My focus article in Language Under Discussion from 2018 argued that dependency syntax is both simpler and more accurate than phrase structure syntax with respect to the results delivered by tests for constituents. Four linguists (Richard Hudson, Lachlan Mackenzie, Stefan Müller, and Matthew Reeve) have responded to my focus article with discussion notes, challenging aspects of my message in various ways. In this article, I respond to the counterarguments produced in the discussion notes. In order to address one of the main counterarguments, having to do with scope and meaning compositionality, I introduce a new unit of dependency syntax, namely the colocant. My claim is that aspects of scope and meaning compositionality, for which phrase structure is deemed necessary, can be addressed in terms of colocants. Hence, scope phenomena and the manner in which meaning is composed can no longer be construed as an argument against dependency syntax and in favor of the necessity of phrase structure.
Discontinuous Statistical Machine Translation with Target-Side Dependency Syntax
