aboutsummaryrefslogtreecommitdiffstats
path: root/native/jni/src
diff options
context:
space:
mode:
Diffstat (limited to 'native/jni/src')
-rw-r--r--native/jni/src/bigram_dictionary.cpp4
-rw-r--r--native/jni/src/bigram_dictionary.h4
-rw-r--r--native/jni/src/binary_format.h15
-rw-r--r--native/jni/src/correction.cpp26
-rw-r--r--native/jni/src/correction.h14
-rw-r--r--native/jni/src/defines.h2
-rw-r--r--native/jni/src/dictionary.cpp2
-rw-r--r--native/jni/src/dictionary.h12
-rw-r--r--native/jni/src/unigram_dictionary.cpp93
-rw-r--r--native/jni/src/unigram_dictionary.h29
10 files changed, 113 insertions, 88 deletions
diff --git a/native/jni/src/bigram_dictionary.cpp b/native/jni/src/bigram_dictionary.cpp
index 320b0af68..927381fdb 100644
--- a/native/jni/src/bigram_dictionary.cpp
+++ b/native/jni/src/bigram_dictionary.cpp
@@ -96,7 +96,7 @@ bool BigramDictionary::addWordBigram(unsigned short *word, int length, int frequ
* and the bigrams are used to boost unigram result scores, it makes little sense to
* reduce their scope to the ones that match the first letter.
*/
-int BigramDictionary::getBigrams(unsigned short *prevWord, int prevWordLength, int *codes,
+int BigramDictionary::getBigrams(const int32_t *prevWord, int prevWordLength, int *codes,
int codesSize, unsigned short *bigramChars, int *bigramFreq, int maxWordLength,
int maxBigrams) {
// TODO: remove unused arguments, and refrain from storing stuff in members of this class
@@ -134,7 +134,7 @@ int BigramDictionary::getBigrams(unsigned short *prevWord, int prevWordLength, i
// Returns a pointer to the start of the bigram list.
// If the word is not found or has no bigrams, this function returns 0.
int BigramDictionary::getBigramListForWord(const uint8_t* const root,
- const unsigned short *prevWord, const int prevWordLength) {
+ const int32_t *prevWord, const int prevWordLength) {
int pos = BinaryFormat::getTerminalPosition(root, prevWord, prevWordLength);
if (NOT_VALID_WORD == pos) return 0;
diff --git a/native/jni/src/bigram_dictionary.h b/native/jni/src/bigram_dictionary.h
index 1612131c4..07e47f059 100644
--- a/native/jni/src/bigram_dictionary.h
+++ b/native/jni/src/bigram_dictionary.h
@@ -25,10 +25,10 @@ class Dictionary;
class BigramDictionary {
public:
BigramDictionary(const unsigned char *dict, int maxWordLength, Dictionary *parentDictionary);
- int getBigrams(unsigned short *word, int length, int *codes, int codesSize,
+ int getBigrams(const int32_t *word, int length, int *codes, int codesSize,
unsigned short *outWords, int *frequencies, int maxWordLength, int maxBigrams);
int getBigramListForWord(const uint8_t* const root,
- const unsigned short *prevWord, const int prevWordLength);
+ const int32_t *prevWord, const int prevWordLength);
~BigramDictionary();
private:
bool addWordBigram(unsigned short *word, int length, int frequency);
diff --git a/native/jni/src/binary_format.h b/native/jni/src/binary_format.h
index f59302460..d5d67c108 100644
--- a/native/jni/src/binary_format.h
+++ b/native/jni/src/binary_format.h
@@ -62,10 +62,11 @@ class BinaryFormat {
static bool hasChildrenInFlags(const uint8_t flags);
static int getAttributeAddressAndForwardPointer(const uint8_t* const dict, const uint8_t flags,
int *pos);
- static int getTerminalPosition(const uint8_t* const root, const uint16_t* const inWord,
+ static int getTerminalPosition(const uint8_t* const root, const int32_t* const inWord,
const int length);
static int getWordAtAddress(const uint8_t* const root, const int address, const int maxDepth,
uint16_t* outWord);
+ static int getProbability(const int bigramListPosition, const int unigramFreq);
// Flags for special processing
// Those *must* match the flags in makedict (BinaryDictInputOutput#*_PROCESSING_FLAG) or
@@ -304,7 +305,7 @@ inline int BinaryFormat::getAttributeAddressAndForwardPointer(const uint8_t* con
// This function gets the byte position of the last chargroup of the exact matching word in the
// dictionary. If no match is found, it returns NOT_VALID_WORD.
