The effects of sentence length on dependency distance, dependency direction and the implications–Based on a parallel English–Chinese dependency treebank

Abstract

Dependency distance is closely related to human working memory capacity, but is also influenced by other non-cognitive factors. Studies of dependency distance contribute to the understanding of the universalities and peculiarities of languages as well as human cognitive processes in language. Forty two sentence sets were selected from a parallel English–Chinese dependency treebank to examine the progressive properties of dependency distance with the change of sentence length in the two languages. It was found that: (1) the probability distribution models of dependency distance of both languages are not affected by either sentence length or the type of language; (2) the quantity of adjacent dependencies in the two languages are identical, but the quantity of adjacent dependencies of Chinese fluctuates within a limited range, while that of English shows a falling tendency; (3) the mean dependency distances (MDDs) of Chinese are always higher than those of English, and both MDDs show slight ascending trends; (4) compared with dependency distance, dependency direction is a more reliable metric for language classification. These findings suggest that: (1) the universal cognition mechanism may be the major factor affecting the general traits of dependency distance, while language-related factors such as sentence length may affect certain traits of dependency distance; and (2) Chinese taxes working memory more than English.