Implement ver4 dictionary bigram removing methods.

Bug: 11073222
Change-Id: I0c42e283d1ff983dc2c51e91b167bc39cdfd3da8
main
Keisuke Kuroyanagi 2013-11-13 16:40:01 +09:00
parent fe05173198
commit ff4b1d1fd0
8 changed files with 122 additions and 7 deletions

View File

@ -91,6 +91,33 @@ bool Ver4BigramListPolicy::addNewEntry(const int terminalId, const int newTarget
return mBigramDictContent->copyBigramList(bigramListPos, writingPos);
}
bool Ver4BigramListPolicy::removeEntry(const int terminalId, const int targetTerminalId) {
const int bigramListPos = mBigramDictContent->getBigramListHeadPos(terminalId);
if (bigramListPos == NOT_A_DICT_POS) {
// Bigram list does't exist.
return false;
}
const int entryPosToUpdate = getEntryPosToUpdate(targetTerminalId, bigramListPos);
if (entryPosToUpdate == NOT_A_DICT_POS) {
// Bigram entry doesn't exist.
return false;
}
int readingPos = entryPosToUpdate;
bool hasNext = false;
int probability = NOT_A_PROBABILITY;
int originalTargetTerminalId = Ver4DictConstants::NOT_A_TERMINAL_ID;
mBigramDictContent->getBigramEntryAndAdvancePosition(&probability, &hasNext,
&originalTargetTerminalId, &readingPos);
if (targetTerminalId != originalTargetTerminalId) {
// Bigram entry doesn't exist.
return false;
}
int writingPos = entryPosToUpdate;
// Remove bigram entry by overwriting target terminal Id.
return mBigramDictContent->writeBigramEntryAndAdvancePosition(probability, hasNext,
Ver4DictConstants::NOT_A_TERMINAL_ID /* targetTerminalId */, &writingPos);
}
int Ver4BigramListPolicy::getEntryPosToUpdate(const int targetTerminalIdToFind,
const int bigramListPos) const {
bool hasNext = true;

View File

@ -42,6 +42,8 @@ class Ver4BigramListPolicy : public DictionaryBigramsStructurePolicy {
bool addNewEntry(const int terminalId, const int newTargetTerminalId, const int newProbability,
bool *const outAddedNewEntry);
bool removeEntry(const int terminalId, const int targetTerminalId);
private:
DISALLOW_IMPLICIT_CONSTRUCTORS(Ver4BigramListPolicy);

View File

@ -31,9 +31,12 @@ void BigramDictContent::getBigramEntryAndAdvancePosition(int *const outProbabili
if (outHasNext) {
*outHasNext = (bigramFlags & Ver4DictConstants::BIGRAM_HAS_NEXT_MASK) != 0;
}
const int targetTerminalId = bigramListBuffer->readUintAndAdvancePosition(
Ver4DictConstants::BIGRAM_TARGET_TERMINAL_ID_FIELD_SIZE, bigramEntryPos);
if (outTargetTerminalId) {
*outTargetTerminalId = bigramListBuffer->readUintAndAdvancePosition(
Ver4DictConstants::BIGRAM_TARGET_TERMINAL_ID_FIELD_SIZE, bigramEntryPos);
*outTargetTerminalId =
(targetTerminalId == Ver4DictConstants::INVALID_BIGRAM_TARGET_TERMINAL_ID) ?
Ver4DictConstants::NOT_A_TERMINAL_ID : targetTerminalId;
}
}
@ -45,7 +48,10 @@ bool BigramDictContent::writeBigramEntryAndAdvancePosition(const int probability
Ver4DictConstants::BIGRAM_FLAGS_FIELD_SIZE, entryWritingPos)) {
return false;
}
return bigramListBuffer->writeUintAndAdvancePosition(targetTerminalId,
const int targetTerminalIdToWrite =
(targetTerminalId == Ver4DictConstants::NOT_A_TERMINAL_ID) ?
Ver4DictConstants::INVALID_BIGRAM_TARGET_TERMINAL_ID : targetTerminalId;
return bigramListBuffer->writeUintAndAdvancePosition(targetTerminalIdToWrite,
Ver4DictConstants::BIGRAM_TARGET_TERMINAL_ID_FIELD_SIZE, entryWritingPos);
}

View File

@ -42,6 +42,10 @@ const int Ver4DictConstants::SHORTCUT_ADDRESS_TABLE_BLOCK_SIZE = 16;
const int Ver4DictConstants::SHORTCUT_ADDRESS_TABLE_DATA_SIZE = 4;
const int Ver4DictConstants::BIGRAM_TARGET_TERMINAL_ID_FIELD_SIZE = 3;
// Unsigned int max value of BIGRAM_TARGET_TERMINAL_ID_FIELD_SIZE-byte is used for representing
// invalid terminal ID in bigram lists.
const int Ver4DictConstants::INVALID_BIGRAM_TARGET_TERMINAL_ID =
(1 << (BIGRAM_TARGET_TERMINAL_ID_FIELD_SIZE * 8)) - 1;
const int Ver4DictConstants::BIGRAM_FLAGS_FIELD_SIZE = 1;
const int Ver4DictConstants::BIGRAM_PROBABILITY_MASK = 0x0F;
const int Ver4DictConstants::BIGRAM_HAS_NEXT_MASK = 0x80;

