Recognizers and parsers
Explain the difference between a recognizer and a parser. Describe a scenario where each might be approriate.
Moving from a CKY Recognizer to a CKY Parser
Write a modification of the CKY-Recognizer algorithm (as given on page 6 of Chapter 11) so that it would be possible to recover a tree.
Parsers as Language Models
How would the spell checker you wrote for Lab 5 need to change if it used a parser instead of an -gram model as its language model?
Converting to Chomsky Normal Form
What 3 changes need to be made to a CFG to convert it to Chomsky Normal Form?
1.
2.
3.
For each of the above, show how probabilities should be altered to convert a PCFG to CNF. Your explanation should include an example rule with a probability and the probability-labeled rules that result from the conversion.
Head Annotation
Apply head annotation to phrasal nodes in each of the trees in Figure 11.2 (page 2 of Chapter 11).
Probabilistic CCG Parsing
Why is it reasonable to ignore production rules and only ‘count’ probabilities of lexical rules in probabilistic CCG parsing?
Why does probabilistic CCG parsing use algorithms like A* instead of table-based dynamic programming (like P-CKY does)?
Evaluating Parses
What is the advantage of counting cross-brackets in addition to constituent precision, recall, and F-measure?