GitHub

Trees

Dependency structure in natural language consists of directed relations between words in a sentence.

Simple API for building dependency trees:

DependencyTree functions

DependencyTrees.arcs — Function
arcs(tree::DependencyTree)

Return a vector of (src, dst) tuples, representing all dependency arcs in tree.

source

Projectivity

"Projectivity" is a characteristic of dependency trees.

An arc from a head to a dependent is said to be projective if there is a path from the head to every word that lies between the headprojective and the dependent in the sentence. A dependency tree is then said to be projective if all the arcs that make it up are projective. [1]

Drawing the trees will show that non-projective trees have crossing arcs, and projective trees do not.

DependencyTrees.isprojective — Function
isprojective(tree::DependencyTree)

True when a tree is projective, false when non-projective.

julia> tree = DependencyTree([("john", 2), ("saw", 0), ("a", 4), ("dog", 2), ("yesterday", 2), ("which", 7), ("was", 4), ("a", 9), ("yorkshire", 10), ("terrier", 7)])
   ┌─────────── 0 ROOT
   │        ┌─► 1 john
   └─►┌──┌──└── 2 saw
      │  │  ┌─► 3 a
   ┌──│  └─►└── 4 dog
   │  └───────► 5 yesterday
   │        ┌─► 6 which
┌──└───────►└── 7 was
│           ┌─► 8 a
│        ┌─►└── 9 yorkshire
└───────►└───── 10 terrier

julia> isprojective(tree)
false

julia> tree2 = DependencyTree([("john", 2), ("saw", 0), ("a", 4), ("dog", 2), ("which", 6), ("was", 4), ("a", 8), ("yorkshire", 9), ("terrier", 6), ("yesterday", 2)])
┌───────────────── 0 ROOT
│              ┌─► 1 john
└─►┌────────┌──└── 2 saw
   │        │  ┌─► 3 a
   │     ┌──└─►└── 4 dog
   │     │     ┌─► 5 which
   │  ┌──└────►└── 6 was
   │  │        ┌─► 7 a
   │  │     ┌─►└── 8 yorkshire
   │  └────►└───── 9 terrier
   └─────────────► 10 yesterday


julia> isprojective(tree2)
true
source

Tokens

Visualizing Trees

DependencyTrees.jl by default displays DependencyTrees by drawing arrows representing the dependency arcs. For example,