Actually add shortcut-only entries.

Change-Id: I84bec8fb560cec2ad9cc857397a3f77a96b1d12d
main
Jean Chalard 2012-01-16 12:51:46 +09:00
parent d64b8c97fe
commit c599f2e9d6
3 changed files with 64 additions and 23 deletions

View File

@ -606,7 +606,9 @@ public class BinaryDictInputOutput {
} }
flags |= FLAG_HAS_BIGRAMS; flags |= FLAG_HAS_BIGRAMS;
} }
// TODO: fill in the FLAG_IS_SHORTCUT_ONLY if (group.mIsShortcutOnly) {
flags |= FLAG_IS_SHORTCUT_ONLY;
}
return flags; return flags;
} }

View File

@ -170,6 +170,24 @@ public class FusionDictionary implements Iterable<Word> {
return array; return array;
} }
/**
* Helper method to add all words in a list as 0-frequency entries
*
* These words are added when shortcuts targets or bigrams are not found in the dictionary
* yet. The same words may be added later with an actual frequency - this is handled by
* the private version of add().
*/
private void addNeutralWords(final ArrayList<WeightedString> words) {
if (null != words) {
for (WeightedString word : words) {
final CharGroup t = findWordInTree(mRoot, word.mWord);
if (null == t) {
add(getCodePoints(word.mWord), 0, null, null, false /* isShortcutOnly */);
}
}
}
}
/** /**
* Helper method to add a word as a string. * Helper method to add a word as a string.
* *
@ -186,22 +204,12 @@ public class FusionDictionary implements Iterable<Word> {
final ArrayList<WeightedString> shortcutTargets, final ArrayList<WeightedString> shortcutTargets,
final ArrayList<WeightedString> bigrams) { final ArrayList<WeightedString> bigrams) {
if (null != shortcutTargets) { if (null != shortcutTargets) {
for (WeightedString target : shortcutTargets) { addNeutralWords(shortcutTargets);
final CharGroup t = findWordInTree(mRoot, target.mWord);
if (null == t) {
add(getCodePoints(target.mWord), 0, null, null);
}
}
} }
if (null != bigrams) { if (null != bigrams) {
for (WeightedString bigram : bigrams) { addNeutralWords(bigrams);
final CharGroup t = findWordInTree(mRoot, bigram.mWord);
if (null == t) {
add(getCodePoints(bigram.mWord), 0, null, null);
} }
} add(getCodePoints(word), frequency, shortcutTargets, bigrams, false /* isShortcutOnly */);
}
add(getCodePoints(word), frequency, shortcutTargets, bigrams);
} }
/** /**
@ -222,6 +230,22 @@ public class FusionDictionary implements Iterable<Word> {
} }
} }
/**
* Helper method to add a shortcut that should not be a dictionary word.
*
* @param word the word to add.
* @param frequency the frequency of the word, in the range [0..255].
* @param shortcutTargets a list of shortcut targets. May not be null.
*/
public void addShortcutOnly(final String word, final int frequency,
final ArrayList<WeightedString> shortcutTargets) {
if (null == shortcutTargets) {
throw new RuntimeException("Can't add a shortcut without targets");
}
addNeutralWords(shortcutTargets);
add(getCodePoints(word), frequency, shortcutTargets, null, true /* isShortcutOnly */);
}
/** /**
* Add a word to this dictionary. * Add a word to this dictionary.
* *
@ -232,10 +256,12 @@ public class FusionDictionary implements Iterable<Word> {
* @param frequency the frequency of the word, in the range [0..255]. * @param frequency the frequency of the word, in the range [0..255].
* @param shortcutTargets an optional list of shortcut targets for this word (null if none). * @param shortcutTargets an optional list of shortcut targets for this word (null if none).
* @param bigrams an optional list of bigrams for this word (null if none). * @param bigrams an optional list of bigrams for this word (null if none).
* @param isShortcutOnly whether this should be a shortcut only.
*/ */
private void add(final int[] word, final int frequency, private void add(final int[] word, final int frequency,
final ArrayList<WeightedString> shortcutTargets, final ArrayList<WeightedString> shortcutTargets,
final ArrayList<WeightedString> bigrams) { final ArrayList<WeightedString> bigrams,
final boolean isShortcutOnly) {
assert(frequency >= 0 && frequency <= 255); assert(frequency >= 0 && frequency <= 255);
Node currentNode = mRoot; Node currentNode = mRoot;
int charIndex = 0; int charIndex = 0;
@ -260,7 +286,7 @@ public class FusionDictionary implements Iterable<Word> {
final int insertionIndex = findInsertionIndex(currentNode, word[charIndex]); final int insertionIndex = findInsertionIndex(currentNode, word[charIndex]);
final CharGroup newGroup = new CharGroup( final CharGroup newGroup = new CharGroup(
Arrays.copyOfRange(word, charIndex, word.length), Arrays.copyOfRange(word, charIndex, word.length),
shortcutTargets, bigrams, frequency, false /* isShortcutOnly */); shortcutTargets, bigrams, frequency, isShortcutOnly);
currentNode.mData.add(insertionIndex, newGroup); currentNode.mData.add(insertionIndex, newGroup);
checkStack(currentNode); checkStack(currentNode);
} else { } else {
@ -275,7 +301,7 @@ public class FusionDictionary implements Iterable<Word> {
} else { } else {
final CharGroup newNode = new CharGroup(currentGroup.mChars, final CharGroup newNode = new CharGroup(currentGroup.mChars,
shortcutTargets, bigrams, frequency, currentGroup.mChildren, shortcutTargets, bigrams, frequency, currentGroup.mChildren,
false /* isShortcutOnly */); isShortcutOnly);
currentNode.mData.set(nodeIndex, newNode); currentNode.mData.set(nodeIndex, newNode);
checkStack(currentNode); checkStack(currentNode);
} }
@ -284,8 +310,7 @@ public class FusionDictionary implements Iterable<Word> {
// We only have to create a new node and add it to the end of this. // We only have to create a new node and add it to the end of this.
final CharGroup newNode = new CharGroup( final CharGroup newNode = new CharGroup(
Arrays.copyOfRange(word, charIndex + differentCharIndex, word.length), Arrays.copyOfRange(word, charIndex + differentCharIndex, word.length),
shortcutTargets, bigrams, frequency, shortcutTargets, bigrams, frequency, isShortcutOnly);
false /* isShortcutOnly */);
currentGroup.mChildren = new Node(); currentGroup.mChildren = new Node();
currentGroup.mChildren.mData.add(newNode); currentGroup.mChildren.mData.add(newNode);
} }
@ -300,7 +325,8 @@ public class FusionDictionary implements Iterable<Word> {
} }
final CharGroup newGroup = new CharGroup(word, final CharGroup newGroup = new CharGroup(word,
currentGroup.mShortcutTargets, currentGroup.mBigrams, currentGroup.mShortcutTargets, currentGroup.mBigrams,
frequency, currentGroup.mChildren, false /* isShortcutOnly */); frequency, currentGroup.mChildren,
currentGroup.mIsShortcutOnly && isShortcutOnly);
currentNode.mData.set(nodeIndex, newGroup); currentNode.mData.set(nodeIndex, newGroup);
} }
} else { } else {
@ -318,16 +344,18 @@ public class FusionDictionary implements Iterable<Word> {
if (charIndex + differentCharIndex >= word.length) { if (charIndex + differentCharIndex >= word.length) {
newParent = new CharGroup( newParent = new CharGroup(
Arrays.copyOfRange(currentGroup.mChars, 0, differentCharIndex), Arrays.copyOfRange(currentGroup.mChars, 0, differentCharIndex),
shortcutTargets, bigrams, frequency, newChildren, shortcutTargets, bigrams, frequency, newChildren, isShortcutOnly);
false /* isShortcutOnly */);
} else { } else {
// isShortcutOnly makes no sense for non-terminal nodes. The following node
// is non-terminal (frequency 0 in FusionDictionary representation) so we
// pass false for isShortcutOnly
newParent = new CharGroup( newParent = new CharGroup(
Arrays.copyOfRange(currentGroup.mChars, 0, differentCharIndex), Arrays.copyOfRange(currentGroup.mChars, 0, differentCharIndex),
null, null, -1, newChildren, false /* isShortcutOnly */); null, null, -1, newChildren, false /* isShortcutOnly */);
final CharGroup newWord = new CharGroup( final CharGroup newWord = new CharGroup(
Arrays.copyOfRange(word, charIndex + differentCharIndex, Arrays.copyOfRange(word, charIndex + differentCharIndex,
word.length), shortcutTargets, bigrams, frequency, word.length), shortcutTargets, bigrams, frequency,
false /* isShortcutOnly */); isShortcutOnly);
final int addIndex = word[charIndex + differentCharIndex] final int addIndex = word[charIndex + differentCharIndex]
> currentGroup.mChars[differentCharIndex] ? 1 : 0; > currentGroup.mChars[differentCharIndex] ? 1 : 0;
newChildren.mData.add(addIndex, newWord); newChildren.mData.add(addIndex, newWord);

