- Constituency parsing is what you learn in an introductory linguistics class
- Break the sentence into a hierarchy of phrases
- NP, VP, old-school Chomsky stuff
- Rule-based methods worked fine for this, though statistical methods exist
- Not super useful for most businesses, but active in research
- Dependency parsing involves directly modeling the interactions between words
- Identifies main noun and verb through dependency relationships
- Lends itself well to statistical methods
- Still useful for some NER and POS tasks
- Though language models have come to dominate most of these tasks in everyday applications