Dependency Parsing is a syntactic analysis task that extracts the grammatical structure of a sentence by identifying binary relationships (dependencies) between "head" words and "dependent" words — representing the sentence as a directed graph (tree) where edges have labels like "subject", "object", "modifier".

Structure

• Head: The governor of the relation (e.g., the main verb).
• Dependent: The modifier (e.g., the subject noun).
• Root: The central node of the sentence (usually the main verb).
• Example: "John hit the ball." (hit $ o$ John [nsubj], hit $ o$ ball [dobj], ball $ o$ the [det]).

Why It Matters

• Information Extraction: "Who did what to whom?" is directly answered by the (Subject, Verb, Object) edges.
• Free Word Order: Better for languages with free word order (Russian, Latin) than Constituency Parsing.
• Efficiency: Linear-time transition-based parsers are very fast.

Dependency Parsing is connecting specific words — defining grammar as a web of relationships between individual words rather than nested phrases.