Dependency distance minimization in discourse structure: universality and individuality compared with that in syntactic structure

Abstract

Dependency distance (DD) traditionally refers to the linear distance between two related words in the dependency structure of a sentence (i.e., syntactic-DD). Based on this concept, previous studies have found that language follows a universal rule of DD minimization (DDM), which states that short DDs occur more frequently than long DDs. But if the DDM as a structural rule is universal, does it extend to other levels of language structure? If so, are there any individualities at different levels? This paper proposes the concept of discourse-DD, i.e., the linear distance between two related clauses in a text, to investigate whether the DDM exists at the discourse level. We present a statistical study of discourse-DD using a discourse dependency corpus and compare the statistical results with the syntactic structure. The results show that discourse structure does indeed show the DDM tendency, and even more strongly than syntactic structure. The reasons for this may be that discourse patterns are more flexible than syntactic patterns and clause functions are more dynamic than word functions, and in this case, a stronger DDM is required for discourse comprehension. This paper also explores how linguistic factors such as discourse length, complex sentences, and connectives influence discourse-DDM. The paper deepens the research on DDM of language.