de.up.ling.irtg.automata

Interface EdgeEvaluator

public interface EdgeEvaluator

This class is used to predict the outside score of a state. This is used in the agenda based intersection algorithms which attempt to find the based tree based on an A* approach.

Author:

koller

Method Summary

Modifier and Type	Method and Description
double	evaluate(int outputState, IntersectionAutomaton auto) This method is called each time the intersection algorithm discovers a new state to put on the agenda.
default void	ruleAdded(Rule rule) This method is called each time a rule is added to the output automaton.

Method Detail

ruleAdded

default void ruleAdded(Rule rule)

This method is called each time a rule is added to the output automaton. Its standard implementation does nothing, but subclasses might implement some other behavior, e.g. to keep track of inside probabilities of states.

Parameters:

rule -

evaluate

double evaluate(int outputState,
                IntersectionAutomaton auto)

This method is called each time the intersection algorithm discovers a new state to put on the agenda. This is a state of the intersection automaton, and thus it represents a pair (p,q) of states from the original left and and right automata. The evaluate method is supposed to return a numeric "evaluation" of the new state, so it can be sorted into the right place in the agenda (which is a priority queue, sorted by descending evaluation values). You can use IntersectionAutomaton.getLeftState(int) and IntersectionAutomaton.getRightState(int) with the "auto" argument to obtain p and q.

Parameters:

outputState -

auto -

Returns: