So far, we have represented syntactic structure by means of tree diagrams, but it is sometimes convenient to use bracketed structures (also known as labeled bracketings) instead. In a tree diagram, the domination relation is represented by arranging the dominating node above the material that is dominated. In a bracketed structure, the dominated material is instead enclosed in square brackets that are annotated with the label of the dominating node. The label is usually placed on the inside of the left bracket.
The correspondence between tree diagrams and bracketed structures is illustrated in the following examples.
| (1) | a. |
| b. |
| c. |
|
| (2) | a. | [IP [NPi he ] [I ' [I will ] [VP [NPi t ] [V ' [V wait ] ] ] ] ] | |
| b. | [IP [NPj we ] [I ' [I will ] [VP [NPj t ] [V ' [V ask ] [CP [C ' [C if ] [IP [NPi she ] [I ' [I [past] ] [VP [NPi t ] [V ' [V left ] ] ] ] ] ] ] ] ] ] ] | ||
| c. | [IP [NPk we ] [I ' [I [pres] ] [VP [NPk t ] [V ' [V ' [V drink ] [NP green tea ] ] [PP after lunch ] ] ] ] ] |
Bracketed structures represent exactly the same amount of structural information as tree diagrams. Notice, for instance, that noun phrases and prepositional phrases are both represented without internal structure in the bracketed structures, just as they are in the corresponding tree diagrams. On the other hand, both representations in (1c) and (2c) contain a recursive V' node.
Although tree diagrams tend to be much easier for humans to process, bracketed structures are often preferable as representations of syntactic structure for various reasons. For one thing, they are cheaper to typeset. In large syntactically annotated corpora such as the Penn Treebank, syntactic structure is represented by means of labeled bracketing because the corpora can then be quickly searched using either dedicated query languages or general-purpose string manipulation languages. Since both types of representation are equivalent, the bracketed structures can be translated by graphic interface programs into tree diagrams if desired, or they can be formatted as in (3) to facilitate processing by people.
| (3) | a. |
[IP [NPi he ]
[I' [I will ]
[VP [NPi t ]
[V' [V wait ] ] ] ] ]
|
| b. |
[IP [NPj we ]
[I' [I will ]
[VP [NPj t ]
[V' [V ask ]
[CP [C' [C if ]
[IP [NPi she ]
[I' [I [past] ]
[VP [NPi t ]
[V' [V left ] ] ] ] ] ] ] ] ] ] ]
| |
| c. |
[IP [NPi we ]
[I' [I [pres] ]
[VP [NPi t ]
[V' [V' [V drink ]
[NP green tea ]
[PP after lunch ] ] ] ] ] ]
|