mirror of
https://gitlab.futo.org/keyboard/latinime.git
synced 2024-09-28 14:54:30 +01:00
Merge changes I94a64865,I62725bfe
* changes: Refactor step3: const unigram_dictionary refactor native step 2
This commit is contained in:
Satoshi Kataoka
Android (Google) Code Review
commit
730b49fffc
@ -98,7 +98,7 @@ inline static int getCurrentEditDistance(int *editDistanceTable, const int editD
|
||||
static const char QUOTE = '\'';
|
||||
|
||||
inline bool Correction::isQuote(const unsigned short c) {
|
||||
const unsigned short userTypedChar = mProximityInfo->getPrimaryCharAt(mInputIndex);
|
||||
const unsigned short userTypedChar = mProximityInfoState.getPrimaryCharAt(mInputIndex);
|
||||
return (c == QUOTE && userTypedChar != QUOTE);
|
||||
}
|
||||
|
||||
@ -283,7 +283,7 @@ bool Correction::needsToPrune() const {
|
||||
|
||||
void Correction::addCharToCurrentWord(const int32_t c) {
|
||||
mWord[mOutputIndex] = c;
|
||||
const unsigned short *primaryInputWord = mProximityInfo->getPrimaryInputWord();
|
||||
const unsigned short *primaryInputWord = mProximityInfoState.getPrimaryInputWord();
|
||||
calcEditDistanceOneStep(mEditDistanceTable, primaryInputWord, mInputLength,
|
||||
mWord, mOutputIndex + 1);
|
||||
}
|
||||
@ -335,19 +335,19 @@ Correction::CorrectionType Correction::processCharAndCalcState(
|
||||
bool incremented = false;
|
||||
if (mLastCharExceeded && mInputIndex == mInputLength - 1) {
|
||||
// TODO: Do not check the proximity if EditDistance exceeds the threshold
|
||||
const ProximityType matchId =
|
||||
mProximityInfo->getMatchedProximityId(mInputIndex, c, true, &proximityIndex);
|
||||
const ProximityType matchId = mProximityInfoState.getMatchedProximityId(
|
||||
mInputIndex, c, true, &proximityIndex);
|
||||
if (isEquivalentChar(matchId)) {
|
||||
mLastCharExceeded = false;
|
||||
--mExcessiveCount;
|
||||
mDistances[mOutputIndex] =
|
||||
mProximityInfo->getNormalizedSquaredDistance(mInputIndex, 0);
|
||||
mProximityInfoState.getNormalizedSquaredDistance(mInputIndex, 0);
|
||||
} else if (matchId == NEAR_PROXIMITY_CHAR) {
|
||||
mLastCharExceeded = false;
|
||||
--mExcessiveCount;
|
||||
++mProximityCount;
|
||||
mDistances[mOutputIndex] =
|
||||
mProximityInfo->getNormalizedSquaredDistance(mInputIndex, proximityIndex);
|
||||
mDistances[mOutputIndex] = mProximityInfoState.getNormalizedSquaredDistance(
|
||||
mInputIndex, proximityIndex);
|
||||
}
|
||||
if (!isQuote(c)) {
|
||||
incrementInputIndex();
|
||||
@ -388,7 +388,8 @@ Correction::CorrectionType Correction::processCharAndCalcState(
|
||||
|
||||
bool secondTransposing = false;
|
||||
if (mTransposedCount % 2 == 1) {
|
||||
if (isEquivalentChar(mProximityInfo->getMatchedProximityId(mInputIndex - 1, c, false))) {
|
||||
if (isEquivalentChar(mProximityInfoState.getMatchedProximityId(
|
||||
mInputIndex - 1, c, false))) {
|
||||
++mTransposedCount;
|
||||
secondTransposing = true;
|
||||
} else if (mCorrectionStates[mOutputIndex].mExceeding) {
|
||||
@ -419,7 +420,7 @@ Correction::CorrectionType Correction::processCharAndCalcState(
|
||||
|
||||
ProximityType matchedProximityCharId = secondTransposing
|
||||
? EQUIVALENT_CHAR
|
||||
: mProximityInfo->getMatchedProximityId(
|
||||
: mProximityInfoState.getMatchedProximityId(
|
||||
mInputIndex, c, checkProximityChars, &proximityIndex);
|
||||
|
||||
if (UNRELATED_CHAR == matchedProximityCharId
|
||||
@ -427,7 +428,7 @@ Correction::CorrectionType Correction::processCharAndCalcState(
|
||||
if (canTryCorrection && mOutputIndex > 0
|
||||
&& mCorrectionStates[mOutputIndex].mProximityMatching
|
||||
&& mCorrectionStates[mOutputIndex].mExceeding
|
||||
&& isEquivalentChar(mProximityInfo->getMatchedProximityId(
|
||||
&& isEquivalentChar(mProximityInfoState.getMatchedProximityId(
|
||||
mInputIndex, mWord[mOutputIndex - 1], false))) {
|
||||
if (DEBUG_CORRECTION
|
||||
&& (INPUTLENGTH_FOR_DEBUG <= 0 || INPUTLENGTH_FOR_DEBUG == mInputLength)
|
||||
@ -446,7 +447,7 @@ Correction::CorrectionType Correction::processCharAndCalcState(
|
||||
// Here, we are doing something equivalent to matchedProximityCharId,
|
||||
// but we already know that "excessive char correction" just happened
|
||||
// so that we just need to check "mProximityCount == 0".
|
||||
matchedProximityCharId = mProximityInfo->getMatchedProximityId(
|
||||
matchedProximityCharId = mProximityInfoState.getMatchedProximityId(
|
||||
mInputIndex, c, mProximityCount == 0, &proximityIndex);
|
||||
}
|
||||
}
|
||||
@ -463,10 +464,10 @@ Correction::CorrectionType Correction::processCharAndCalcState(
|
||||
if (mInputIndex < mInputLength - 1 && mOutputIndex > 0 && mTransposedCount > 0
|
||||
&& !mCorrectionStates[mOutputIndex].mTransposing
|
||||
&& mCorrectionStates[mOutputIndex - 1].mTransposing
|
||||
&& isEquivalentChar(mProximityInfo->getMatchedProximityId(
|
||||
&& isEquivalentChar(mProximityInfoState.getMatchedProximityId(
|
||||
mInputIndex, mWord[mOutputIndex - 1], false))
|
||||
&& isEquivalentChar(
|
||||
mProximityInfo->getMatchedProximityId(mInputIndex + 1, c, false))) {
|
||||
mProximityInfoState.getMatchedProximityId(mInputIndex + 1, c, false))) {
|
||||
// Conversion t->e
|
||||
// Example:
|
||||
// occaisional -> occa sional
|
||||
@ -478,7 +479,7 @@ Correction::CorrectionType Correction::processCharAndCalcState(
|
||||
&& !mCorrectionStates[mOutputIndex].mTransposing
|
||||
&& mCorrectionStates[mOutputIndex - 1].mTransposing
|
||||
&& isEquivalentChar(
|
||||
mProximityInfo->getMatchedProximityId(mInputIndex - 1, c, false))) {
|
||||
mProximityInfoState.getMatchedProximityId(mInputIndex - 1, c, false))) {
|
||||
// Conversion t->s
|
||||
// Example:
|
||||
// chcolate -> chocolate
|
||||
@ -490,7 +491,7 @@ Correction::CorrectionType Correction::processCharAndCalcState(
|
||||
&& mCorrectionStates[mOutputIndex].mProximityMatching
|
||||
&& mCorrectionStates[mOutputIndex].mSkipping
|
||||
&& isEquivalentChar(
|
||||
mProximityInfo->getMatchedProximityId(mInputIndex - 1, c, false))) {
|
||||
mProximityInfoState.getMatchedProximityId(mInputIndex - 1, c, false))) {
|
||||
// Conversion p->s
|
||||
// Note: This logic tries saving cases like contrst --> contrast -- "a" is one of
|
||||
// proximity chars of "s", but it should rather be handled as a skipped char.
