syntactic parsing

Syntactic parsing is working out the grammatical skeleton of a sentence — which words group into phrases and how those phrases hang together. It is the machine version of sentence diagramming. Given "The old man the boats," a parser has to discover the surprising structure: "the old" are the people, and "man" is the verb (to staff). The output is usually a tree showing the subject, the verb, the objects, and how each modifier attaches to what it describes.

Two flavors dominate. A constituency parse groups words into nested phrases (a noun phrase inside a sentence, an adjective inside a noun phrase) like Russian dolls. A dependency parse instead draws arrows from each word to the word it depends on — the verb at the center, with its subject and object hanging off it. Either way, the goal is to make the hidden architecture of a sentence explicit, because the same words in the same order can encode very different structures.

Why care? Because structure changes meaning. "I saw the man with the telescope" can mean you used the telescope, or the man was holding one — and which reading you pick depends entirely on where one phrase attaches. Parsing feeds translation, question-answering, and grammar-checking, and historically it was the proving ground of the whole field. The honest news: today's large language models often handle these structural problems implicitly, without ever producing an explicit tree — so visible parsing is now one useful tool among many, not the obligatory first step it once was.