition to their own emotional state of the Respondent) identified the optimal set of scales provides sufficient total significance. Questionnaires to assess emotional state showed that the individual significance of semantic scales for them are much lower than for other objects. There are only two groups of scales. Average, with approximately equal proportions of respondents across the range of significance. And bad, with the maximum distribution function at low important part. Signs with good importance at all. As optimum it is possible to take a set, the same as for other objects. But neither he, nor even a complete set of 20 scales will not provide the proper total value. When interpreting data on the emotional state in the work it is recommended to use the conventional Cartesian distance, as previously, we tested a weighted metric of urban areas (the Minkowski metric). Individual significance for each scale as coefficients of balanced. Scale with zero significance of this will automatically be eliminated, thereby improving the total significance of the feature set.
Keywords: the method of semantic differential, affective component, social objects, semantic scale, ideal constructs, significance, distribution function, cumulative importance, a multidimensional model, sections, sections, the choice of the optimal set