Constituency Parsing is a syntactic analysis task that breaks a sentence into a hierarchy of nested phrases (constituents) — representing structure as a tree where leaves are words and internal nodes are phrase types (NP: Noun Phrase, VP: Verb Phrase, PP: Prepositional Phrase).

Structure

• Hierarchy: Sentences are typically divided recursively: S $ o$ NP VP.
• Non-terminal nodes: Abstract categories (NP, VP, S).
• Terminal nodes: The actual words.
• Example: "The black cat" $ o$ [NP [Det The] [Adj black] [N cat]].

Why It Matters

• Linguistics: Aligns with Noam Chomsky's Transformational Grammar / Phrase Structure Grammar.
• Scope: Useful for resolving scope ambiguity ("old men and women" — is "old" modifying just "men" or both?).
• Recursive Neural Networks: Tree-structured networks (Tree-LSTMs) run over constituency trees.

Constituency Parsing is nested shelving — organizing words into small phrases, which fit into larger phrases, forming a complete sentence structure.