Syntax and Parsing
- The syntax of a sentence really does matter.
- John killed Bob. is different from Bob killed John.
- The syntax of a sentence is typically represented by a
tree. (You can do this with brackets).
- (S(NP(NP The syntax) (PP of (NP a sentence)))
(VP (VG is typically represented) (PP by (NP a tree)))).
- Generating that tree from the surface form is called parsing.
- Chart parsers are typically used.
- People parse left to right with little backtracking.
- Cascaded FSAs are also used.
- Language is generally considered to be context free.
- Evaluate by comparison to a gold standard (e.g. wall street journal
corpus in the penn tree bank).