inline int BinaryFormat::getTerminalPosition(const uint8_t* const root,
- const uint16_t* const inWord, const int length) {
+ const int32_t* const inWord, const int length) {
int pos = 0;
int wordPos = 0;
@@ -313,7 +314,7 @@ inline int BinaryFormat::getTerminalPosition(const uint8_t* const root,
// there was no match (or we would have found it).
if (wordPos > length) return NOT_VALID_WORD;
int charGroupCount = BinaryFormat::getGroupCountAndForwardPointer(root, &pos);
- const uint16_t wChar = inWord[wordPos];
+ const int32_t wChar = inWord[wordPos];
while (true) {
// If there are no more character groups in this node, it means we could not
// find a matching character for this depth, therefore there is no match.
@@ -517,6 +518,14 @@ inline int BinaryFormat::getWordAtAddress(const uint8_t* const root, const int a
return 0;
}
+// This should probably return a probability in log space.
+inline int BinaryFormat::getProbability(const int bigramListPosition, const int unigramFreq) {
+ // TODO: use the bigram list position to get the bigram probability. If the bigram
+ // is not found, use the unigram frequency.
+ // TODO: if the unigram frequency is used, compute the actual probability
+ return unigramFreq;
+}
+
} // namespace latinime
#endif // LATINIME_BINARY_FORMAT_H
diff --git a/native/jni/src/correction.cpp b/native/jni/src/correction.cpp
index 087219ed4..376e9a10e 100644
--- a/native/jni/src/correction.cpp
+++ b/native/jni/src/correction.cpp
@@ -165,28 +165,28 @@ int Correction::getFreqForSplitMultipleWords(const int *freqArray, const int *wo
wordCount, this, isSpaceProximity, word);
}
-int Correction::getFinalFreq(const int freq, unsigned short **word, int *wordLength) {
- return getFinalFreqInternal(freq, word, wordLength, mInputLength);
+int Correction::getFinalProbability(const int probability, unsigned short **word, int *wordLength) {
+ return getFinalProbabilityInternal(probability, word, wordLength, mInputLength);
}
-int Correction::getFinalFreqForSubQueue(const int freq, unsigned short **word, int *wordLength,
- const int inputLength) {
- return getFinalFreqInternal(freq, word, wordLength, inputLength);
+int Correction::getFinalProbabilityForSubQueue(const int probability, unsigned short **word,
+ int *wordLength, const int inputLength) {
+ return getFinalProbabilityInternal(probability, word, wordLength, inputLength);
}
-int Correction::getFinalFreqInternal(const int freq, unsigned short **word, int *wordLength,
- const int inputLength) {
+int Correction::getFinalProbabilityInternal(const int probability, unsigned short **word,
+ int *wordLength, const int inputLength) {
const int outputIndex = mTerminalOutputIndex;
const int inputIndex = mTerminalInputIndex;
*wordLength = outputIndex + 1;
if (outputIndex < MIN_SUGGEST_DEPTH) {
- return NOT_A_FREQUENCY;
+ return NOT_A_PROBABILITY;
}
*word = mWord;
- int finalFreq = Correction::RankingAlgorithm::calculateFinalFreq(
- inputIndex, outputIndex, freq, mEditDistanceTable, this, inputLength);
- return finalFreq;
+ int finalProbability= Correction::RankingAlgorithm::calculateFinalProbability(
+ inputIndex, outputIndex, probability, mEditDistanceTable, this, inputLength);
+ return finalProbability;
}
bool Correction::initProcessState(const int outputIndex) {
@@ -649,8 +649,8 @@ inline static bool isUpperCase(unsigned short c) {
//////////////////////
/* static */
-int Correction::RankingAlgorithm::calculateFinalFreq(const int inputIndex, const int outputIndex,
- const int freq, int* editDistanceTable, const Correction* correction,
+int Correction::RankingAlgorithm::calculateFinalProbability(const int inputIndex,
