edu.umass.cs.mallet.base.fst
Class Transducer.ViterbiPath
java.lang.Object
edu.umass.cs.mallet.base.types.SequencePair
edu.umass.cs.mallet.base.types.SequencePairAlignment
edu.umass.cs.mallet.base.fst.Transducer.ViterbiPath
- Enclosing class:
- Transducer
- public class Transducer.ViterbiPath
- extends SequencePairAlignment
Methods inherited from class java.lang.Object |
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait |
Transducer.ViterbiPath
protected Transducer.ViterbiPath(Sequence inputSequence,
Sequence outputSequence)
Transducer.ViterbiPath
protected Transducer.ViterbiPath(Sequence inputSequence,
Sequence outputSequence,
boolean saveLattice)
getDelta
public double getDelta(int ip,
int stateIndex)
getBestState
public Transducer.State getBestState(int ip)
getStateAtRank
public Transducer.State getStateAtRank(int ip,
int rank)
getViterbiNode
protected edu.umass.cs.mallet.base.fst.Transducer.ViterbiPath.ViterbiNode getViterbiNode(edu.umass.cs.mallet.base.fst.Transducer.ViterbiPath.ViterbiNode[][] nodes,
int ip,
int stateIndex)
incrementTransducerCounts
public void incrementTransducerCounts()
trimStateInfo
public SequencePairAlignment trimStateInfo()
tokenAccuracy
public double tokenAccuracy(Sequence referenceOutput)
tokenAccuracy
public double tokenAccuracy(Sequence referenceOutput,
java.io.PrintWriter out)
getTransducer
public Transducer getTransducer()
getBestStates
public Sequence getBestStates()
- Returns the maximum-value state sequence from this Lattice.
This is distinct from
SequencePair.output()
, which returns the Transducer's output from the
most likely state sequence.
Almost all applications will want to use the output method instead.
An exception are debugging tools that care specifically about the state sequence.
- Returns:
- A Sequence of Strings. Its length will be the input length plus 1. (for start state).