Conjunct Lengths in English, Dependency Length Minimization, and Dependency Structure of Coordination

Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 15494–15512 July 9-14, 2023 ©2023 Association for Computational Linguistics Conjunct Lengths in English, Dependency Length Minimization, and Dependency Structure of Coordination Adam Przepiórkowski ICS Polish Academy of Sciences and University of Warsaw adamp@ipipan.waw.pl Michał Wo´zniak University of Warsaw m.wozniak60@student.uw.edu.pl Abstract This paper confirms that, in English binary co- ordinations, left conjuncts tend to be shorter than right conjuncts, regardless of the position of the governor of the coordination. We demon- strate that this tendency becomes stronger when length differences are greater, but only when the governor is on the left or absent, not when it is on the right. We explain this effect via De- pendency Length Minimization and we show that this explanation provides support for sym- metrical dependency structures of coordination, as opposed to structures headed by the first con- junct. 1 Introduction It has been observed for various particular types of coordination in English that left-most conjuncts tend to be shorter than right-most conjuncts (e.g., Gibson et al. 1996, Temperley 2005, Lohmann 2014). This is illustrated in (1) from the Penn Treebank (PTB; Marcus et al. 1993), where the left conjunct, ship, is shorter than the right conjunct, hope I get paid, in terms of the number of words (1 vs. 4), the number of syllables (1 vs. 4), and the number of characters (4 vs. 15, including spaces). (1) I’m going to [[ship] and [hope I get paid]]. However, to the best of our knowledge, there are no demonstrations of this effect that would take all kinds of coordinations into account and that would use various length metrics. Filling this gap is the first contribution of this paper. There is even less work that asks whether it is re- ally the left-to-right order of conjuncts that matters here, or whether it is perhaps the closeness to the external head – henceforth, governor. (1), where the governor to is on the left, lends support to both – “leftness” and “closeness” – hypotheses. But the two hypotheses make different predictions when the governor is on the right, as in (2)–(3). (2) [[Walter Sisulu] and [the African National Congress]] came home yesterday. (3) [[My younger daughter] and [I]] are fine. (2), where the left conjunct is shorter (2 vs. 4 words, 5 vs. 9 syllables, 13 vs. 29 characters), only sup- ports the “leftness” hypothesis, while (3), where the right conjunct is shorter (1 vs. 3 words, 1 vs. 5 syllables, 1 vs. 19 characters), only supports the “closeness” hypothesis. The second contribution of this paper is to estab- lish two facts regarding the influence of the gover- nor on conjunct lengths in English. The first fact is that left conjuncts tend to be shorter even when the governor is on the right, which immediately invali- dates the “closeness” hypothesis. The second ob- servation is more interesting and has important con- sequences for dependency theories of coordination: the position of the governor …

Content truncated to 500 words