+ const int outputIndex, const int freq, int* editDistanceTable, const Correction* correction,
const int inputLength) {
const int excessivePos = correction->getExcessivePos();
const int typedLetterMultiplier = correction->TYPED_LETTER_MULTIPLIER;
diff --git a/native/jni/src/correction.h b/native/jni/src/correction.h
index ee55c9604..1b4e4bf4e 100644
--- a/native/jni/src/correction.h
+++ b/native/jni/src/correction.h
@@ -132,9 +132,9 @@ class Correction {
int getFreqForSplitMultipleWords(
const int *freqArray, const int *wordLengthArray, const int wordCount,
const bool isSpaceProximity, const unsigned short *word);
- int getFinalFreq(const int freq, unsigned short **word, int* wordLength);
- int getFinalFreqForSubQueue(const int freq, unsigned short **word, int* wordLength,
- const int inputLength);
+ int getFinalProbability(const int probability, unsigned short **word, int* wordLength);
+ int getFinalProbabilityForSubQueue(const int probability, unsigned short **word,
+ int* wordLength, const int inputLength);
CorrectionType processCharAndCalcState(const int32_t c, const bool isTerminal);
@@ -156,8 +156,8 @@ class Correction {
class RankingAlgorithm {
public:
- static int calculateFinalFreq(const int inputIndex, const int depth,
- const int freq, int *editDistanceTable, const Correction* correction,
+ static int calculateFinalProbability(const int inputIndex, const int depth,
+ const int probability, int *editDistanceTable, const Correction* correction,
const int inputLength);
static int calcFreqForSplitMultipleWords(const int *freqArray, const int *wordLengthArray,
const int wordCount, const Correction* correction, const bool isSpaceProximity,
@@ -182,8 +182,8 @@ class Correction {
const int32_t c, const bool isTerminal, const bool inputIndexIncremented);
inline CorrectionType processUnrelatedCorrectionType();
inline void addCharToCurrentWord(const int32_t c);
- inline int getFinalFreqInternal(const int freq, unsigned short **word, int* wordLength,
- const int inputLength);
+ inline int getFinalProbabilityInternal(const int probability, unsigned short **word,
+ int* wordLength, const int inputLength);
const int TYPED_LETTER_MULTIPLIER;
const int FULL_WORD_MULTIPLIER;
diff --git a/native/jni/src/defines.h b/native/jni/src/defines.h
index e882c3714..c99f8a8b2 100644
--- a/native/jni/src/defines.h
+++ b/native/jni/src/defines.h
@@ -172,7 +172,7 @@ static inline void prof_out(void) {
#define PROXIMITY_CHAR_WITHOUT_DISTANCE_INFO -3
#define ADDITIONAL_PROXIMITY_CHAR_DISTANCE_INFO -4
#define NOT_AN_INDEX -1
-#define NOT_A_FREQUENCY -1
+#define NOT_A_PROBABILITY -1
#define KEYCODE_SPACE ' '
diff --git a/native/jni/src/dictionary.cpp b/native/jni/src/dictionary.cpp
index 90ec207f0..9dc207223 100644
--- a/native/jni/src/dictionary.cpp
+++ b/native/jni/src/dictionary.cpp
@@ -54,7 +54,7 @@ Dictionary::~Dictionary() {
delete mBigramDictionary;
}
-bool Dictionary::isValidWord(unsigned short *word, int length) {
+bool Dictionary::isValidWord(const int32_t *word, int length) {
return mUnigramDictionary->isValidWord(word, length);
}
diff --git a/native/jni/src/dictionary.h b/native/jni/src/dictionary.h
index 66a5c2150..5b9ddb3e9 100644
--- a/native/jni/src/dictionary.h
+++ b/native/jni/src/dictionary.h
@@ -35,18 +35,22 @@ class Dictionary {
int getSuggestions(ProximityInfo *proximityInfo, int *xcoordinates, int *ycoordinates,
int *codes, int codesSize, bool useFullEditDistance, unsigned short *outWords,
int *frequencies) {
+ // bigramListPosition is, as an int, the offset of the bigram list in the file.
