edu.cmu.sphinx.research.parallel
Class SameTimeScoreCombiner
java.lang.Object
edu.cmu.sphinx.research.parallel.SameTimeScoreCombiner
- All Implemented Interfaces:
- ScoreCombiner
- public class SameTimeScoreCombiner
- extends java.lang.Object
- implements ScoreCombiner
Combines the various unit feature stream scores in a CombineToken.
This SameTimeScoreCombiner assumes that there can be more than one token
from each feature stream. It will take the highest scoring token
from each stream at each last combine time, combine their
scores, and choose the one with the highest combined score.
For example, if tokens T1 and T2 from feature F1 had time stamps
(i.e., the last combine time) t1 and t2, and tokens Ta and Tb from
feature F2 had time stampes t1 and t2, we must compare
combinedScore(T1,Ta) and combineScore(T2,Tb), and retain
the one for which the combined score is higher.
All scores are maintained internally in the LogMath logbase
Method Summary |
void |
combineScore(CombineToken combineToken)
Combines the scores from all the feature stream in the given
CombineToken. |
Methods inherited from class java.lang.Object |
equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait |
SameTimeScoreCombiner
public SameTimeScoreCombiner(int timeDifference)
- Constructs a SameTimeScoreCombiner.
- Parameters:
timeDifference
- the maximum difference in time between
tokens for them to be considered approximately
the 'same time'
combineScore
public void combineScore(CombineToken combineToken)
- Combines the scores from all the feature stream in the given
CombineToken. The feature stream(s) are represented in the form
of ParallelTokens in the given CombineToken.
- Specified by:
combineScore
in interface ScoreCombiner
- Parameters:
combineToken
- the CombineToken on which to combine the feature
stream scores