Week 4 - Chunking

Question 1

Chunking

Answer 1

Finding groups of words that have specific rules, instead of just single words.

e.g. “Dave gave Mary a book” - Verb phrase

Question 2

Noun Phrase (NP) or nominal (phrase)

Answer 2

a phrase (sequence of words) that has a noun (at its head) and performs the same grammatical function as that noun

Question 3

Verb Phrase (VP)

Answer 3

A syntactic unit composed of at least one verb and its dependents

Question 4

Context Free Grammars

Answer 4

Terminals
- Takes these to be words

Non-Terminals
- The constituents in a language, Like NP, VP or sentence

Rules
Equations that consist of a single non-terminal on the left and any number of terminals and non-terminals on the right
A -> a B c | d c

Question 5

Noun Phrases (NP) determiners

Answer 5

NPs can start with determiners
They can be:
- Simple lexical items
A car

• Simple possessives
  John’s car
• complex recursive versions of that
  John’s sister’s husband’s son’s car

Question 6

Noun Phrases (NP) nominals

Answer 6

Contains the head and any pre- or post- modifiers of the head

Pre-modifiers:
Quantifiers, cardinals, ordinals…
Three cars

Adjectives
large cars

Ordering pre-modifiers?
three large cars
large three cars

Post modifiers:
propositional phrases (e.g. from Seattle)
Non-finite clauses (e.g. arriving before noon)
Relative clauses (e.g. that server breakfast)

There are many many rules for this, it is very complicated

Question 7

Treebanks

Answer 7

corpora in which each sentence has been paired with the (presumably correct) syntax tree.

Instead of paying linguists to write a grammar, pay them to annotate real sentences with parse trees. Use that annotated data to learn the rules

Penn treebank is a widely used treebank

Question 8

Probabilistic CFGs

Answer 8

Probabilistic context free grammar
Each production (rule) has a probability

From a treebank, get the probabilities for all rules and their lexicalised variants