+ // If none, it's zero.
+ // TODO: get this from the bigram dictionary instance
+ const int bigramListPosition = 0;
return mUnigramDictionary->getSuggestions(proximityInfo, mWordsPriorityQueuePool,
- mCorrection, xcoordinates, ycoordinates, codes,
- codesSize, useFullEditDistance, outWords, frequencies);
+ mCorrection, xcoordinates, ycoordinates, codes, codesSize, bigramListPosition,
+ useFullEditDistance, outWords, frequencies);
}
- int getBigrams(unsigned short *word, int length, int *codes, int codesSize,
+ int getBigrams(const int32_t *word, int length, int *codes, int codesSize,
unsigned short *outWords, int *frequencies, int maxWordLength, int maxBigrams) {
return mBigramDictionary->getBigrams(word, length, codes, codesSize, outWords, frequencies,
maxWordLength, maxBigrams);
}
- bool isValidWord(unsigned short *word, int length);
+ bool isValidWord(const int32_t *word, int length);
void *getDict() { return (void *)mDict; }
int getDictSize() { return mDictSize; }
int getMmapFd() { return mMmapFd; }
diff --git a/native/jni/src/unigram_dictionary.cpp b/native/jni/src/unigram_dictionary.cpp
index ab8570e6f..0c759d438 100644
--- a/native/jni/src/unigram_dictionary.cpp
+++ b/native/jni/src/unigram_dictionary.cpp
@@ -98,7 +98,8 @@ int UnigramDictionary::getDigraphReplacement(const int *codes, const int i, cons
void UnigramDictionary::getWordWithDigraphSuggestionsRec(ProximityInfo *proximityInfo,
const int *xcoordinates, const int *ycoordinates, const int *codesBuffer,
int *xCoordinatesBuffer, int *yCoordinatesBuffer,
- const int codesBufferSize, const bool useFullEditDistance, const int *codesSrc,
+ const int codesBufferSize, const int bigramListPosition,
+ 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) {
@@ -127,8 +128,8 @@ void UnigramDictionary::getWordWithDigraphSuggestionsRec(ProximityInfo *proximit
replacementCodePoint;
getWordWithDigraphSuggestionsRec(proximityInfo, xcoordinates, ycoordinates,
codesBuffer, xCoordinatesBuffer, yCoordinatesBuffer, codesBufferSize,
- useFullEditDistance, codesSrc + i + 1, codesRemain - i - 1,
- currentDepth + 1, codesDest + i, correction,
+ bigramListPosition, useFullEditDistance, codesSrc + i + 1,
+ codesRemain - i - 1, currentDepth + 1, codesDest + i, correction,
queuePool, digraphs, digraphsSize);
// Copy the second char of the digraph in place, then continue processing on
@@ -137,9 +138,9 @@ void UnigramDictionary::getWordWithDigraphSuggestionsRec(ProximityInfo *proximit
memcpy(codesDest + i, codesSrc + i, BYTES_IN_ONE_CHAR);
getWordWithDigraphSuggestionsRec(proximityInfo, xcoordinates, ycoordinates,
codesBuffer, xCoordinatesBuffer, yCoordinatesBuffer, codesBufferSize,
- useFullEditDistance, codesSrc + i, codesRemain - i, currentDepth + 1,
- codesDest + i, correction, queuePool,
- digraphs, digraphsSize);
+ bigramListPosition, useFullEditDistance, codesSrc + i, codesRemain - i,
+ currentDepth + 1, codesDest + i, correction, queuePool, digraphs,
+ digraphsSize);
return;
}
}
@@ -160,14 +161,16 @@ void UnigramDictionary::getWordWithDigraphSuggestionsRec(ProximityInfo *proximit
}
getWordSuggestions(proximityInfo, xCoordinatesBuffer, yCoordinatesBuffer, codesBuffer,
- startIndex + codesRemain, useFullEditDistance, correction,
+ startIndex + codesRemain, bigramListPosition, useFullEditDistance, correction,
queuePool);
}
+// bigramListPosition is the offset in the file to the list of bigrams for the previous word.