|
||||
@ -502,7 +503,7 @@ Correction::CorrectionType Correction::processCharAndCalcState(
|
||||
&& mCorrectionStates[mOutputIndex].mSkipping
|
||||
&& mCorrectionStates[mOutputIndex].mAdditionalProximityMatching
|
||||
&& isProximityCharOrEquivalentChar(
|
||||
mProximityInfo->getMatchedProximityId(mInputIndex + 1, c, false))) {
|
||||
mProximityInfoState.getMatchedProximityId(mInputIndex + 1, c, false))) {
|
||||
// Conversion s->a
|
||||
incrementInputIndex();
|
||||
--mSkippedCount;
|
||||
@ -511,7 +512,7 @@ Correction::CorrectionType Correction::processCharAndCalcState(
|
||||
mDistances[mOutputIndex] = ADDITIONAL_PROXIMITY_CHAR_DISTANCE_INFO;
|
||||
} else if ((mExceeding || mTransposing) && mInputIndex - 1 < mInputLength
|
||||
&& isEquivalentChar(
|
||||
mProximityInfo->getMatchedProximityId(mInputIndex + 1, c, false))) {
|
||||
mProximityInfoState.getMatchedProximityId(mInputIndex + 1, c, false))) {
|
||||
// 1.2. Excessive or transpose correction
|
||||
if (mTransposing) {
|
||||
++mTransposedCount;
|
||||
@ -573,12 +574,12 @@ Correction::CorrectionType Correction::processCharAndCalcState(
|
||||
} else if (isEquivalentChar(matchedProximityCharId)) {
|
||||
mMatching = true;
|
||||
++mEquivalentCharCount;
|
||||
mDistances[mOutputIndex] = mProximityInfo->getNormalizedSquaredDistance(mInputIndex, 0);
|
||||
mDistances[mOutputIndex] = mProximityInfoState.getNormalizedSquaredDistance(mInputIndex, 0);
|
||||
} else if (NEAR_PROXIMITY_CHAR == matchedProximityCharId) {
|
||||
mProximityMatching = true;
|
||||
++mProximityCount;
|
||||
mDistances[mOutputIndex] =
|
||||
mProximityInfo->getNormalizedSquaredDistance(mInputIndex, proximityIndex);
|
||||
mProximityInfoState.getNormalizedSquaredDistance(mInputIndex, proximityIndex);
|
||||
if (DEBUG_CORRECTION
|
||||
&& (INPUTLENGTH_FOR_DEBUG <= 0 || INPUTLENGTH_FOR_DEBUG == mInputLength)
|
||||
&& (MIN_OUTPUT_INDEX_FOR_DEBUG <= 0
|
||||
@ -662,7 +663,7 @@ int Correction::RankingAlgorithm::calculateFinalProbability(const int inputIndex
|
||||
const int excessivePos = correction->getExcessivePos();
|
||||
const int typedLetterMultiplier = correction->TYPED_LETTER_MULTIPLIER;
|
||||
const int fullWordMultiplier = correction->FULL_WORD_MULTIPLIER;
|
||||
const ProximityInfo *proximityInfo = correction->mProximityInfo;
|
||||
const ProximityInfoState *proximityInfoState = &correction->mProximityInfoState;
|
||||
const int skippedCount = correction->mSkippedCount;
|
||||
const int transposedCount = correction->mTransposedCount / 2;
|
||||
const int excessiveCount = correction->mExcessiveCount + correction->mTransposedCount % 2;
|
||||
@ -685,7 +686,7 @@ int Correction::RankingAlgorithm::calculateFinalProbability(const int inputIndex
|
||||
const bool skipped = skippedCount > 0;
|
||||
|
||||
const int quoteDiffCount = max(0, getQuoteCount(word, outputLength)
|
||||
- getQuoteCount(proximityInfo->getPrimaryInputWord(), inputLength));
|
||||
- getQuoteCount(proximityInfoState->getPrimaryInputWord(), inputLength));
|
||||
|
||||
// TODO: Calculate edit distance for transposed and excessive
|
||||
int ed = 0;
|
||||
@ -737,7 +738,7 @@ int Correction::RankingAlgorithm::calculateFinalProbability(const int inputIndex
|
||||
multiplyIntCapped(matchWeight, &finalFreq);
|
||||
}
|
||||
|
||||
if (proximityInfo->getMatchedProximityId(0, word[0], true) == UNRELATED_CHAR) {
|
||||
if (proximityInfoState->getMatchedProximityId(0, word[0], true) == UNRELATED_CHAR) {
|
||||
multiplyRate(FIRST_CHAR_DIFFERENT_DEMOTION_RATE, &finalFreq);
|
||||
}
|
||||
|
||||
@ -763,7 +764,7 @@ int Correction::RankingAlgorithm::calculateFinalProbability(const int inputIndex
|
||||
// Demotion for a word with excessive character
|
||||
if (excessiveCount > 0) {
|
||||
multiplyRate(WORDS_WITH_EXCESSIVE_CHARACTER_DEMOTION_RATE, &finalFreq);
|
||||
if (!lastCharExceeded && !proximityInfo->existsAdjacentProximityChars(excessivePos)) {
|
||||
if (!lastCharExceeded && !proximityInfoState->existsAdjacentProximityChars(excessivePos)) {
|
||||
if (DEBUG_DICT_FULL) {
|
||||
AKLOGI("Double excessive demotion");
|
||||
}
|
||||
@ -774,8 +775,9 @@ int Correction::RankingAlgorithm::calculateFinalProbability(const int inputIndex
|
||||
}
|
||||
|
||||
const bool performTouchPositionCorrection =
|
||||
CALIBRATE_SCORE_BY_TOUCH_COORDINATES && proximityInfo->touchPositionCorrectionEnabled()
|
||||
&& skippedCount == 0 && excessiveCount == 0 && transposedCount == 0;
|
||||
CALIBRATE_SCORE_BY_TOUCH_COORDINATES
|
||||
&& proximityInfoState->touchPositionCorrectionEnabled()
|
||||
&& skippedCount == 0 && excessiveCount == 0 && transposedCount == 0;
|
||||
// Score calibration by touch coordinates is being done only for pure-fat finger typing error
|
||||
// cases.
|
||||
int additionalProximityCount = 0;
|
||||
@ -1145,5 +1147,4 @@ float Correction::RankingAlgorithm::calcNormalizedScore(const unsigned short* be
|
||||
const float weight = 1.0 - (float) distance / afterLength;
|
||||
return (score / maxScore) * weight;
|
||||
}
|
||||
|
||||
} // namespace latinime
|
||||
|
||||
@ -19,9 +19,10 @@
|
||||
|
||||
#include <assert.h>
|
||||
#include <stdint.h>
|
||||
#include "correction_state.h"
|
||||
|
||||
#include "correction_state.h"
|
||||
#include "defines.h"
|
||||
#include "proximity_info_state.h"
|
||||
|
||||
namespace latinime {
|
||||
|
||||
@ -178,6 +179,21 @@ class Correction {
|
||||
static const int FULL_WORD_MULTIPLIER = 2;
|
||||
};
|
||||
|
||||
// proximity info state
|
||||
void initInputParams(const ProximityInfo *proximityInfo, const int32_t *inputCodes,
|
||||
const int inputLength, const int *xCoordinates, const int *yCoordinates) {
|
||||
mProximityInfoState.initInputParams(
|
||||
proximityInfo, inputCodes, inputLength, xCoordinates, yCoordinates);
|
||||
}
|
||||
|
||||
const unsigned short* getPrimaryInputWord() const {
|
||||
return mProximityInfoState.getPrimaryInputWord();
|
||||
}
|
||||
|
||||
unsigned short getPrimaryCharAt(const int index) const {
|
||||
return mProximityInfoState.getPrimaryCharAt(index);
|
||||
}
|
||||
|
||||
private:
|
||||
inline void incrementInputIndex();
|
||||
inline void incrementOutputIndex();
|
||||
@ -240,7 +256,7 @@ class Correction {
|
||||
bool mExceeding;
|
||||
bool mTransposing;
|
||||
bool mSkipping;
|
||||
|
||||
ProximityInfoState mProximityInfoState;
|
||||
};
|
||||
} // namespace latinime
|
||||
#endif // LATINIME_CORRECTION_H
|
||||
|
||||
