|
|
@ -210,6 +210,7 @@ bool Correction::initProcessState(const int outputIndex) {
|
|
|
|
|
|
|
|
|
|
|
|
mMatching = false;
|
|
|
|
mMatching = false;
|
|
|
|
mProximityMatching = false;
|
|
|
|
mProximityMatching = false;
|
|
|
|
|
|
|
|
mAdditionalProximityMatching = false;
|
|
|
|
mTransposing = false;
|
|
|
|
mTransposing = false;
|
|
|
|
mExceeding = false;
|
|
|
|
mExceeding = false;
|
|
|
|
mSkipping = false;
|
|
|
|
mSkipping = false;
|
|
|
@ -256,6 +257,7 @@ void Correction::incrementOutputIndex() {
|
|
|
|
|
|
|
|
|
|
|
|
mCorrectionStates[mOutputIndex].mMatching = mMatching;
|
|
|
|
mCorrectionStates[mOutputIndex].mMatching = mMatching;
|
|
|
|
mCorrectionStates[mOutputIndex].mProximityMatching = mProximityMatching;
|
|
|
|
mCorrectionStates[mOutputIndex].mProximityMatching = mProximityMatching;
|
|
|
|
|
|
|
|
mCorrectionStates[mOutputIndex].mAdditionalProximityMatching = mAdditionalProximityMatching;
|
|
|
|
mCorrectionStates[mOutputIndex].mTransposing = mTransposing;
|
|
|
|
mCorrectionStates[mOutputIndex].mTransposing = mTransposing;
|
|
|
|
mCorrectionStates[mOutputIndex].mExceeding = mExceeding;
|
|
|
|
mCorrectionStates[mOutputIndex].mExceeding = mExceeding;
|
|
|
|
mCorrectionStates[mOutputIndex].mSkipping = mSkipping;
|
|
|
|
mCorrectionStates[mOutputIndex].mSkipping = mSkipping;
|
|
|
@ -304,6 +306,11 @@ inline bool isEquivalentChar(ProximityInfo::ProximityType type) {
|
|
|
|
return type == ProximityInfo::EQUIVALENT_CHAR;
|
|
|
|
return type == ProximityInfo::EQUIVALENT_CHAR;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
inline bool isProximityCharOrEquivalentChar(ProximityInfo::ProximityType type) {
|
|
|
|
|
|
|
|
return type == ProximityInfo::EQUIVALENT_CHAR
|
|
|
|
|
|
|
|
|| type == ProximityInfo::NEAR_PROXIMITY_CHAR;
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
Correction::CorrectionType Correction::processCharAndCalcState(
|
|
|
|
Correction::CorrectionType Correction::processCharAndCalcState(
|
|
|
|
const int32_t c, const bool isTerminal) {
|
|
|
|
const int32_t c, const bool isTerminal) {
|
|
|
|
const int correctionCount = (mSkippedCount + mExcessiveCount + mTransposedCount);
|
|
|
|
const int correctionCount = (mSkippedCount + mExcessiveCount + mTransposedCount);
|
|
|
@ -438,6 +445,9 @@ Correction::CorrectionType Correction::processCharAndCalcState(
|
|
|
|
|
|
|
|
|
|
|
|
if (ProximityInfo::UNRELATED_CHAR == matchedProximityCharId
|
|
|
|
if (ProximityInfo::UNRELATED_CHAR == matchedProximityCharId
|
|
|
|
|| ProximityInfo::ADDITIONAL_PROXIMITY_CHAR == matchedProximityCharId) {
|
|
|
|
|| ProximityInfo::ADDITIONAL_PROXIMITY_CHAR == matchedProximityCharId) {
|
|
|
|
|
|
|
|
if (ProximityInfo::ADDITIONAL_PROXIMITY_CHAR == matchedProximityCharId) {
|
|
|
|
|
|
|
|
mAdditionalProximityMatching = true;
|
|
|
|
|
|
|
|
}
|
|
|
|
// TODO: Optimize
|
|
|
|
// TODO: Optimize
|
|
|
|
// As the current char turned out to be an unrelated char,
|
|
|
|
// As the current char turned out to be an unrelated char,
|
|
|
|
// we will try other correction-types. Please note that mCorrectionStates[mOutputIndex]
|
|
|
|
// we will try other correction-types. Please note that mCorrectionStates[mOutputIndex]
|
|
|
@ -479,6 +489,18 @@ Correction::CorrectionType Correction::processCharAndCalcState(