int UnigramDictionary::getSuggestions(ProximityInfo *proximityInfo,
WordsPriorityQueuePool *queuePool, Correction *correction, const int *xcoordinates,
const int *ycoordinates, const int *codes, const int codesSize,
- const bool useFullEditDistance, unsigned short *outWords, int *frequencies) {
+ const int bigramListPosition, const bool useFullEditDistance, unsigned short *outWords,
+ int *frequencies) {
queuePool->clearAll();
Correction* masterCorrection = correction;
@@ -177,8 +180,8 @@ int UnigramDictionary::getSuggestions(ProximityInfo *proximityInfo,
int xCoordinatesBuffer[codesSize];
int yCoordinatesBuffer[codesSize];
getWordWithDigraphSuggestionsRec(proximityInfo, xcoordinates, ycoordinates, codesBuffer,
- xCoordinatesBuffer, yCoordinatesBuffer,
- codesSize, useFullEditDistance, codes, codesSize, 0, codesBuffer, masterCorrection,
+ xCoordinatesBuffer, yCoordinatesBuffer, codesSize, bigramListPosition,
+ useFullEditDistance, codes, codesSize, 0, codesBuffer, masterCorrection,
queuePool, GERMAN_UMLAUT_DIGRAPHS,
sizeof(GERMAN_UMLAUT_DIGRAPHS) / sizeof(GERMAN_UMLAUT_DIGRAPHS[0]));
} else if (BinaryFormat::REQUIRES_FRENCH_LIGATURES_PROCESSING & FLAGS) {
@@ -186,13 +189,13 @@ int UnigramDictionary::getSuggestions(ProximityInfo *proximityInfo,
int xCoordinatesBuffer[codesSize];
int yCoordinatesBuffer[codesSize];
getWordWithDigraphSuggestionsRec(proximityInfo, xcoordinates, ycoordinates, codesBuffer,
- xCoordinatesBuffer, yCoordinatesBuffer,
- codesSize, useFullEditDistance, codes, codesSize, 0, codesBuffer, masterCorrection,
+ xCoordinatesBuffer, yCoordinatesBuffer, codesSize, bigramListPosition,
+ useFullEditDistance, codes, codesSize, 0, codesBuffer, masterCorrection,
queuePool, FRENCH_LIGATURES_DIGRAPHS,
sizeof(FRENCH_LIGATURES_DIGRAPHS) / sizeof(FRENCH_LIGATURES_DIGRAPHS[0]));
} else { // Normal processing
getWordSuggestions(proximityInfo, xcoordinates, ycoordinates, codes, codesSize,
- useFullEditDistance, masterCorrection, queuePool);
+ bigramListPosition, useFullEditDistance, masterCorrection, queuePool);
}
PROF_START(20);
@@ -225,16 +228,16 @@ int UnigramDictionary::getSuggestions(ProximityInfo *proximityInfo,
void UnigramDictionary::getWordSuggestions(ProximityInfo *proximityInfo,
const int *xcoordinates, const int *ycoordinates, const int *codes,
- const int inputLength, const bool useFullEditDistance, Correction *correction,
- WordsPriorityQueuePool *queuePool) {
+ const int inputLength, const int bigramListPosition, const bool useFullEditDistance,
+ Correction *correction, WordsPriorityQueuePool *queuePool) {
PROF_OPEN;
PROF_START(0);
PROF_END(0);
PROF_START(1);
- getOneWordSuggestions(proximityInfo, xcoordinates, ycoordinates, codes, useFullEditDistance,
- inputLength, correction, queuePool);
+ getOneWordSuggestions(proximityInfo, xcoordinates, ycoordinates, codes, bigramListPosition,
+ useFullEditDistance, inputLength, correction, queuePool);
PROF_END(1);
PROF_START(2);
@@ -305,15 +308,16 @@ static const char SPACE = ' ';
void UnigramDictionary::getOneWordSuggestions(ProximityInfo *proximityInfo,
const int *xcoordinates, const int *ycoordinates, const int *codes,
- const bool useFullEditDistance, const int inputLength, Correction *correction,
- WordsPriorityQueuePool *queuePool) {
+ const int bigramListPosition, const bool useFullEditDistance, const int inputLength,
