Dependency Distance as a Metric of Language Comprehension Difficulty

Abstract

Linguistic complexity is a measure of the cognitive difficulty of human \nlanguage processing. The present paper proposes dependency distance, in the \nframework of dependency grammar, as an insightful metric of complexity. \nThree hypotheses are formulated: (1) The human language parser prefers linear \norders that minimize the average dependency distance of the recognized \nsentence (2) There is a threshold that the average dependency distance of most \nsentences or texts of human languages does not exceed (3) Grammar and \ncognition combine to keep dependency distance within the threshold. Twenty \ncorpora from different languages with dependency syntactic annotation are used \nto test these hypotheses. The paper reports the average dependency distance in \nthese corpora and analyzes the factors which influence dependency distance. \nThe findings — that average dependency distance has a tendency to be \nminimized in human language and that there is a threshold of less than 3 words \nin average dependency distance and grammar plays an important role in \nconstraining distance —support all three hypotheses, although some questions \nare still open for further research.