|
|
|
|
++mSkippedCount;
|
|
|
|
++mSkippedCount;
|
|
|
|
--mProximityCount;
|
|
|
|
--mProximityCount;
|
|
|
|
return processSkipChar(c, isTerminal, false);
|
|
|
|
return processSkipChar(c, isTerminal, false);
|
|
|
|
|
|
|
|
} else if (mInputIndex - 1 < mInputLength
|
|
|
|
|
|
|
|
&& mSkippedCount > 0
|
|
|
|
|
|
|
|
&& mCorrectionStates[mOutputIndex].mSkipping
|
|
|
|
|
|
|
|
&& mCorrectionStates[mOutputIndex].mAdditionalProximityMatching
|
|
|
|
|
|
|
|
&& isProximityCharOrEquivalentChar(
|
|
|
|
|
|
|
|
mProximityInfo->getMatchedProximityId(mInputIndex + 1, c, false))) {
|
|
|
|
|
|
|
|
// Conversion s->a
|
|
|
|
|
|
|
|
incrementInputIndex();
|
|
|
|
|
|
|
|
--mSkippedCount;
|
|
|
|
|
|
|
|
mProximityMatching = true;
|
|
|
|
|
|
|
|
++mProximityCount;
|
|
|
|
|
|
|
|
mDistances[mOutputIndex] = ADDITIONAL_PROXIMITY_CHAR_DISTANCE_INFO;
|
|
|
|
} else if ((mExceeding || mTransposing) && mInputIndex - 1 < mInputLength
|
|
|
|
} else if ((mExceeding || mTransposing) && mInputIndex - 1 < mInputLength
|
|
|
|
&& isEquivalentChar(
|
|
|
|
&& isEquivalentChar(
|
|
|
|
mProximityInfo->getMatchedProximityId(mInputIndex + 1, c, false))) {
|
|
|
|
mProximityInfo->getMatchedProximityId(mInputIndex + 1, c, false))) {
|
|
|
@ -666,6 +688,10 @@ int Correction::RankingAlgorithm::calculateFinalFreq(const int inputIndex, const
|
|
|
|
|
|
|
|
|
|
|
|
int finalFreq = freq;
|
|
|
|
int finalFreq = freq;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if (DEBUG_CORRECTION_FREQ
|
|
|
|
|
|
|
|
&& (INPUTLENGTH_FOR_DEBUG <= 0 || INPUTLENGTH_FOR_DEBUG == inputLength)) {
|
|
|
|
|
|
|
|
AKLOGI("FinalFreq0: %d", finalFreq);
|
|
|
|
|
|
|
|
}
|
|
|
|
// TODO: Optimize this.
|
|
|
|
// TODO: Optimize this.
|
|
|
|
if (transposedCount > 0 || proximityMatchedCount > 0 || skipped || excessiveCount > 0) {
|
|
|
|
if (transposedCount > 0 || proximityMatchedCount > 0 || skipped || excessiveCount > 0) {
|
|
|
|
ed = getCurrentEditDistance(editDistanceTable, correction->mInputLength, outputLength,
|
|
|
|
ed = getCurrentEditDistance(editDistanceTable, correction->mInputLength, outputLength,
|
|
|
@ -681,12 +707,15 @@ int Correction::RankingAlgorithm::calculateFinalFreq(const int inputIndex, const
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
ed = max(0, ed - quoteDiffCount);
|
|
|
|
ed = max(0, ed - quoteDiffCount);
|
|
|
|
|
|
|
|
adjustedProximityMatchedCount = min(max(0, ed - (outputLength - inputLength)),
|
|
|
|
|
|
|
|
proximityMatchedCount);
|
|
|
|
if (transposedCount < 1) {
|
|
|
|
if (transposedCount < 1) {
|
|
|
|
if (ed == 1 && (inputLength == outputLength - 1 || inputLength == outputLength + 1)) {
|
|
|
|
if (ed == 1 && (inputLength == outputLength - 1 || inputLength == outputLength + 1)) {
|
|
|
|
// Promote a word with just one skipped or excessive char
|
|
|
|
// Promote a word with just one skipped or excessive char
|
|
|
|
if (sameLength) {
|
|
|
|
if (sameLength) {
|
|
|
|
multiplyRate(WORDS_WITH_JUST_ONE_CORRECTION_PROMOTION_RATE, &finalFreq);
|
|
|
|
multiplyRate(WORDS_WITH_JUST_ONE_CORRECTION_PROMOTION_RATE
|
|
|
|
|
|
|
|
+ WORDS_WITH_JUST_ONE_CORRECTION_PROMOTION_MULTIPLIER * outputLength,
|
|
|
|
|
|
|
|
&finalFreq);
|
|
|
|
} else {
|
|
|
|
} else {
|
|
|
|
multiplyIntCapped(typedLetterMultiplier, &finalFreq);