@ -73,9 +73,6 @@ ProximityInfo::ProximityInfo(const std::string localeStr, const int maxProximity
|
||||
copyOrFillZero(mSweetSpotRadii, sweetSpotRadii, KEY_COUNT * sizeof(mSweetSpotRadii[0]));
|
||||
|
||||
initializeCodeToKeyIndex();
|
||||
mProximityInfoState = new ProximityInfoState(this, MAX_PROXIMITY_CHARS_SIZE,
|
||||
HAS_TOUCH_POSITION_CORRECTION_DATA, MOST_COMMON_KEY_WIDTH_SQUARE, mLocaleStr,
|
||||
KEY_COUNT, CELL_HEIGHT, CELL_WIDTH, GRID_WIDTH, GRID_HEIGHT);
|
||||
}
|
||||
|
||||
// Build the reversed look up table from the char code to the index in mKeyXCoordinates,
|
||||
@ -92,7 +89,6 @@ void ProximityInfo::initializeCodeToKeyIndex() {
|
||||
|
||||
ProximityInfo::~ProximityInfo() {
|
||||
delete[] mProximityCharsArray;
|
||||
delete mProximityInfoState;
|
||||
}
|
||||
|
||||
inline int ProximityInfo::getStartIndexFromCoordinates(const int x, const int y) const {
|
||||
@ -203,12 +199,6 @@ void ProximityInfo::calculateNearbyKeyCodes(
|
||||
}
|
||||
}
|
||||
|
||||
// TODO: remove
|
||||
void ProximityInfo::initInputParams(const int32_t *inputCodes, const int inputLength,
|
||||
const int *xCoordinates, const int *yCoordinates) {
|
||||
mProximityInfoState->initInputParams(inputCodes, inputLength, xCoordinates, yCoordinates);
|
||||
}
|
||||
|
||||
int ProximityInfo::getKeyIndex(const int c) const {
|
||||
if (KEY_COUNT == 0) {
|
||||
// We do not have the coordinate data
|
||||
@ -220,47 +210,4 @@ int ProximityInfo::getKeyIndex(const int c) const {
|
||||
}
|
||||
return mCodeToKeyIndex[baseLowerC];
|
||||
}
|
||||
|
||||
// TODO: remove
|
||||
inline const int* ProximityInfo::getProximityCharsAt(const int index) const {
|
||||
return mProximityInfoState->getProximityCharsAt(index);
|
||||
}
|
||||
|
||||
// TODO: remove
|
||||
unsigned short ProximityInfo::getPrimaryCharAt(const int index) const {
|
||||
return mProximityInfoState->getPrimaryCharAt(index);
|
||||
}
|
||||
|
||||
// TODO: remove
|
||||
bool ProximityInfo::existsCharInProximityAt(const int index, const int c) const {
|
||||
return mProximityInfoState->existsCharInProximityAt(index, c);
|
||||
}
|
||||
|
||||
// TODO: remove
|
||||
bool ProximityInfo::existsAdjacentProximityChars(const int index) const {
|
||||
return mProximityInfoState->existsAdjacentProximityChars(index);
|
||||
}
|
||||
|
||||
// TODO: remove
|
||||
ProximityType ProximityInfo::getMatchedProximityId(const int index,
|
||||
const unsigned short c, const bool checkProximityChars, int *proximityIndex) const {
|
||||
return mProximityInfoState->getMatchedProximityId(
|
||||
index, c, checkProximityChars, proximityIndex);
|
||||
}
|
||||
|
||||
// TODO: remove
|
||||
int ProximityInfo::getNormalizedSquaredDistance(
|
||||
const int inputIndex, const int proximityIndex) const {
|
||||
return mProximityInfoState->getNormalizedSquaredDistance(inputIndex, proximityIndex);
|
||||
}
|
||||
|
||||
// TODO: remove
|
||||
const unsigned short* ProximityInfo::getPrimaryInputWord() const {
|
||||
return mProximityInfoState->getPrimaryInputWord();
|
||||
}
|
||||
|
||||
// TODO: remove
|
||||
bool ProximityInfo::touchPositionCorrectionEnabled() const {
|
||||
return mProximityInfoState->touchPositionCorrectionEnabled();
|
||||
}
|
||||
} // namespace latinime
|
||||
|
||||
@ -25,11 +25,9 @@
|
||||
namespace latinime {
|
||||
|
||||
class Correction;
|
||||
class ProximityInfoState;
|
||||
|
||||
class ProximityInfo {
|
||||
public:
|
||||
|
||||
ProximityInfo(const std::string localeStr, const int maxProximityCharsSize,
|
||||
const int keyboardWidth, const int keyboardHeight, const int gridWidth,
|
||||
const int gridHeight, const int mostCommonkeyWidth,
|
||||
@ -68,21 +66,37 @@ class ProximityInfo {
|
||||
void calculateNearbyKeyCodes(
|
||||
const int x, const int y, const int32_t primaryKey, int *inputCodes) const;
|
||||
|
||||
////////////////////////////////////
|
||||
// Access to proximity info state //
|
||||
// TODO: remove //
|
||||
////////////////////////////////////
|
||||
void initInputParams(const int32_t *inputCodes, const int inputLength,
|
||||
const int *xCoordinates, const int *yCoordinates);
|
||||
const int* getProximityCharsAt(const int index) const;
|
||||
unsigned short getPrimaryCharAt(const int index) const;
|
||||
bool existsCharInProximityAt(const int index, const int c) const;
|
||||
bool existsAdjacentProximityChars(const int index) const;
|
||||
ProximityType getMatchedProximityId(const int index, const unsigned short c,
|
||||
const bool checkProximityChars, int *proximityIndex = 0) const;
|
||||
const unsigned short* getPrimaryInputWord() const;
|
||||
bool touchPositionCorrectionEnabled() const;
|
||||
////////////////////////////////////
|
||||
bool hasTouchPositionCorrectionData() const {
|
||||
return HAS_TOUCH_POSITION_CORRECTION_DATA;
|
||||
}
|
||||
|
||||
int getMostCommonKeyWidthSquare() const {
|
||||
return MOST_COMMON_KEY_WIDTH_SQUARE;
|
||||
}
|
||||
|
||||
std::string getLocaleStr() const {
|
||||
return mLocaleStr;
|
||||
}
|
||||
|
||||
int getKeyCount() const {
|
||||
return KEY_COUNT;
|
||||
}
|
||||
|
||||
int getCellHeight() const {
|
||||
return CELL_HEIGHT;
|
||||
}
|
||||
|
||||
int getCellWidth() const {
|
||||
return CELL_WIDTH;
|
||||
}
|
||||
|
||||
int getGridWidth() const {
|
||||
return GRID_WIDTH;
|
||||
}
|
||||
|
||||
int getGridHeight() const {
|
||||
return GRID_HEIGHT;
|
||||
}
|
||||
|
||||
private:
|
||||
// The max number of the keys in one keyboard layout
|
||||
@ -121,7 +135,6 @@ class ProximityInfo {
|
||||
float mSweetSpotRadii[MAX_KEY_COUNT_IN_A_KEYBOARD];
|
||||
int mCodeToKeyIndex[MAX_CHAR_CODE + 1];
|
||||
// TODO: move to correction.h
|
||||
ProximityInfoState *mProximityInfoState;
|
||||
};
|
||||
|
||||
} // namespace latinime
|
||||
|
||||
@ -27,25 +27,41 @@
|
||||
#include "proximity_info_state.h"
|
||||
|
||||
namespace latinime {
|
||||
void ProximityInfoState::initInputParams(const int32_t* inputCodes, const int inputLength,
|
||||
void ProximityInfoState::initInputParams(
|
||||
const ProximityInfo* proximityInfo, const int32_t* inputCodes, const int inputLength,
|
||||
const int* xCoordinates, const int* yCoordinates) {
|
||||
mProximityInfo = proximityInfo;
|
||||
mHasTouchPositionCorrectionData = proximityInfo->hasTouchPositionCorrectionData();
|
||||
mMostCommonKeyWidthSquare = proximityInfo->getMostCommonKeyWidthSquare();
|
||||
mLocaleStr = proximityInfo->getLocaleStr();
|
||||
mKeyCount = proximityInfo->getKeyCount();
|
||||
mCellHeight = proximityInfo->getCellHeight();
