org.omegat.core.statistics

Class FindMatches

java.lang.Object

org.omegat.core.statistics.FindMatches

public class FindMatches
extends java.lang.Object

Class to find matches by specified criteria. Since we can use stemmers to prepare tokens, we should use 3-pass comparison of similarity. Similarity will be calculated in 3 steps: 1. Split original segment into word-only tokens using stemmer (with stop words list), then compare tokens. 2. Split original segment into word-only tokens without stemmer, then compare tokens. 3. Split original segment into not-only-words tokens (including numbers and tags) without stemmer, then compare tokens. This class is not thread safe ! Must be used in the one thread only.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	FindMatches.StoppedException Process will throw this exception if it stopped.All callers must catch it and just skip.

Constructor Summary

Constructors

Constructor and Description
FindMatches(IProject project, int maxCount, boolean allowSeparateSegmentMatch, boolean searchExactlyTheSame)

Method Summary

Modifier and Type	Method and Description
protected void	addNearString(EntryKey key, java.lang.String source, java.lang.String translation, NearString.MATCH_SOURCE comesFrom, boolean fuzzy, int similarity, int similarityNoStem, int simAdjusted, byte[] similarityData, java.lang.String tmxName, java.lang.String creator, long creationDate, java.lang.String changer, long changedDate, java.util.List<TMXProp> tuProperties) Add near string into result list.
protected void	checkStopped(IStopped stop)
protected boolean	haveChanceToAdd(int simStem, int simNoStem, int simExactly) Check if entry have a chance to be added to result list.
protected void	processEntry(EntryKey key, java.lang.String source, java.lang.String translation, NearString.MATCH_SOURCE comesFrom, boolean fuzzy, int penalty, java.lang.String tmxName, java.lang.String creator, long creationDate, java.lang.String changer, long changedDate, java.util.List<TMXProp> props) Compare one entry with original entry.
java.util.List<NearString>	search(java.lang.String searchText, boolean requiresTranslation, boolean fillSimilarityData, IStopped stop)
Token[]	tokenizeAll(java.lang.String str)
Token[]	tokenizeNoStem(java.lang.String str)
Token[]	tokenizeStem(java.lang.String str)

Methods inherited from class java.lang.Object

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

Constructor Detail

FindMatches

public FindMatches(IProject project,
                   int maxCount,
                   boolean allowSeparateSegmentMatch,
                   boolean searchExactlyTheSame)

Parameters:

searchExactlyTheSame - allows to search similarities with the same text as source segment. This mode used only for separate sentence match in paragraph project, i.e. where source is just part of current source.

Method Detail

search

public java.util.List<NearString> search(java.lang.String searchText,
                                         boolean requiresTranslation,
                                         boolean fillSimilarityData,
                                         IStopped stop)
                                  throws FindMatches.StoppedException

Throws:

FindMatches.StoppedException

processEntry

protected void processEntry(EntryKey key,
                            java.lang.String source,
                            java.lang.String translation,
                            NearString.MATCH_SOURCE comesFrom,
                            boolean fuzzy,
                            int penalty,
                            java.lang.String tmxName,
                            java.lang.String creator,
                            long creationDate,
                            java.lang.String changer,
                            long changedDate,
                            java.util.List<TMXProp> props)

Compare one entry with original entry.

Parameters:

candEntry - entry to compare

haveChanceToAdd

protected boolean haveChanceToAdd(int simStem,
                                  int simNoStem,
                                  int simExactly)

Check if entry have a chance to be added to result list. If no, there is no sense to calculate other parameters.

Parameters:

simStem - similarity with stemming

simNoStem - similarity without stemming

simExactly - exactly similarity

Returns:

true if we have chance

addNearString

protected void addNearString(EntryKey key,
                             java.lang.String source,
                             java.lang.String translation,
                             NearString.MATCH_SOURCE comesFrom,
                             boolean fuzzy,
                             int similarity,
                             int similarityNoStem,
                             int simAdjusted,
                             byte[] similarityData,
                             java.lang.String tmxName,
                             java.lang.String creator,
                             long creationDate,
                             java.lang.String changer,
                             long changedDate,
                             java.util.List<TMXProp> tuProperties)

Add near string into result list. Near strings sorted by "similarity,simAdjusted"

tokenizeStem

public Token[] tokenizeStem(java.lang.String str)

tokenizeNoStem

public Token[] tokenizeNoStem(java.lang.String str)

tokenizeAll

public Token[] tokenizeAll(java.lang.String str)

checkStopped

protected void checkStopped(IStopped stop)
                     throws FindMatches.StoppedException

Throws:

FindMatches.StoppedException