|
|
|
|
multiplyIntCapped(typedLetterMultiplier, &finalFreq);
|
|
|
|
}
|
|
|
|
}
|
|
|
@ -695,8 +724,6 @@ int Correction::RankingAlgorithm::calculateFinalFreq(const int inputIndex, const
|
|
|
|
sameLength = true;
|
|
|
|
sameLength = true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
adjustedProximityMatchedCount = min(max(0, ed - (outputLength - inputLength)),
|
|
|
|
|
|
|
|
proximityMatchedCount);
|
|
|
|
|
|
|
|
} else {
|
|
|
|
} else {
|
|
|
|
const int matchWeight = powerIntCapped(typedLetterMultiplier, matchCount);
|
|
|
|
const int matchWeight = powerIntCapped(typedLetterMultiplier, matchCount);
|
|
|
|
multiplyIntCapped(matchWeight, &finalFreq);
|
|
|
|
multiplyIntCapped(matchWeight, &finalFreq);
|
|
|
@ -744,6 +771,7 @@ int Correction::RankingAlgorithm::calculateFinalFreq(const int inputIndex, const
|
|
|
|
&& skippedCount == 0 && excessiveCount == 0 && transposedCount == 0;
|
|
|
|
&& skippedCount == 0 && excessiveCount == 0 && transposedCount == 0;
|
|
|
|
// Score calibration by touch coordinates is being done only for pure-fat finger typing error
|
|
|
|
// Score calibration by touch coordinates is being done only for pure-fat finger typing error
|
|
|
|
// cases.
|
|
|
|
// cases.
|
|
|
|
|
|
|
|
int additionalProximityCount = 0;
|
|
|
|
// TODO: Remove this constraint.
|
|
|
|
// TODO: Remove this constraint.
|
|
|
|
if (performTouchPositionCorrection) {
|
|
|
|
if (performTouchPositionCorrection) {
|
|
|
|
for (int i = 0; i < outputLength; ++i) {
|
|
|
|
for (int i = 0; i < outputLength; ++i) {
|
|
|
@ -776,12 +804,12 @@ int Correction::RankingAlgorithm::calculateFinalFreq(const int inputIndex, const
|
|
|
|
} else if (squaredDistance == PROXIMITY_CHAR_WITHOUT_DISTANCE_INFO) {
|
|
|
|
} else if (squaredDistance == PROXIMITY_CHAR_WITHOUT_DISTANCE_INFO) {
|
|
|
|
multiplyRate(WORDS_WITH_PROXIMITY_CHARACTER_DEMOTION_RATE, &finalFreq);
|
|
|
|
multiplyRate(WORDS_WITH_PROXIMITY_CHARACTER_DEMOTION_RATE, &finalFreq);
|
|
|
|
} else if (squaredDistance == ADDITIONAL_PROXIMITY_CHAR_DISTANCE_INFO) {
|
|
|
|
} else if (squaredDistance == ADDITIONAL_PROXIMITY_CHAR_DISTANCE_INFO) {
|
|
|
|
|
|
|
|
++additionalProximityCount;
|
|
|
|
multiplyRate(WORDS_WITH_ADDITIONAL_PROXIMITY_CHARACTER_DEMOTION_RATE, &finalFreq);
|
|
|
|
multiplyRate(WORDS_WITH_ADDITIONAL_PROXIMITY_CHARACTER_DEMOTION_RATE, &finalFreq);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
} else {
|
|
|
|
// Demote additional proximity characters
|
|
|
|
// Demote additional proximity characters
|
|
|
|
int additionalProximityCount = 0;
|
|
|
|
|
|
|
|
for (int i = 0; i < outputLength; ++i) {
|
|
|
|
for (int i = 0; i < outputLength; ++i) {
|
|
|
|
const int squaredDistance = correction->mDistances[i];
|
|
|
|
const int squaredDistance = correction->mDistances[i];
|
|
|
|
if (squaredDistance == ADDITIONAL_PROXIMITY_CHAR_DISTANCE_INFO) {
|
|
|
|
if (squaredDistance == ADDITIONAL_PROXIMITY_CHAR_DISTANCE_INFO) {
|
|
|
@ -803,6 +831,13 @@ int Correction::RankingAlgorithm::calculateFinalFreq(const int inputIndex, const
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
// If the user types too many(three or more) proximity characters with additional proximity
|
|
|
|
|
|
|
|
// character,do not treat as the same length word.