|
||||
mCellWidth = proximityInfo->getCellWidth();
|
||||
mGridHeight = proximityInfo->getGridWidth();
|
||||
mGridWidth = proximityInfo->getGridHeight();
|
||||
const int normalizedSquaredDistancesLength =
|
||||
MAX_PROXIMITY_CHARS_SIZE_INTERNAL * MAX_WORD_LENGTH_INTERNAL;
|
||||
for (int i = 0; i < normalizedSquaredDistancesLength; ++i) {
|
||||
mNormalizedSquaredDistances[i] = NOT_A_DISTANCE;
|
||||
}
|
||||
|
||||
memset(mInputCodes, 0,
|
||||
MAX_WORD_LENGTH_INTERNAL * MAX_PROXIMITY_CHARS_SIZE * sizeof(mInputCodes[0]));
|
||||
MAX_WORD_LENGTH_INTERNAL * MAX_PROXIMITY_CHARS_SIZE_INTERNAL * sizeof(mInputCodes[0]));
|
||||
|
||||
for (int i = 0; i < inputLength; ++i) {
|
||||
const int32_t primaryKey = inputCodes[i];
|
||||
const int x = xCoordinates[i];
|
||||
const int y = yCoordinates[i];
|
||||
int *proximities = &mInputCodes[i * MAX_PROXIMITY_CHARS_SIZE];
|
||||
int *proximities = &mInputCodes[i * MAX_PROXIMITY_CHARS_SIZE_INTERNAL];
|
||||
mProximityInfo->calculateNearbyKeyCodes(x, y, primaryKey, proximities);
|
||||
}
|
||||
|
||||
if (DEBUG_PROXIMITY_CHARS) {
|
||||
for (int i = 0; i < inputLength; ++i) {
|
||||
AKLOGI("---");
|
||||
for (int j = 0; j < MAX_PROXIMITY_CHARS_SIZE; ++j) {
|
||||
int icc = mInputCodes[i * MAX_PROXIMITY_CHARS_SIZE + j];
|
||||
int icfjc = inputCodes[i * MAX_PROXIMITY_CHARS_SIZE + j];
|
||||
for (int j = 0; j < MAX_PROXIMITY_CHARS_SIZE_INTERNAL; ++j) {
|
||||
int icc = mInputCodes[i * MAX_PROXIMITY_CHARS_SIZE_INTERNAL + j];
|
||||
int icfjc = inputCodes[i * MAX_PROXIMITY_CHARS_SIZE_INTERNAL + j];
|
||||
icc += 0;
|
||||
icfjc += 0;
|
||||
AKLOGI("--- (%d)%c,%c", i, icc, icfjc); AKLOGI("--- A<%d>,B<%d>", icc, icfjc);
|
||||
@ -54,8 +70,8 @@ void ProximityInfoState::initInputParams(const int32_t* inputCodes, const int in
|
||||
}
|
||||
mInputXCoordinates = xCoordinates;
|
||||
mInputYCoordinates = yCoordinates;
|
||||
mTouchPositionCorrectionEnabled = HAS_TOUCH_POSITION_CORRECTION_DATA && xCoordinates
|
||||
&& yCoordinates;
|
||||
mTouchPositionCorrectionEnabled =
|
||||
mHasTouchPositionCorrectionData && xCoordinates && yCoordinates;
|
||||
mInputLength = inputLength;
|
||||
for (int i = 0; i < inputLength; ++i) {
|
||||
mPrimaryInputWord[i] = getPrimaryCharAt(i);
|
||||
@ -74,17 +90,17 @@ void ProximityInfoState::initInputParams(const int32_t* inputCodes, const int in
|
||||
a += 0;
|
||||
AKLOGI("--- Primary = %c, x = %d, y = %d", primaryKey, x, y);
|
||||
}
|
||||
for (int j = 0; j < MAX_PROXIMITY_CHARS_SIZE && proximityChars[j] > 0; ++j) {
|
||||
for (int j = 0; j < MAX_PROXIMITY_CHARS_SIZE_INTERNAL && proximityChars[j] > 0; ++j) {
|
||||
const int currentChar = proximityChars[j];
|
||||
const float squaredDistance =
|
||||
hasInputCoordinates() ? calculateNormalizedSquaredDistance(
|
||||
mProximityInfo->getKeyIndex(currentChar), i) :
|
||||
NOT_A_DISTANCE_FLOAT;
|
||||
if (squaredDistance >= 0.0f) {
|
||||
mNormalizedSquaredDistances[i * MAX_PROXIMITY_CHARS_SIZE + j] =
|
||||
mNormalizedSquaredDistances[i * MAX_PROXIMITY_CHARS_SIZE_INTERNAL + j] =
|
||||
(int) (squaredDistance * NORMALIZED_SQUARED_DISTANCE_SCALING_FACTOR);
|
||||
} else {
|
||||
mNormalizedSquaredDistances[i * MAX_PROXIMITY_CHARS_SIZE + j] =
|
||||
mNormalizedSquaredDistances[i * MAX_PROXIMITY_CHARS_SIZE_INTERNAL + j] =
|
||||
(j == 0) ? EQUIVALENT_CHAR_WITHOUT_DISTANCE_INFO :
|
||||
PROXIMITY_CHAR_WITHOUT_DISTANCE_INFO;
|
||||
}
|
||||
@ -94,4 +110,30 @@ void ProximityInfoState::initInputParams(const int32_t* inputCodes, const int in
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
float ProximityInfoState::calculateNormalizedSquaredDistance(
|
||||
const int keyIndex, const int inputIndex) const {
|
||||
if (keyIndex == NOT_AN_INDEX) {
|
||||
return NOT_A_DISTANCE_FLOAT;
|
||||
}
|
||||
if (!mProximityInfo->hasSweetSpotData(keyIndex)) {
|
||||
return NOT_A_DISTANCE_FLOAT;
|
||||
}
|
||||
if (NOT_A_COORDINATE == mInputXCoordinates[inputIndex]) {
|
||||
return NOT_A_DISTANCE_FLOAT;
|
||||
}
|
||||
const float squaredDistance = calculateSquaredDistanceFromSweetSpotCenter(
|
||||
keyIndex, inputIndex);
|
||||
const float squaredRadius = square(mProximityInfo->getSweetSpotRadiiAt(keyIndex));
|
||||
return squaredDistance / squaredRadius;
|
||||
}
|
||||
|
||||
float ProximityInfoState::calculateSquaredDistanceFromSweetSpotCenter(
|
||||
const int keyIndex, const int inputIndex) const {
|
||||
const float sweetSpotCenterX = mProximityInfo->getSweetSpotCenterXAt(keyIndex);
|
||||
const float sweetSpotCenterY = mProximityInfo->getSweetSpotCenterYAt(keyIndex);
|
||||
const float inputX = (float)mInputXCoordinates[inputIndex];
|
||||
const float inputY = (float)mInputYCoordinates[inputIndex];
|
||||
return square(inputX - sweetSpotCenterX) + square(inputY - sweetSpotCenterY);
|
||||
}
|
||||
} // namespace latinime
|
||||
|
||||
@ -22,6 +22,7 @@
|
||||
#include <string>
|
||||
|
||||
#include "additional_proximity_chars.h"
|
||||
#include "char_utils.h"
|
||||
#include "defines.h"
|
||||
|
||||
namespace latinime {
|
||||
@ -33,8 +34,6 @@ class ProximityInfoState {
|
||||
static const int NORMALIZED_SQUARED_DISTANCE_SCALING_FACTOR_LOG_2 = 10;
|
||||
static const int NORMALIZED_SQUARED_DISTANCE_SCALING_FACTOR =
|
||||
1 << NORMALIZED_SQUARED_DISTANCE_SCALING_FACTOR_LOG_2;
|
||||
// The max number of the keys in one keyboard layout
|
||||
static const int MAX_KEY_COUNT_IN_A_KEYBOARD = 64;
|
||||
// The upper limit of the char code in mCodeToKeyIndex
|
||||
static const int MAX_CHAR_CODE = 127;
|
||||
static const float NOT_A_DISTANCE_FLOAT = -1.0f;
|
||||
@ -43,39 +42,15 @@ class ProximityInfoState {
|
||||
/////////////////////////////////////////
|
||||
// Defined in proximity_info_state.cpp //
|
||||
/////////////////////////////////////////
|
||||
void initInputParams(const int32_t* inputCodes, const int inputLength,
|
||||
const int* xCoordinates, const int* yCoordinates);
|
||||
void initInputParams(
|
||||
const ProximityInfo* proximityInfo, const int32_t* inputCodes, const int inputLength,
|
||||
const int* xCoordinates, const int* yCoordinates);
|
||||
|
||||
/////////////////////////////////////////
|
||||
// Defined here //
|
||||
/////////////////////////////////////////
|
||||
// TODO: Move the constructor to initInputParams
|
||||
ProximityInfoState(ProximityInfo* proximityInfo, const int maxProximityCharsSize,
|
||||
const bool hasTouchPositionCorrectionData, const int mostCommonKeyWidthSquare,
|
||||
