The Invisible Skeleton

Language isn't just a string of words (linear); it's a structure of nested boxes (hierarchical). We call these boxes Constituents.

Visualizing Structural Ambiguity

One sentence, two different hierarchical trees.

Analysis

The sentence "I shot an elephant in my pajamas" is ambiguous. The confusion arises not from the words, but from where the Prepositional Phrase (PP) "in my pajamas" attaches.

Meaning 1: Absurdity

The PP attaches to the Noun Phrase "an elephant". The elephant is wearing the pajamas.

Key Concepts

Hierarchy: Words group into phrases.
Recursion: NPs can contain PPs, which contain NPs.
Head: The core word (e.g., Noun) defines the phrase.

Form vs. Function

In syntax, we distinguish between what a word is (Syntactic Category) and the role it plays (Grammatical Function).

Interactive Sentence Parser

The strict professor assigned homework yesterday.

Click a word above to analyze its properties.

Concept	Definition	Examples
Category (Form)	Inherent lexical class or phrase type.	Noun, Verb, NP, VP
Function (Role)	Relationship to other parts of the sentence.	Subject, Object, Adjunct

👆

Select a Word

Tap a word in the sentence to reveal its syntactic secrets.

Theory, Evolution & AI

From Chomsky's trees to Google's Transformers, understanding syntax is key to understanding intelligence.

Theoretical Foundations

Structuralism (Early 20th C.)

Language as a catalog of patterns. Focus on surface arrangement. "Beads on a string."

Generative Grammar (1950s)

Chomsky revolution. Syntax is innate and algorithmic. Introduction of "Deep Structure" and transformations.

X-Bar Theory (1970s)

Scientific unification. All phrases (NP, VP, AP) share the same internal "X-Bar" geometry.

Constituency vs. Dependency

Two ways to map the same sentence. AI uses both.

Constituency

Groups words into Phrases

[[The cat] [sat]]

Used in: Theoretical Linguistics

Dependency

Links words to Heads

Cat ← sat → on

Used in: NLP, Google Translate

Language Acquisition: Human vs. Statistical Model

Humans (Hierarchical Bias) learn syntax with less data than basic statistical models. This "Poverty of the Stimulus" suggests innate structure.

Modern NLP Focus

While theories debate structure, modern Large Language Models (LLMs) implicitly learn both dependency and constituent patterns.