|
|
|
|
|
|
|
|
if (sameLength && additionalProximityCount > 0 && (adjustedProximityMatchedCount >= 3
|
|
|
|
|
|
|
|
|| transposedCount > 0 || skipped || excessiveCount > 0)) {
|
|
|
|
|
|
|
|
sameLength = false;
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
const int errorCount = adjustedProximityMatchedCount > 0
|
|
|
|
const int errorCount = adjustedProximityMatchedCount > 0
|
|
|
|
? adjustedProximityMatchedCount
|
|
|
|
? adjustedProximityMatchedCount
|
|
|
|
: (proximityMatchedCount + transposedCount);
|
|
|
|
: (proximityMatchedCount + transposedCount);
|
|
|
@ -813,13 +848,14 @@ int Correction::RankingAlgorithm::calculateFinalFreq(const int inputIndex, const
|
|
|
|
if (ed == 0) {
|
|
|
|
if (ed == 0) {
|
|
|
|
// Full exact match
|
|
|
|
// Full exact match
|
|
|
|
if (sameLength && transposedCount == 0 && !skipped && excessiveCount == 0
|
|
|
|
if (sameLength && transposedCount == 0 && !skipped && excessiveCount == 0
|
|
|
|
&& quoteDiffCount == 0) {
|
|
|
|
&& quoteDiffCount == 0 && additionalProximityCount == 0) {
|
|
|
|
finalFreq = capped255MultForFullMatchAccentsOrCapitalizationDifference(finalFreq);
|
|
|
|
finalFreq = capped255MultForFullMatchAccentsOrCapitalizationDifference(finalFreq);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
// Promote a word with no correction
|
|
|
|
// Promote a word with no correction
|
|
|
|
if (proximityMatchedCount == 0 && transposedCount == 0 && !skipped && excessiveCount == 0) {
|
|
|
|
if (proximityMatchedCount == 0 && transposedCount == 0 && !skipped && excessiveCount == 0
|
|
|
|
|
|
|
|
&& additionalProximityCount == 0) {
|
|
|
|
multiplyRate(FULL_MATCHED_WORDS_PROMOTION_RATE, &finalFreq);
|
|
|
|
multiplyRate(FULL_MATCHED_WORDS_PROMOTION_RATE, &finalFreq);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
@ -863,10 +899,11 @@ int Correction::RankingAlgorithm::calculateFinalFreq(const int inputIndex, const
|
|
|
|
|
|
|
|
|
|
|
|
if (DEBUG_CORRECTION_FREQ
|
|
|
|
if (DEBUG_CORRECTION_FREQ
|
|
|
|
&& (INPUTLENGTH_FOR_DEBUG <= 0 || INPUTLENGTH_FOR_DEBUG == inputLength)) {
|
|
|
|
&& (INPUTLENGTH_FOR_DEBUG <= 0 || INPUTLENGTH_FOR_DEBUG == inputLength)) {
|
|
|
|
|
|
|
|
DUMP_WORD(proximityInfo->getPrimaryInputWord(), inputLength);
|
|
|
|
DUMP_WORD(correction->mWord, outputLength);
|
|
|
|
DUMP_WORD(correction->mWord, outputLength);
|
|
|
|
AKLOGI("FinalFreq: [P%d, S%d, T%d, E%d] %d, %d, %d, %d, %d, %d", proximityMatchedCount,
|
|
|
|
AKLOGI("FinalFreq: [P%d, S%d, T%d, E%d, A%d] %d, %d, %d, %d, %d, %d", proximityMatchedCount,
|
|
|
|
skippedCount, transposedCount, excessiveCount, outputLength, lastCharExceeded,
|
|
|
|
skippedCount, transposedCount, excessiveCount, additionalProximityCount,
|
|
|
|
sameLength, quoteDiffCount, ed, finalFreq);
|
|
|
|
outputLength, lastCharExceeded, sameLength, quoteDiffCount, ed, finalFreq);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
return finalFreq;
|
|
|
|
return finalFreq;
|
|
|
|