marytts.tools.analysis

Class TranscriptionAligner

java.lang.Object

marytts.tools.analysis.TranscriptionAligner

Direct Known Subclasses:

MaryTranscriptionAligner

public class TranscriptionAligner
extends Object

This class aligns a label file with an XML file in MARY ALLOPHONES format, modifying the structure of the XML file as needed to match the label file. After calling alignXMLTranscriptions(), it is guaranteed that an iteration through all PHONE and BOUNDARY nodes of the XML file matches the label file.

Author:

marc

Field Summary

Fields

Modifier and Type	Field and Description
protected Map<String,Integer>	aligncost
protected AllophoneSet	allophoneSet
protected int	defaultBoundaryCost
protected int	defaultcost
protected boolean	ensureInitialBoundary
protected String	entrySeparator
protected String	possibleBnd
protected int	skipcost

Constructor Summary

Constructors

Constructor and Description
TranscriptionAligner()
TranscriptionAligner(AllophoneSet allophoneSet)
TranscriptionAligner(AllophoneSet allophoneSet, String entrySeparator)

Method Summary

Methods

Modifier and Type	Method and Description
AlignedLabels	alignLabels(Labels first, Labels second) Align the two given sequences of labels and return a mapping array indicating which index in first should be aligned to which index in second.
protected String	distanceAlign(String in, String out) This computes the alignment that has the lowest distance between two Strings.
boolean	getEnsureInitialBoundary()
String	getEntrySeparator()
static String	readLabelFile(String entrySeparator, boolean ensureInitialBoundary, String trfname) This reads in a label file and returns a String of the phonetic symbols, separated by the entry separator character entrySeparator.
void	SetEnsureInitialBoundary(boolean value)

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

aligncost

protected Map<String,Integer> aligncost

defaultcost

protected int defaultcost

defaultBoundaryCost

protected int defaultBoundaryCost

skipcost

protected int skipcost

allophoneSet

protected AllophoneSet allophoneSet

possibleBnd

protected String possibleBnd

entrySeparator

protected String entrySeparator

ensureInitialBoundary

protected boolean ensureInitialBoundary

Constructor Detail

TranscriptionAligner

public TranscriptionAligner()

TranscriptionAligner

public TranscriptionAligner(AllophoneSet allophoneSet)

TranscriptionAligner

public TranscriptionAligner(AllophoneSet allophoneSet,
                    String entrySeparator)

Method Detail

SetEnsureInitialBoundary

public void SetEnsureInitialBoundary(boolean value)

getEnsureInitialBoundary

public boolean getEnsureInitialBoundary()

getEntrySeparator

public String getEntrySeparator()

readLabelFile

public static String readLabelFile(String entrySeparator,
                   boolean ensureInitialBoundary,
                   String trfname)
                            throws IOException

This reads in a label file and returns a String of the phonetic symbols, separated by the entry separator character entrySeparator.

Parameters:

entrySeparator - entry separator

ensureInitialBoundary - ensure initial boundary

trfname - trf name

Returns:

result

Throws:

IOException - if something goes wrong with opening/reading the file

distanceAlign

protected String distanceAlign(String in,
                   String out)

This computes the alignment that has the lowest distance between two Strings. There are three differences to the normal Levenshtein-distance: 1. Only insertions and deletions are allowed, no replacements (i.e. no "diagonal" transitions) 2. insertion costs are dependent on a particular phone on the input side (the one they are aligned to) 3. deletion is equivalent to a symbol on the input side that is not aligned. There are costs associated with that. The method returns the output string with alignment boundaries ('#') inserted.

Parameters:

in - in

out - out

Returns:

p_al[ostr.length]

alignLabels

public AlignedLabels alignLabels(Labels first,
                        Labels second)

Align the two given sequences of labels and return a mapping array indicating which index in first should be aligned to which index in second.

Parameters:

first - first

second - second

Returns:

an array m of integers -- for each index i in first, m[i] gives the (rightmost) corresponding index in second.