const std::string localeStr, const int keyCount, const int cellHeight,
|
||||
const int cellWidth, const int gridHeight, const int gridWidth)
|
||||
: mProximityInfo(proximityInfo),
|
||||
MAX_PROXIMITY_CHARS_SIZE(maxProximityCharsSize),
|
||||
HAS_TOUCH_POSITION_CORRECTION_DATA(hasTouchPositionCorrectionData),
|
||||
MOST_COMMON_KEY_WIDTH_SQUARE(mostCommonKeyWidthSquare),
|
||||
LOCALE_STR(localeStr),
|
||||
KEY_COUNT(keyCount),
|
||||
CELL_HEIGHT(cellHeight),
|
||||
CELL_WIDTH(cellWidth),
|
||||
GRID_HEIGHT(gridHeight),
|
||||
GRID_WIDTH(gridWidth),
|
||||
mInputXCoordinates(0),
|
||||
mInputYCoordinates(0),
|
||||
mTouchPositionCorrectionEnabled(false) {
|
||||
const int normalizedSquaredDistancesLength =
|
||||
MAX_PROXIMITY_CHARS_SIZE_INTERNAL * MAX_WORD_LENGTH_INTERNAL;
|
||||
for (int i = 0; i < normalizedSquaredDistancesLength; ++i) {
|
||||
mNormalizedSquaredDistances[i] = NOT_A_DISTANCE;
|
||||
}
|
||||
}
|
||||
|
||||
inline const int* getProximityCharsAt(const int index) const {
|
||||
return mInputCodes + (index * MAX_PROXIMITY_CHARS_SIZE);
|
||||
return mInputCodes + (index * MAX_PROXIMITY_CHARS_SIZE_INTERNAL);
|
||||
}
|
||||
|
||||
inline unsigned short getPrimaryCharAt(const int index) const {
|
||||
@ -85,7 +60,7 @@ class ProximityInfoState {
|
||||
inline bool existsCharInProximityAt(const int index, const int c) const {
|
||||
const int *chars = getProximityCharsAt(index);
|
||||
int i = 0;
|
||||
while (chars[i] > 0 && i < MAX_PROXIMITY_CHARS_SIZE) {
|
||||
while (chars[i] > 0 && i < MAX_PROXIMITY_CHARS_SIZE_INTERNAL) {
|
||||
if (chars[i++] == c) {
|
||||
return true;
|
||||
}
|
||||
@ -120,7 +95,7 @@ class ProximityInfoState {
|
||||
// in their list. The non-accented version of the character should be considered
|
||||
// "close", but not the other keys close to the non-accented version.
|
||||
inline ProximityType getMatchedProximityId(const int index,
|
||||
const unsigned short c, const bool checkProximityChars, int *proximityIndex) const {
|
||||
const unsigned short c, const bool checkProximityChars, int *proximityIndex = 0) const {
|
||||
const int *currentChars = getProximityCharsAt(index);
|
||||
const int firstChar = currentChars[0];
|
||||
const unsigned short baseLowerC = toBaseLowerCase(c);
|
||||
@ -141,7 +116,7 @@ class ProximityInfoState {
|
||||
|
||||
// Not an exact nor an accent-alike match: search the list of close keys
|
||||
int j = 1;
|
||||
while (j < MAX_PROXIMITY_CHARS_SIZE
|
||||
while (j < MAX_PROXIMITY_CHARS_SIZE_INTERNAL
|
||||
&& currentChars[j] > ADDITIONAL_PROXIMITY_CHAR_DELIMITER_CODE) {
|
||||
const bool matched = (currentChars[j] == baseLowerC || currentChars[j] == c);
|
||||
if (matched) {
|
||||
@ -152,10 +127,10 @@ class ProximityInfoState {
|
||||
}
|
||||
++j;
|
||||
}
|
||||
if (j < MAX_PROXIMITY_CHARS_SIZE
|
||||
if (j < MAX_PROXIMITY_CHARS_SIZE_INTERNAL
|
||||
&& currentChars[j] == ADDITIONAL_PROXIMITY_CHAR_DELIMITER_CODE) {
|
||||
++j;
|
||||
while (j < MAX_PROXIMITY_CHARS_SIZE
|
||||
while (j < MAX_PROXIMITY_CHARS_SIZE_INTERNAL
|
||||
&& currentChars[j] > ADDITIONAL_PROXIMITY_CHAR_DELIMITER_CODE) {
|
||||
const bool matched = (currentChars[j] == baseLowerC || currentChars[j] == c);
|
||||
if (matched) {
|
||||
@ -174,7 +149,8 @@ class ProximityInfoState {
|
||||
|
||||
inline int getNormalizedSquaredDistance(
|
||||
const int inputIndex, const int proximityIndex) const {
|
||||
return mNormalizedSquaredDistances[inputIndex * MAX_PROXIMITY_CHARS_SIZE + proximityIndex];
|
||||
return mNormalizedSquaredDistances[
|
||||
inputIndex * MAX_PROXIMITY_CHARS_SIZE_INTERNAL + proximityIndex];
|
||||
}
|
||||
|
||||
inline const unsigned short* getPrimaryInputWord() const {
|
||||
@ -186,38 +162,23 @@ class ProximityInfoState {
|
||||
}
|
||||
|
||||
private:
|
||||
inline float square(const float x) const { return x * x; }
|
||||
/////////////////////////////////////////
|
||||
// Defined in proximity_info_state.cpp //
|
||||
/////////////////////////////////////////
|
||||
float calculateNormalizedSquaredDistance(const int keyIndex, const int inputIndex) const;
|
||||
|
||||
float calculateNormalizedSquaredDistance(
|
||||
const int keyIndex, const int inputIndex) const {
|
||||
if (keyIndex == NOT_AN_INDEX) {
|
||||
return NOT_A_DISTANCE_FLOAT;
|
||||
}
|
||||
if (!mProximityInfo->hasSweetSpotData(keyIndex)) {
|
||||
return NOT_A_DISTANCE_FLOAT;
|
||||
}
|
||||
if (NOT_A_COORDINATE == mInputXCoordinates[inputIndex]) {
|
||||
return NOT_A_DISTANCE_FLOAT;
|
||||
}
|
||||
const float squaredDistance = calculateSquaredDistanceFromSweetSpotCenter(
|
||||
keyIndex, inputIndex);
|
||||
const float squaredRadius = square(mProximityInfo->getSweetSpotRadiiAt(keyIndex));
|
||||
return squaredDistance / squaredRadius;
|
||||
}
|
||||
float calculateSquaredDistanceFromSweetSpotCenter(
|
||||
const int keyIndex, const int inputIndex) const;
|
||||
|
||||
/////////////////////////////////////////
|
||||
// Defined here //
|
||||
/////////////////////////////////////////
|
||||
inline float square(const float x) const { return x * x; }
|
||||
|
||||
bool hasInputCoordinates() const {
|
||||
return mInputXCoordinates && mInputYCoordinates;
|
||||
}
|
||||
|
||||
float calculateSquaredDistanceFromSweetSpotCenter(
|
||||
const int keyIndex, const int inputIndex) const {
|
||||
const float sweetSpotCenterX = mProximityInfo->getSweetSpotCenterXAt(keyIndex);
|
||||
const float sweetSpotCenterY = mProximityInfo->getSweetSpotCenterYAt(keyIndex);
|
||||
const float inputX = (float)mInputXCoordinates[inputIndex];
|
||||
const float inputY = (float)mInputYCoordinates[inputIndex];
|
||||
return square(inputX - sweetSpotCenterX) + square(inputY - sweetSpotCenterY);
|
||||
}
|
||||
|
||||
bool sameAsTyped(const unsigned short *word, int length) const {
|
||||
if (length != mInputLength) {
|
||||
return false;
|
||||
@ -227,23 +188,22 @@ class ProximityInfoState {
|
||||
if ((unsigned int) *inputCodes != (unsigned int) *word) {
|
||||
return false;
|
||||
}
|
||||
inputCodes += MAX_PROXIMITY_CHARS_SIZE;
|
||||
inputCodes += MAX_PROXIMITY_CHARS_SIZE_INTERNAL;
|
||||
word++;
|
||||
}
|
||||
return true;
|
||||
}
|
||||
|
||||
// TODO: const
|
||||
ProximityInfo *mProximityInfo;
|
||||
const int MAX_PROXIMITY_CHARS_SIZE;
|
||||
const bool HAS_TOUCH_POSITION_CORRECTION_DATA;
|
||||
const int MOST_COMMON_KEY_WIDTH_SQUARE;
|
||||
const std::string LOCALE_STR;
|
||||
const int KEY_COUNT;
|
||||
const int CELL_HEIGHT;
|
||||