+ Correction *correction, WordsPriorityQueuePool *queuePool) {
initSuggestions(proximityInfo, xcoordinates, ycoordinates, codes, inputLength, correction);
- getSuggestionCandidates(useFullEditDistance, inputLength, correction, queuePool,
- true /* doAutoCompletion */, DEFAULT_MAX_ERRORS, FIRST_WORD_INDEX);
+ getSuggestionCandidates(useFullEditDistance, inputLength, bigramListPosition, correction,
+ queuePool, true /* doAutoCompletion */, DEFAULT_MAX_ERRORS, FIRST_WORD_INDEX);
}
void UnigramDictionary::getSuggestionCandidates(const bool useFullEditDistance,
- const int inputLength, Correction *correction, WordsPriorityQueuePool *queuePool,
+ const int inputLength, const int bigramListPosition,
+ Correction *correction, WordsPriorityQueuePool *queuePool,
const bool doAutoCompletion, const int maxErrors, const int currentWordIndex) {
// TODO: Remove setCorrectionParams
correction->setCorrectionParams(0, 0, 0,
@@ -333,8 +337,8 @@ void UnigramDictionary::getSuggestionCandidates(const bool useFullEditDistance,
int firstChildPos;
const bool needsToTraverseChildrenNodes = processCurrentNode(siblingPos,
- correction, &childCount, &firstChildPos, &siblingPos, queuePool,
- currentWordIndex);
+ bigramListPosition, correction, &childCount, &firstChildPos, &siblingPos,
+ queuePool, currentWordIndex);
// Update next sibling pos
correction->setTreeSiblingPos(outputIndex, siblingPos);
@@ -349,7 +353,7 @@ void UnigramDictionary::getSuggestionCandidates(const bool useFullEditDistance,
}
}
-inline void UnigramDictionary::onTerminal(const int freq,
+inline void UnigramDictionary::onTerminal(const int probability,
const TerminalAttributes& terminalAttributes, Correction *correction,
WordsPriorityQueuePool *queuePool, const bool addToMasterQueue,
const int currentWordIndex) {
@@ -361,26 +365,28 @@ inline void UnigramDictionary::onTerminal(const int freq,
if ((currentWordIndex == FIRST_WORD_INDEX) && addToMasterQueue) {
WordsPriorityQueue *masterQueue = queuePool->getMasterQueue();
- const int finalFreq = correction->getFinalFreq(freq, &wordPointer, &wordLength);
- if (finalFreq != NOT_A_FREQUENCY) {
- addWord(wordPointer, wordLength, finalFreq, masterQueue);
+ const int finalProbability =
+ correction->getFinalProbability(probability, &wordPointer, &wordLength);
+ if (finalProbability != NOT_A_PROBABILITY) {
+ addWord(wordPointer, wordLength, finalProbability, masterQueue);
- const int shortcutFreq = finalFreq > 0 ? finalFreq - 1 : 0;
+ const int shortcutProbability = finalProbability > 0 ? finalProbability - 1 : 0;
// Please note that the shortcut candidates will be added to the master queue only.
TerminalAttributes::ShortcutIterator iterator =
terminalAttributes.getShortcutIterator();
while (iterator.hasNextShortcutTarget()) {
// TODO: addWord only supports weak ordering, meaning we have no means
// to control the order of the shortcuts relative to one another or to the word.
- // We need to either modulate the frequency of each shortcut according
- // to its own shortcut frequency or to make the queue
+ // We need to either modulate the probability of each shortcut according
+ // to its own shortcut probability or to make the queue
// so that the insert order is protected inside the queue for words
// with the same score. For the moment we use -1 to make sure the shortcut will
// never be in front of the word.
