Dependency Grammar constraints

Question

I study dependency grammar, DG and I have a question regarding constraints of DG.

I do understand why do we need constituency grammars, CG and DG, however I don't completely understand the connections between them.

Constraints of DG:

1) every constituent has only one governor.

Very simple constraint, the same we can observe on the tree of CG, there is just no nodes with two parents.

2) dependency grammar forms tree, there is no cycles.

Completely makes sense, the same is in CG.

3) Projectivity. There is no intersections between the arcs of dependency tree.

This is a strange constraint, it doesn't follow from the CG, because the dependency relation we define only on the level of DG, there is no such a thing in CG. And for some reason we should define them so that the arcs don't intersect. But what the reason to do so? And in order to do so we need to change intuitive and simple rules of dependency.

As I understand the DG is not a particular case of CG, we cannot derive DG from CG, just because there is no constant set of rules of defining dependency, so DG contains some new information CG doesn't have. If not projectivity constraint, could we talk about constant set of rules and say the DG is the special case of CF?

In short, what are the connection between DG and CG? What the role of projectivity in DG?

Answer 1

Projectivity is a concept that applies to both dependency grammars (DGs) and constituency grammars (CGs). The extent to which it is applicable to both approaches to syntax is discussed and illustrated with numerous trees here:

https://en.wikipedia.org/wiki/Discontinuity_%28linguistics%29

The term itself, i.e. projectivity, was introduced in the 1960s in early explorations of word order in DG formalisms. The term is generally NOT used in CGs, although the concept is perfectly valid for both formalisms (DGs and CGs). More common terms that the projectivity concept underlies are long distance dependency, discontinuous constituent, and discontinuity. These terms all point to same thing, i.e. crossing lines in the syntax tree (that are then overcome in various ways, e.g. by movement or feature passing).

Thus to answer part of the question (What is the role of projectivity in DG?) as directly as possible, the role of projectivity is to identify discontinuities (i.e. crossing lines in the syntax tree). When the crossing lines obtain, one is dealing with a discontinuity. There are numerous types of discontinuities, those associated with wh-fronting, topicalization, scrambling, and extraposition. Wikipedia has articles on each of these types of discontinuities, and each article has both DG trees and CG trees for illustration.

Addressing the other part of the question (What is the connection between DG and CG?) is a difficult task. I disagree with Jlawler's comment in this area (assuming I understand it). There are big differences between DGs and CGs. DGs are much, much simpler than CGs at the most basic level. They often posit about half the amount of syntactic structure as (measured in terms of the number of nodes and edges in the syntax trees). You can teach a kid to diagram simple sentences using dependency, but the same cannot be said for constituency. The extra structure that CGs posit allow one to do things that one cannot do with dependency. For instance, c-command is not possible in dependency-based structures. DGs are, however, likely to challenge all the extra structure that CGs necessitate. They argue that it does not increase our understanding of syntax; the simple DG structures get the job done with much less effort.

Finally, the point about whether dependency and constituency are notational variants of the same thing is mostly decided in DG circles. Most DG people reject the notion that they are notational variants of the same thing. This means that, as stated, you can do things with dependency-based structures that you cannot do with constituency-based structures, and vice versa.