const int CELL_WIDTH;
|
||||
const int GRID_HEIGHT;
|
||||
const int GRID_WIDTH;
|
||||
// const
|
||||
const ProximityInfo *mProximityInfo;
|
||||
bool mHasTouchPositionCorrectionData;
|
||||
int mMostCommonKeyWidthSquare;
|
||||
std::string mLocaleStr;
|
||||
int mKeyCount;
|
||||
int mCellHeight;
|
||||
int mCellWidth;
|
||||
int mGridHeight;
|
||||
int mGridWidth;
|
||||
|
||||
const int *mInputXCoordinates;
|
||||
const int *mInputYCoordinates;
|
||||
|
||||
@ -103,7 +103,7 @@ void UnigramDictionary::getWordWithDigraphSuggestionsRec(ProximityInfo *proximit
|
||||
const bool useFullEditDistance, const int *codesSrc,
|
||||
const int codesRemain, const int currentDepth, int *codesDest, Correction *correction,
|
||||
WordsPriorityQueuePool *queuePool,
|
||||
const digraph_t* const digraphs, const unsigned int digraphsSize) {
|
||||
const digraph_t* const digraphs, const unsigned int digraphsSize) const {
|
||||
|
||||
const int startIndex = codesDest - codesBuffer;
|
||||
if (currentDepth < MAX_DIGRAPH_SEARCH_DEPTH) {
|
||||
@ -173,7 +173,7 @@ int UnigramDictionary::getSuggestions(ProximityInfo *proximityInfo,
|
||||
WordsPriorityQueuePool *queuePool, Correction *correction, const int *xcoordinates,
|
||||
const int *ycoordinates, const int *codes, const int codesSize,
|
||||
const std::map<int, int> *bigramMap, const uint8_t *bigramFilter,
|
||||
const bool useFullEditDistance, unsigned short *outWords, int *frequencies) {
|
||||
const bool useFullEditDistance, unsigned short *outWords, int *frequencies) const {
|
||||
|
||||
queuePool->clearAll();
|
||||
Correction* masterCorrection = correction;
|
||||
@ -205,17 +205,17 @@ int UnigramDictionary::getSuggestions(ProximityInfo *proximityInfo,
|
||||
PROF_START(20);
|
||||
if (DEBUG_DICT) {
|
||||
float ns = queuePool->getMasterQueue()->getHighestNormalizedScore(
|
||||
proximityInfo->getPrimaryInputWord(), codesSize, 0, 0, 0);
|
||||
correction->getPrimaryInputWord(), codesSize, 0, 0, 0);
|
||||
ns += 0;
|
||||
AKLOGI("Max normalized score = %f", ns);
|
||||
}
|
||||
const int suggestedWordsCount =
|
||||
queuePool->getMasterQueue()->outputSuggestions(
|
||||
proximityInfo->getPrimaryInputWord(), codesSize, frequencies, outWords);
|
||||
correction->getPrimaryInputWord(), codesSize, frequencies, outWords);
|
||||
|
||||
if (DEBUG_DICT) {
|
||||
float ns = queuePool->getMasterQueue()->getHighestNormalizedScore(
|
||||
proximityInfo->getPrimaryInputWord(), codesSize, 0, 0, 0);
|
||||
correction->getPrimaryInputWord(), codesSize, 0, 0, 0);
|
||||
ns += 0;
|
||||
AKLOGI("Returning %d words", suggestedWordsCount);
|
||||
/// Print the returned words
|
||||
@ -235,7 +235,8 @@ int UnigramDictionary::getSuggestions(ProximityInfo *proximityInfo,
|
||||
void UnigramDictionary::getWordSuggestions(ProximityInfo *proximityInfo,
|
||||
const int *xcoordinates, const int *ycoordinates, const int *codes,
|
||||
const int inputLength, const std::map<int, int> *bigramMap, const uint8_t *bigramFilter,
|
||||
const bool useFullEditDistance, Correction *correction, WordsPriorityQueuePool *queuePool) {
|
||||
const bool useFullEditDistance, Correction *correction,
|
||||
WordsPriorityQueuePool *queuePool) const {
|
||||
|
||||
PROF_OPEN;
|
||||
PROF_START(0);
|
||||
@ -259,7 +260,7 @@ void UnigramDictionary::getWordSuggestions(ProximityInfo *proximityInfo,
|
||||
WordsPriorityQueue* masterQueue = queuePool->getMasterQueue();
|
||||
if (masterQueue->size() > 0) {
|
||||
float nsForMaster = masterQueue->getHighestNormalizedScore(
|
||||
proximityInfo->getPrimaryInputWord(), inputLength, 0, 0, 0);
|
||||
correction->getPrimaryInputWord(), inputLength, 0, 0, 0);
|
||||
hasAutoCorrectionCandidate = (nsForMaster > START_TWO_WORDS_CORRECTION_THRESHOLD);
|
||||
}
|
||||
PROF_END(4);
|
||||
@ -288,11 +289,11 @@ void UnigramDictionary::getWordSuggestions(ProximityInfo *proximityInfo,
|
||||
const unsigned short* word = sw->mWord;
|
||||
const int wordLength = sw->mWordLength;
|
||||
float ns = Correction::RankingAlgorithm::calcNormalizedScore(
|
||||
proximityInfo->getPrimaryInputWord(), i, word, wordLength, score);
|
||||
correction->getPrimaryInputWord(), i, word, wordLength, score);
|
||||
ns += 0;
|
||||
AKLOGI("--- TOP SUB WORDS for %d --- %d %f [%d]", i, score, ns,
|
||||
(ns > TWO_WORDS_CORRECTION_WITH_OTHER_ERROR_THRESHOLD));
|
||||
DUMP_WORD(proximityInfo->getPrimaryInputWord(), i);
|
||||
DUMP_WORD(correction->getPrimaryInputWord(), i);
|
||||
DUMP_WORD(word, wordLength);
|
||||
}
|
||||
}
|
||||
@ -300,12 +301,13 @@ void UnigramDictionary::getWordSuggestions(ProximityInfo *proximityInfo,
|
||||
}
|
||||
|
||||
void UnigramDictionary::initSuggestions(ProximityInfo *proximityInfo, const int *xCoordinates,
|
||||
const int *yCoordinates, const int *codes, const int inputLength, Correction *correction) {
|
||||
const int *yCoordinates, const int *codes, const int inputLength,
|
||||
Correction *correction) const {
|
||||
if (DEBUG_DICT) {
|
||||
AKLOGI("initSuggest");
|
||||
DUMP_WORD_INT(codes, inputLength);
|
||||
}
|
||||
proximityInfo->initInputParams(codes, inputLength, xCoordinates, yCoordinates);
|
||||
correction->initInputParams(proximityInfo, codes, inputLength, xCoordinates, yCoordinates);
|
||||
const int maxDepth = min(inputLength * MAX_DEPTH_MULTIPLIER, MAX_WORD_LENGTH);
|
||||
correction->initCorrection(proximityInfo, inputLength, maxDepth);
|
||||
}
|
||||
@ -317,7 +319,7 @@ void UnigramDictionary::getOneWordSuggestions(ProximityInfo *proximityInfo,
|
||||
const int *xcoordinates, const int *ycoordinates, const int *codes,
|
||||
const std::map<int, int> *bigramMap, const uint8_t *bigramFilter,
|
||||
const bool useFullEditDistance, const int inputLength,
|
||||
Correction *correction, WordsPriorityQueuePool *queuePool) {
|
||||
Correction *correction, WordsPriorityQueuePool *queuePool) const {
|
||||
initSuggestions(proximityInfo, xcoordinates, ycoordinates, codes, inputLength, correction);
|
||||
getSuggestionCandidates(useFullEditDistance, inputLength, bigramMap, bigramFilter, correction,
|
||||
queuePool, true /* doAutoCompletion */, DEFAULT_MAX_ERRORS, FIRST_WORD_INDEX);
|
||||
@ -326,7 +328,7 @@ void UnigramDictionary::getOneWordSuggestions(ProximityInfo *proximityInfo,
|
||||
void UnigramDictionary::getSuggestionCandidates(const bool useFullEditDistance,
|
||||
const int inputLength, const std::map<int, int> *bigramMap, const uint8_t *bigramFilter,
|
||||
Correction *correction, WordsPriorityQueuePool *queuePool,