View File

@ -47,6 +47,7 @@ class Ver4DictConstants {
static const int BIGRAM_FLAGS_FIELD_SIZE;
static const int BIGRAM_TARGET_TERMINAL_ID_FIELD_SIZE;
static const int INVALID_BIGRAM_TARGET_TERMINAL_ID;
static const int BIGRAM_PROBABILITY_MASK;
static const int BIGRAM_HAS_NEXT_MASK;

View File

@ -192,8 +192,8 @@ bool Ver4PatriciaTrieNodeWriter::addNewBigramEntry(
bool Ver4PatriciaTrieNodeWriter::removeBigramEntry(
const PtNodeParams *const sourcePtNodeParams, const PtNodeParams *const targetPtNodeParam) {
// TODO: Implement.
return false;
return mBigramPolicy->removeEntry(sourcePtNodeParams->getTerminalId(),
targetPtNodeParam->getTerminalId());
}
}

View File

@ -188,8 +188,31 @@ bool Ver4PatriciaTriePolicy::addBigramWords(const int *const word0, const int le
bool Ver4PatriciaTriePolicy::removeBigramWords(const int *const word0, const int length0,
const int *const word1, const int length1) {
// TODO: Implement.
return false;
if (!mBuffers.get()->isUpdatable()) {
AKLOGI("Warning: addBigramWords() is called for non-updatable dictionary.");
return false;
}
if (mDictBuffer.getTailPosition() >= MIN_DICT_SIZE_TO_REFUSE_DYNAMIC_OPERATIONS) {
AKLOGE("The dictionary is too large to dynamically update. Dictionary size: %d",
mDictBuffer.getTailPosition());
return false;
}
const int word0Pos = getTerminalPtNodePositionOfWord(word0, length0,
false /* forceLowerCaseSearch */);
if (word0Pos == NOT_A_DICT_POS) {
return false;
}
const int word1Pos = getTerminalPtNodePositionOfWord(word1, length1,
false /* forceLowerCaseSearch */);
if (word1Pos == NOT_A_DICT_POS) {
return false;
}
if (mUpdatingHelper.removeBigramWords(word0Pos, word1Pos)) {
mBigramCount--;
return true;
} else {
return false;
}
}
void Ver4PatriciaTriePolicy::flush(const char *const filePath) {

View File

@ -246,4 +246,56 @@ public class Ver4BinaryDictionaryTests extends AndroidTestCase {
assertEquals(probability, binaryDictionary.getBigramProbability("abb", "aaa"));
assertEquals(probability, binaryDictionary.getBigramProbability("abb", "bcc"));
}
public void testRemoveBigramWords() {
final String dictVersion = Long.toString(System.currentTimeMillis());
final FusionDictionary dict = new FusionDictionary(new PtNodeArray(),
getDictionaryOptions(TEST_LOCALE, dictVersion));
final DictEncoder encoder = new Ver4DictEncoder(getContext().getCacheDir());
try {
encoder.writeDictionary(dict, FORMAT_OPTIONS);
} catch (IOException e) {
Log.e(TAG, "IOException while writing dictionary", e);
} catch (UnsupportedFormatException e) {
Log.e(TAG, "Unsupported format", e);
}
final File trieFile = getTrieFile(TEST_LOCALE, dictVersion);
final BinaryDictionary binaryDictionary = new BinaryDictionary(trieFile.getAbsolutePath(),
0 /* offset */, trieFile.length(), true /* useFullEditDistance */,
Locale.getDefault(), TEST_LOCALE, true /* isUpdatable */);
assertTrue(binaryDictionary.isValidDictionary());
final int unigramProbability = 100;
final int bigramProbability = 10;
binaryDictionary.addUnigramWord("aaa", unigramProbability);
binaryDictionary.addUnigramWord("abb", unigramProbability);
binaryDictionary.addUnigramWord("bcc", unigramProbability);
binaryDictionary.addBigramWords("aaa", "abb", bigramProbability);
binaryDictionary.addBigramWords("aaa", "bcc", bigramProbability);
binaryDictionary.addBigramWords("abb", "aaa", bigramProbability);
binaryDictionary.addBigramWords("abb", "bcc", bigramProbability);
assertEquals(true, binaryDictionary.isValidBigram("aaa", "abb"));
assertEquals(true, binaryDictionary.isValidBigram("aaa", "bcc"));
assertEquals(true, binaryDictionary.isValidBigram("abb", "aaa"));
assertEquals(true, binaryDictionary.isValidBigram("abb", "bcc"));
binaryDictionary.removeBigramWords("aaa", "abb");
assertEquals(false, binaryDictionary.isValidBigram("aaa", "abb"));
binaryDictionary.addBigramWords("aaa", "abb", bigramProbability);
assertEquals(true, binaryDictionary.isValidBigram("aaa", "abb"));
binaryDictionary.removeBigramWords("aaa", "bcc");
assertEquals(false, binaryDictionary.isValidBigram("aaa", "bcc"));
binaryDictionary.removeBigramWords("abb", "aaa");
assertEquals(false, binaryDictionary.isValidBigram("abb", "aaa"));
binaryDictionary.removeBigramWords("abb", "bcc");
assertEquals(false, binaryDictionary.isValidBigram("abb", "bcc"));
binaryDictionary.removeBigramWords("aaa", "abb");
// Test remove non-existing bigram operation.
binaryDictionary.removeBigramWords("aaa", "abb");
binaryDictionary.removeBigramWords("bcc", "aaa");
}
}