STATISTICAL FEATURES OF ENGINEERING TEXTS

The presented article describes statistical data that characterize probabilistic-statistical models of three technical specialties.The material was text corpora of the specialties chemical engineering, automotive engineering and electrical engineering.The text corpora were compiled on the basis of scientific articles published in the corresponding journals in the USA and Great Britain.At the beginning, the authors assumed that since all three text corpora relate to scientific and technical discourse and do not interact in terms of general scientific topics, but represent completely different branches of technology, they could create an example of style-distinguishing marker in the course of the procedure of comparison.However, different approaches to the formation of text corpora, and, above all, the principles and goals of creating the semantic space of all three specialties led to a sharp difference between the quantitative values of all three frequency dictionaries.The authors of the probabilistic-statistical model of chemical engineering sought to evenly distribute text units across the frequency dictionary and, for this purpose, selected texts that had practically common technical problems, and, accordingly, repeated lexical units that described these problems.This was easily achieved by creating a simple and fairly limited scheme of semantic space, including a few problems that relate to the subjects of this specialty.The same was done by the authors of the text corpus of automotive engineering.As to the authors of the text corpus of electrical engineering they attempted to include practically all scientific and technical problems of this specialty in the semantic space.Therefore, firstly, within the frequency dictionary of electrical engineering itself, sharp differences in quantitative values were observed between the high-frequency and low-frequency zones.Secondly, when comparing the quantitative values of three probabilistic-statistical models, which were expected to be practically identical, the data of the dictionary of electrical engineering were completely different from the dictionaries of chemical engineering and automobile engineering.As a result of the study, the authors proposed to observe one of two conditions when forming text corpora: if it is supposed to consider as many technical problems as possible, then it is necessary to either significantly increase the size of text corpus, or to approach the selection of texts more carefully.