Phrase - Representing Phrases

Representing Phrases

Many theories of syntax and grammar represent sentence structure using trees. The trees provide schematic illustrations of how the words of sentences are grouped. These representations show the words, phrases, and at times clauses that make up sentences. Any word combination that corresponds to a complete subtree can be seen as a phrase. There are two competing principles for producing trees, constituency and dependency. Both of these principles are illustrated here using the example sentence from above. The constituency-based tree is on the left, and the dependency-based tree on the right:

The constituency-based tree on the left is associated with a traditional phrase structure grammar, and the tree on the right is one of a dependency grammar. The node labels in the trees (e.g. N, NP, V, VP) mark the syntactic category of the constituents. Both trees take a phrase to be any combination of words that corresponds to a complete subtree. In the constituency tree on the left, each phrasal node (marked with P) identifies a phrase; there are therefore 8 phrases in the constituency tree. In the dependency tree on the right, each node that dominates one or more other nodes corresponds to a phrase; there are therefore 5 (or 6 if the whole sentence is included) phrases in the dependency tree. What the trees and the numbers demonstrate is that theories of syntax differ in what they deem to qualify as a phrase. The constituency tree takes three word combinations to be phrases (house at the end of the street, end of the street, and is red) that the dependency tree does not judge to be phrases. Which of the two tree structures is more plausible can be determined in part by empirical considerations, such as those delivered by constituency tests.