|
||||
const bool doAutoCompletion, const int maxErrors, const int currentWordIndex) {
|
||||
const bool doAutoCompletion, const int maxErrors, const int currentWordIndex) const {
|
||||
uint8_t totalTraverseCount = correction->pushAndGetTotalTraverseCount();
|
||||
if (DEBUG_DICT) {
|
||||
AKLOGI("Traverse count %d", totalTraverseCount);
|
||||
@ -374,7 +376,7 @@ void UnigramDictionary::getSuggestionCandidates(const bool useFullEditDistance,
|
||||
inline void UnigramDictionary::onTerminal(const int probability,
|
||||
const TerminalAttributes& terminalAttributes, Correction *correction,
|
||||
WordsPriorityQueuePool *queuePool, const bool addToMasterQueue,
|
||||
const int currentWordIndex) {
|
||||
const int currentWordIndex) const {
|
||||
const int inputIndex = correction->getInputIndex();
|
||||
const bool addToSubQueue = inputIndex < SUB_QUEUE_MAX_COUNT;
|
||||
|
||||
@ -430,7 +432,7 @@ int UnigramDictionary::getSubStringSuggestion(
|
||||
const bool hasAutoCorrectionCandidate, const int currentWordIndex,
|
||||
const int inputWordStartPos, const int inputWordLength,
|
||||
const int outputWordStartPos, const bool isSpaceProximity, int *freqArray,
|
||||
int*wordLengthArray, unsigned short* outputWord, int *outputWordLength) {
|
||||
int*wordLengthArray, unsigned short* outputWord, int *outputWordLength) const {
|
||||
if (inputWordLength > MULTIPLE_WORDS_SUGGESTION_MAX_WORD_LENGTH) {
|
||||
return FLAG_MULTIPLE_SUGGEST_ABORT;
|
||||
}
|
||||
@ -479,11 +481,12 @@ int UnigramDictionary::getSubStringSuggestion(
|
||||
initSuggestions(proximityInfo, xcoordinates, ycoordinates, codes,
|
||||
inputLength, correction);
|
||||
|
||||
unsigned short word[MAX_WORD_LENGTH_INTERNAL];
|
||||
int freq = getMostFrequentWordLike(
|
||||
inputWordStartPos, inputWordLength, proximityInfo, mWord);
|
||||
inputWordStartPos, inputWordLength, correction, word);
|
||||
if (freq > 0) {
|
||||
nextWordLength = inputWordLength;
|
||||
tempOutputWord = mWord;
|
||||
tempOutputWord = word;
|
||||
} else if (!hasAutoCorrectionCandidate) {
|
||||
if (inputWordStartPos > 0) {
|
||||
const int offset = inputWordStartPos;
|
||||
@ -510,7 +513,7 @@ int UnigramDictionary::getSubStringSuggestion(
|
||||
}
|
||||
int score = 0;
|
||||
const float ns = queue->getHighestNormalizedScore(
|
||||
proximityInfo->getPrimaryInputWord(), inputWordLength,
|
||||
correction->getPrimaryInputWord(), inputWordLength,
|
||||
&tempOutputWord, &score, &nextWordLength);
|
||||
if (DEBUG_DICT) {
|
||||
AKLOGI("NS(%d) = %f, Score = %d", currentWordIndex, ns, score);
|
||||
@ -577,7 +580,7 @@ void UnigramDictionary::getMultiWordsSuggestionRec(ProximityInfo *proximityInfo,
|
||||
Correction *correction, WordsPriorityQueuePool* queuePool,
|
||||
const bool hasAutoCorrectionCandidate, const int startInputPos, const int startWordIndex,
|
||||
const int outputWordLength, int *freqArray, int* wordLengthArray,
|
||||
unsigned short* outputWord) {
|
||||
unsigned short* outputWord) const {
|
||||
if (startWordIndex >= (MULTIPLE_WORDS_SUGGESTION_MAX_WORDS - 1)) {
|
||||
// Return if the last word index
|
||||
return;
|
||||
@ -656,7 +659,7 @@ void UnigramDictionary::getSplitMultipleWordsSuggestions(ProximityInfo *proximit
|
||||
const int *xcoordinates, const int *ycoordinates, const int *codes,
|
||||
const bool useFullEditDistance, const int inputLength,
|
||||
Correction *correction, WordsPriorityQueuePool* queuePool,
|
||||
const bool hasAutoCorrectionCandidate) {
|
||||
const bool hasAutoCorrectionCandidate) const {
|
||||
if (inputLength >= MAX_WORD_LENGTH) return;
|
||||
if (DEBUG_DICT) {
|
||||
AKLOGI("--- Suggest multiple words");
|
||||
@ -678,11 +681,11 @@ void UnigramDictionary::getSplitMultipleWordsSuggestions(ProximityInfo *proximit
|
||||
// Wrapper for getMostFrequentWordLikeInner, which matches it to the previous
|
||||
// interface.
|
||||
inline int UnigramDictionary::getMostFrequentWordLike(const int startInputIndex,
|
||||
const int inputLength, ProximityInfo *proximityInfo, unsigned short *word) {
|
||||
const int inputLength, Correction *correction, unsigned short *word) const {
|
||||
uint16_t inWord[inputLength];
|
||||
|
||||
for (int i = 0; i < inputLength; ++i) {
|
||||
inWord[i] = (uint16_t)proximityInfo->getPrimaryCharAt(startInputIndex + i);
|
||||
inWord[i] = (uint16_t)correction->getPrimaryCharAt(startInputIndex + i);
|
||||
}
|
||||
return getMostFrequentWordLikeInner(inWord, inputLength, word);
|
||||
}
|
||||
@ -751,21 +754,24 @@ static inline void onTerminalWordLike(const int freq, int32_t* newWord, const in
|
||||
// Will find the highest frequency of the words like the one passed as an argument,
|
||||
// that is, everything that only differs by case/accents.
|
||||
int UnigramDictionary::getMostFrequentWordLikeInner(const uint16_t * const inWord,
|
||||
const int length, short unsigned int* outWord) {
|
||||
const int length, short unsigned int* outWord) const {
|
||||
int32_t newWord[MAX_WORD_LENGTH_INTERNAL];
|
||||
int depth = 0;
|
||||
int maxFreq = -1;
|
||||
const uint8_t* const root = DICT_ROOT;
|
||||
int stackChildCount[MAX_WORD_LENGTH_INTERNAL];
|
||||
int stackInputIndex[MAX_WORD_LENGTH_INTERNAL];
|
||||
int stackSiblingPos[MAX_WORD_LENGTH_INTERNAL];
|
||||
|
||||
int startPos = 0;
|
||||
mStackChildCount[0] = BinaryFormat::getGroupCountAndForwardPointer(root, &startPos);
|
||||
mStackInputIndex[0] = 0;
|
||||
mStackSiblingPos[0] = startPos;
|
||||
stackChildCount[0] = BinaryFormat::getGroupCountAndForwardPointer(root, &startPos);
|
||||
stackInputIndex[0] = 0;
|
||||
stackSiblingPos[0] = startPos;
|
||||
while (depth >= 0) {
|
||||
const int charGroupCount = mStackChildCount[depth];
|
||||
int pos = mStackSiblingPos[depth];
|
||||
const int charGroupCount = stackChildCount[depth];
|
||||
int pos = stackSiblingPos[depth];
|
||||
for (int charGroupIndex = charGroupCount - 1; charGroupIndex >= 0; --charGroupIndex) {
|
||||
int inputIndex = mStackInputIndex[depth];
|
||||
int inputIndex = stackInputIndex[depth];
|
||||
const uint8_t flags = BinaryFormat::getFlagsAndForwardPointer(root, &pos);
|
||||
// Test whether all chars in this group match with the word we are searching for. If so,
|
||||
// we want to traverse its children (or if the length match, evaluate its frequency).
|
||||
@ -785,15 +791,15 @@ int UnigramDictionary::getMostFrequentWordLikeInner(const uint16_t * const inWor
|
||||
// anyway, so don't traverse unless inputIndex < length.