uint16_t shortcutTarget[MAX_WORD_LENGTH_INTERNAL];
const int shortcutTargetStringLength = iterator.getNextShortcutTarget(
MAX_WORD_LENGTH_INTERNAL, shortcutTarget);
- addWord(shortcutTarget, shortcutTargetStringLength, shortcutFreq, masterQueue);
+ addWord(shortcutTarget, shortcutTargetStringLength, shortcutProbability,
+ masterQueue);
}
}
}
@@ -393,9 +399,9 @@ inline void UnigramDictionary::onTerminal(const int freq,
if (!subQueue) {
return;
}
- const int finalFreq = correction->getFinalFreqForSubQueue(freq, &wordPointer, &wordLength,
- inputIndex);
- addWord(wordPointer, wordLength, finalFreq, subQueue);
+ const int finalProbability = correction->getFinalProbabilityForSubQueue(
+ probability, &wordPointer, &wordLength, inputIndex);
+ addWord(wordPointer, wordLength, finalProbability, subQueue);
}
}
@@ -424,8 +430,10 @@ bool UnigramDictionary::getSubStringSuggestion(
initSuggestions(proximityInfo, &xcoordinates[offset], &ycoordinates[offset],
codes + offset, inputWordLength, correction);
queuePool->clearSubQueue(currentWordIndex);
- getSuggestionCandidates(useFullEditDistance, inputWordLength, correction,
- queuePool, false, MAX_ERRORS_FOR_TWO_WORDS, currentWordIndex);
+ // TODO: pass the bigram list for substring suggestion
+ getSuggestionCandidates(useFullEditDistance, inputWordLength,
+ 0 /* bigramListPosition */, correction, queuePool, false /* doAutoCompletion */,
+ MAX_ERRORS_FOR_TWO_WORDS, currentWordIndex);
if (DEBUG_DICT) {
if (currentWordIndex < MULTIPLE_WORDS_SUGGESTION_MAX_WORDS) {
AKLOGI("Dump word candidates(%d) %d", currentWordIndex, inputWordLength);
@@ -730,7 +738,7 @@ int UnigramDictionary::getMostFrequentWordLikeInner(const uint16_t * const inWor
return maxFreq;
}
-bool UnigramDictionary::isValidWord(const uint16_t* const inWord, const int length) const {
+bool UnigramDictionary::isValidWord(const int32_t* const inWord, const int length) const {
return NOT_VALID_WORD != BinaryFormat::getTerminalPosition(DICT_ROOT, inWord, length);
}
@@ -755,7 +763,7 @@ int UnigramDictionary::getBigramPosition(int pos, unsigned short *word, int offs
// the current node in nextSiblingPosition. Thus, the caller must keep count of the nodes at any
// given level, as output into newCount when traversing this level's parent.
inline bool UnigramDictionary::processCurrentNode(const int initialPos,
- Correction *correction, int *newCount,
+ const int bigramListPosition, Correction *correction, int *newCount,
int *newChildrenPosition, int *nextSiblingPosition, WordsPriorityQueuePool *queuePool,
const int currentWordIndex) {
if (DEBUG_DICT) {
@@ -834,11 +842,14 @@ inline bool UnigramDictionary::processCurrentNode(const int initialPos,
if (isTerminalNode) {
// The frequency should be here, because we come here only if this is actually
// a terminal node, and we are on its last char.
- const int freq = BinaryFormat::readFrequencyWithoutMovingPointer(DICT_ROOT, pos);
+ const int unigramFreq = BinaryFormat::readFrequencyWithoutMovingPointer(DICT_ROOT, pos);
const int childrenAddressPos = BinaryFormat::skipFrequency(flags, pos);
const int attributesPos = BinaryFormat::skipChildrenPosition(flags, childrenAddressPos);
TerminalAttributes terminalAttributes(DICT_ROOT, flags, attributesPos);
- onTerminal(freq, terminalAttributes, correction, queuePool, needsToInvokeOnTerminal,
+ // The bigramListPosition is the offset in the file of the bigrams for the previous word,
+ // or zero if we don't know of any bigrams for it.
