Version 2 (modified by 11 years ago) (diff) | ,
---|
Action
An action (or action list) follows the template specification in the rule.
An action is basically a function taking arbitrary number of arguments named accordingly to the token indices, starting with 0.
Currently, there are following actions implemented:
MARK
This action is used for marking a token segment. It is used in two cases:- Adding a phrasal element: in this case, the last argument must be the name of the new phrase.
Example:MARK 0 2 4 <coord>
-- marks a coordination on tokens with indices 0, 2 and 4. - Marking a single element: in this case, this element is then assigned a dependency using the
DEP
action (see below).
- Adding a phrasal element: in this case, the last argument must be the name of the new phrase.
AGREE
This action is used for testing grammatical agreement. It takes 3 arguments, the first two are the token indices, the third describes the morphological tags to be tested -g
for the gender,n
for the number andc
for the case.
Example:AGREE 0 2 gnc
-- tests a case-number-gender agreement on tokens with indices 0 and 2.
DEP
This action marks a dependency on a result of theMARK
action. It takes a single argument describing the token index to depend on.
PROB
This action assigns a probability, or rather weight (as it does not need to meet the probability axioms), to the rule. It takes a single argument -- the weight which is any positive natural number.
HEAD
This action marks the head of a phrasal element created by theMARK
action. It takes a single argument -- the token index to be marked as the head.