|
||||
if (isAlike && (-1 != childrenNodePos) && (inputIndex < length)) {
|
||||
// Save position for this depth, to get back to this once children are done
|
||||
mStackChildCount[depth] = charGroupIndex;
|
||||
mStackSiblingPos[depth] = siblingPos;
|
||||
stackChildCount[depth] = charGroupIndex;
|
||||
stackSiblingPos[depth] = siblingPos;
|
||||
// Prepare stack values for next depth
|
||||
++depth;
|
||||
int childrenPos = childrenNodePos;
|
||||
mStackChildCount[depth] =
|
||||
stackChildCount[depth] =
|
||||
BinaryFormat::getGroupCountAndForwardPointer(root, &childrenPos);
|
||||
mStackSiblingPos[depth] = childrenPos;
|
||||
mStackInputIndex[depth] = inputIndex;
|
||||
stackSiblingPos[depth] = childrenPos;
|
||||
stackInputIndex[depth] = inputIndex;
|
||||
pos = childrenPos;
|
||||
// Go to the next depth level.
|
||||
++depth;
|
||||
@ -848,7 +854,7 @@ int UnigramDictionary::getBigramPosition(int pos, unsigned short *word, int offs
|
||||
inline bool UnigramDictionary::processCurrentNode(const int initialPos,
|
||||
const std::map<int, int> *bigramMap, const uint8_t *bigramFilter, Correction *correction,
|
||||
int *newCount, int *newChildrenPosition, int *nextSiblingPosition,
|
||||
WordsPriorityQueuePool *queuePool, const int currentWordIndex) {
|
||||
WordsPriorityQueuePool *queuePool, const int currentWordIndex) const {
|
||||
if (DEBUG_DICT) {
|
||||
correction->checkState();
|
||||
}
|
||||
|
||||
@ -81,7 +81,7 @@ class UnigramDictionary {
|
||||
Correction *correction, const int *xcoordinates, const int *ycoordinates,
|
||||
const int *codes, const int codesSize, const std::map<int, int> *bigramMap,
|
||||
const uint8_t *bigramFilter, const bool useFullEditDistance, unsigned short *outWords,
|
||||
int *frequencies);
|
||||
int *frequencies) const;
|
||||
virtual ~UnigramDictionary();
|
||||
|
||||
private:
|
||||
@ -89,7 +89,7 @@ class UnigramDictionary {
|
||||
const int *ycoordinates, const int *codes, const int inputLength,
|
||||
const std::map<int, int> *bigramMap, const uint8_t *bigramFilter,
|
||||
const bool useFullEditDistance, Correction *correction,
|
||||
WordsPriorityQueuePool *queuePool);
|
||||
WordsPriorityQueuePool *queuePool) const;
|
||||
int getDigraphReplacement(const int *codes, const int i, const int codesSize,
|
||||
const digraph_t* const digraphs, const unsigned int digraphsSize) const;
|
||||
void getWordWithDigraphSuggestionsRec(ProximityInfo *proximityInfo,
|
||||
@ -99,37 +99,36 @@ class UnigramDictionary {
|
||||
const bool useFullEditDistance, const int* codesSrc, const int codesRemain,
|
||||
const int currentDepth, int* codesDest, Correction *correction,
|
||||
WordsPriorityQueuePool* queuePool, const digraph_t* const digraphs,
|
||||
const unsigned int digraphsSize);
|
||||
const unsigned int digraphsSize) const;
|
||||
void initSuggestions(ProximityInfo *proximityInfo, const int *xcoordinates,
|
||||
const int *ycoordinates, const int *codes, const int codesSize, Correction *correction);
|
||||
const int *ycoordinates, const int *codes, const int codesSize,
|
||||
Correction *correction) const;
|
||||
void getOneWordSuggestions(ProximityInfo *proximityInfo, const int *xcoordinates,
|
||||
const int *ycoordinates, const int *codes, const std::map<int, int> *bigramMap,
|
||||
const uint8_t *bigramFilter, const bool useFullEditDistance, const int inputLength,
|
||||
Correction *correction, WordsPriorityQueuePool* queuePool);
|
||||
Correction *correction, WordsPriorityQueuePool* queuePool) const;
|
||||
void getSuggestionCandidates(
|
||||
const bool useFullEditDistance, const int inputLength,
|
||||
const std::map<int, int> *bigramMap, const uint8_t *bigramFilter,
|
||||
Correction *correction, WordsPriorityQueuePool* queuePool, const bool doAutoCompletion,
|
||||
const int maxErrors, const int currentWordIndex);
|
||||
const int maxErrors, const int currentWordIndex) const;
|
||||
void getSplitMultipleWordsSuggestions(ProximityInfo *proximityInfo,
|
||||
const int *xcoordinates, const int *ycoordinates, const int *codes,
|
||||
const bool useFullEditDistance, const int inputLength,
|
||||
Correction *correction, WordsPriorityQueuePool* queuePool,
|
||||
const bool hasAutoCorrectionCandidate);
|
||||
const bool hasAutoCorrectionCandidate) const;
|
||||
void onTerminal(const int freq, const TerminalAttributes& terminalAttributes,
|
||||
Correction *correction, WordsPriorityQueuePool *queuePool, const bool addToMasterQueue,
|
||||
const int currentWordIndex);
|
||||
bool needsToSkipCurrentNode(const unsigned short c,
|
||||
const int inputIndex, const int skipPos, const int depth);
|
||||
const int currentWordIndex) const;
|
||||
// Process a node by considering proximity, missing and excessive character
|
||||
bool processCurrentNode(const int initialPos, const std::map<int, int> *bigramMap,
|
||||
const uint8_t *bigramFilter, Correction *correction, int *newCount,
|
||||
int *newChildPosition, int *nextSiblingPosition, WordsPriorityQueuePool *queuePool,
|
||||
const int currentWordIndex);
|
||||
const int currentWordIndex) const;
|
||||
int getMostFrequentWordLike(const int startInputIndex, const int inputLength,
|
||||
ProximityInfo *proximityInfo, unsigned short *word);
|
||||
Correction *correction, unsigned short *word) const;
|
||||
int getMostFrequentWordLikeInner(const uint16_t* const inWord, const int length,
|
||||
short unsigned int *outWord);
|
||||
short unsigned int *outWord) const;
|
||||
int getSubStringSuggestion(
|
||||
ProximityInfo *proximityInfo, const int *xcoordinates, const int *ycoordinates,
|
||||
const int *codes, const bool useFullEditDistance, Correction *correction,
|
||||
@ -137,14 +136,14 @@ class UnigramDictionary {
|
||||
const bool hasAutoCorrectionCandidate, const int currentWordIndex,
|
||||
const int inputWordStartPos, const int inputWordLength,
|
||||
const int outputWordStartPos, const bool isSpaceProximity, int *freqArray,
|
||||
int *wordLengthArray, unsigned short* outputWord, int *outputWordLength);
|
||||
int *wordLengthArray, unsigned short* outputWord, int *outputWordLength) const;
|
||||
void getMultiWordsSuggestionRec(ProximityInfo *proximityInfo,
|
||||
const int *xcoordinates, const int *ycoordinates, const int *codes,
|
||||
const bool useFullEditDistance, const int inputLength,
|
||||
Correction *correction, WordsPriorityQueuePool* queuePool,
|
||||
const bool hasAutoCorrectionCandidate, const int startPos, const int startWordIndex,
|
||||
const int outputWordLength, int *freqArray, int* wordLengthArray,
|
||||
unsigned short* outputWord);
|
||||
unsigned short* outputWord) const;
|
||||
|
||||
const uint8_t* const DICT_ROOT;
|
||||
const int MAX_WORD_LENGTH;
|
||||
@ -158,12 +157,6 @@ class UnigramDictionary {
|
||||
|
||||
static const digraph_t GERMAN_UMLAUT_DIGRAPHS[];
|
||||
static const digraph_t FRENCH_LIGATURES_DIGRAPHS[];
|
||||
|
||||
// Still bundled members
|
||||
unsigned short mWord[MAX_WORD_LENGTH_INTERNAL];// TODO: remove
|
||||
int mStackChildCount[MAX_WORD_LENGTH_INTERNAL];// TODO: remove
|
||||
int mStackInputIndex[MAX_WORD_LENGTH_INTERNAL];// TODO: remove
|
||||
int mStackSiblingPos[MAX_WORD_LENGTH_INTERNAL];// TODO: remove
|
||||
};
|
||||
} // namespace latinime
|
||||
|
||||
|
||||
Loading…
Reference in New Issue
Block a user