Catalogue of Artificial Intelligence Techniques

   

Jump to: Top | Entry | References | Comments

View Maths as: Images | MathML

Dependency Grammar

Aliases: DG , Dependency Syntax

Keywords: Machine Translation, non-terminal symbol

Categories: Natural Language


Author(s): Nicolas Nicolov

Dependency grammar (DG) describes the syntactic structure of sentences in natural languages in terms of links (dependencies) between individual words rather than constituency trees. The concept of dependency apparently originated with the Arabs and was adopted into Latin traditional grammar in the Middle Ages. The key claim of DG is that every phrase has a most prominent element (the head) which determines its syntactic properties. Modifiers and complements of the head are called dependents. The fundamental relation in dependency grammar is between head and dependent. One word (usually the main verb) is the head of the whole sentence; every other word depends on some head, and may itself be the head of any number of dependents. The rules of grammar then specify what heads can take what dependents (for example, adjectives depend on nouns, not on verbs). Practical DGs distinguish various types of dependents (complement, adjunct, determiner, etc.). DG, unlike Phrase Structure Grammar, does not take the constituency of syntactic elements (how are words grouped into phrases and how these phrases are grouped into bigger phrases, etc.) to be a central notion but rather concentrates on the relations between ultimate syntactic units. DG is therefore concerned with meaningful links, i.e., semantics. The syntactic representations postulated by the grammar are dependency trees. The nodes in a dependency tree correspond to word forms and the links between nodes represent binary relations between them. Thus, in dependency trees there are no non-terminal categories. The linear order between words is not stated in dependency trees and all meaningful distinctions expressed by word order variation should be explicitly stated using appropriate nodes and labelled arcs connecting them. DG has proved particularly useful in treating syntactic phenomena in free word-order languages (like Latin, the Slavic languages, etc.). It has found use in Machine Translation systems because dependency structures are seen to be very close to semantic structures.


References:


Comments:

Add Comment

No comments.