+ const int probability = BinaryFormat::getProbability(bigramListPosition, unigramFreq);
+ onTerminal(probability, terminalAttributes, correction, queuePool, needsToInvokeOnTerminal,
currentWordIndex);
// If there are more chars in this node, then this virtual node has children.
diff --git a/native/jni/src/unigram_dictionary.h b/native/jni/src/unigram_dictionary.h
index 4479cd94e..0cc59bac8 100644
--- a/native/jni/src/unigram_dictionary.h
+++ b/native/jni/src/unigram_dictionary.h
@@ -71,37 +71,38 @@ class UnigramDictionary {
UnigramDictionary(const uint8_t* const streamStart, int typedLetterMultipler,
int fullWordMultiplier, int maxWordLength, int maxWords, const unsigned int flags);
- bool isValidWord(const uint16_t* const inWord, const int length) const;
+ bool isValidWord(const int32_t* const inWord, const int length) const;
int getBigramPosition(int pos, unsigned short *word, int offset, int length) const;
int getSuggestions(ProximityInfo *proximityInfo, WordsPriorityQueuePool *queuePool,
- Correction *correction, const int *xcoordinates,
- const int *ycoordinates, const int *codes, const int codesSize,
+ Correction *correction, const int *xcoordinates, const int *ycoordinates,
+ const int *codes, const int codesSize, const int bigramListPosition,
const bool useFullEditDistance, unsigned short *outWords, int *frequencies);
virtual ~UnigramDictionary();
private:
void getWordSuggestions(ProximityInfo *proximityInfo, const int *xcoordinates,
const int *ycoordinates, const int *codes, const int inputLength,
- const bool useFullEditDistance, Correction *correction,
+ const int bigramListPosition, const bool useFullEditDistance, Correction *correction,
WordsPriorityQueuePool *queuePool);
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,
const int *xcoordinates, const int* ycoordinates, const int *codesBuffer,
- int *xCoordinatesBuffer, int *yCoordinatesBuffer,
- const int codesBufferSize, const bool useFullEditDistance, const int* codesSrc,
+ int *xCoordinatesBuffer, int *yCoordinatesBuffer, const int codesBufferSize,
+ const int bigramListPosition, 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);
void initSuggestions(ProximityInfo *proximityInfo, const int *xcoordinates,
const int *ycoordinates, const int *codes, const int codesSize, Correction *correction);
void getOneWordSuggestions(ProximityInfo *proximityInfo, const int *xcoordinates,
- const int *ycoordinates, const int *codes, const bool useFullEditDistance,
- const int inputLength, Correction *correction, WordsPriorityQueuePool* queuePool);
- void getSuggestionCandidates(
+ const int *ycoordinates, const int *codes, const int bigramListPosition,
const bool useFullEditDistance, const int inputLength, Correction *correction,
- WordsPriorityQueuePool* queuePool, const bool doAutoCompletion, const int maxErrors,
- const int currentWordIndex);
+ WordsPriorityQueuePool* queuePool);
+ void getSuggestionCandidates(
+ const bool useFullEditDistance, const int inputLength, const int bigramListPosition,
+ Correction *correction, WordsPriorityQueuePool* queuePool, const bool doAutoCompletion,
+ const int maxErrors, const int currentWordIndex);
void getSplitMultipleWordsSuggestions(ProximityInfo *proximityInfo,
const int *xcoordinates, const int *ycoordinates, const int *codes,
const bool useFullEditDistance, const int inputLength,
@@ -113,9 +114,9 @@ class UnigramDictionary {
bool needsToSkipCurrentNode(const unsigned short c,
const int inputIndex, const int skipPos, const int depth);
// Process a node by considering proximity, missing and excessive character
- bool processCurrentNode(const int initialPos, Correction *correction, int *newCount,
- int *newChildPosition, int *nextSiblingPosition, WordsPriorityQueuePool *queuePool,
- const int currentWordIndex);
+ bool processCurrentNode(const int initialPos, const int bigramListPosition,
+ Correction *correction, int *newCount, int *newChildPosition, int *nextSiblingPosition,
+ WordsPriorityQueuePool *queuePool, const int currentWordIndex);
int getMostFrequentWordLike(const int startInputIndex, const int inputLength,
ProximityInfo *proximityInfo, unsigned short *word);
int getMostFrequentWordLikeInner(const uint16_t* const inWord, const int length,
generated by cgit v1.2.3-83-g751a (git 2.47.1) at 2025-06-04 12:39:22 +0000