Автоматизированная оценка сложности текста: Корпусное обследование текстов разных категорий и предметных областей.
The paper analyses data of automated evaluation of syntactical complexity of texts
from four comparable corpora - two of expert texts and two of learner texts - in
two subject areas. The research applied two different computer tools– Udpipe
Stanford parser and AntConc set of programs. The following conclusions were
made from the counts in the four corpora:
- Expert scholarly texts demonstrated high clausal and phrasal complexity,
and its level was higher in Business Studies than in Economics;
- Clausal complexity in expert texts was much higher than in student texts,
while phrasal complexity was at about the same level as, or even
insignificantly lower than those of students.
The largest difference was attested for genitive construction with preposition of,
which was 1.7 times as frequent in student texts in both subject areas as in expert
texts. The results contradict the conclusion of the prior research that with the
growth in proficiency clausal complexity decreases, while phrasal complexity