View File

@ -46,6 +46,8 @@ public class XmlDictInputOutput {
private static final String FREQUENCY_ATTR = "f"; private static final String FREQUENCY_ATTR = "f";
private static final String WORD_ATTR = "word"; private static final String WORD_ATTR = "word";
private static final int SHORTCUT_ONLY_DEFAULT_FREQ = 1;
/** /**
* SAX handler for a unigram XML file. * SAX handler for a unigram XML file.
*/ */
@ -232,6 +234,15 @@ public class XmlDictInputOutput {
new UnigramHandler(dict, shortcutHandler.getShortcutMap(), new UnigramHandler(dict, shortcutHandler.getShortcutMap(),
bigramHandler.getBigramMap()); bigramHandler.getBigramMap());
parser.parse(unigrams, unigramHandler); parser.parse(unigrams, unigramHandler);
final HashMap<String, ArrayList<WeightedString>> shortcutMap =
shortcutHandler.getShortcutMap();
for (final String shortcut : shortcutMap.keySet()) {
if (dict.hasWord(shortcut)) continue;
// TODO: list a frequency in the shortcut file and use it here, instead of
// a constant freq
dict.addShortcutOnly(shortcut, SHORTCUT_ONLY_DEFAULT_FREQ, shortcutMap.get(shortcut));
}
return dict; return dict;
} }