mirror of
https://gitlab.futo.org/keyboard/latinime.git
synced 2024-09-28 14:54:30 +01:00
Make native bigram dictionary const
Change-Id: Id883f2d69483c9234877ad42446a582258e1cf91
This commit is contained in:
parent
1bc038c5e4
commit
b1ed1d47ac
@ -27,9 +27,8 @@
|
||||
|
||||
namespace latinime {
|
||||
|
||||
BigramDictionary::BigramDictionary(const unsigned char *dict, int maxWordLength,
|
||||
Dictionary *parentDictionary)
|
||||
: DICT(dict), MAX_WORD_LENGTH(maxWordLength), mParentDictionary(parentDictionary) {
|
||||
BigramDictionary::BigramDictionary(const unsigned char *dict, int maxWordLength)
|
||||
: DICT(dict), MAX_WORD_LENGTH(maxWordLength) {
|
||||
if (DEBUG_DICT) {
|
||||
AKLOGI("BigramDictionary - constructor");
|
||||
}
|
||||
@ -38,7 +37,8 @@ BigramDictionary::BigramDictionary(const unsigned char *dict, int maxWordLength,
|
||||
BigramDictionary::~BigramDictionary() {
|
||||
}
|
||||
|
||||
bool BigramDictionary::addWordBigram(unsigned short *word, int length, int frequency) {
|
||||
bool BigramDictionary::addWordBigram(unsigned short *word, int length, int frequency,
|
||||
const int maxBigrams, int *bigramFreq, unsigned short *bigramChars) const {
|
||||
word[length] = 0;
|
||||
if (DEBUG_DICT) {
|
||||
#ifdef FLAG_DBG
|
||||
@ -50,25 +50,25 @@ bool BigramDictionary::addWordBigram(unsigned short *word, int length, int frequ
|
||||
|
||||
// Find the right insertion point
|
||||
int insertAt = 0;
|
||||
while (insertAt < mMaxBigrams) {
|
||||
if (frequency > mBigramFreq[insertAt] || (mBigramFreq[insertAt] == frequency
|
||||
&& length < Dictionary::wideStrLen(mBigramChars + insertAt * MAX_WORD_LENGTH))) {
|
||||
while (insertAt < maxBigrams) {
|
||||
if (frequency > bigramFreq[insertAt] || (bigramFreq[insertAt] == frequency
|
||||
&& length < Dictionary::wideStrLen(bigramChars + insertAt * MAX_WORD_LENGTH))) {
|
||||
break;
|
||||
}
|
||||
insertAt++;
|
||||
}
|
||||
if (DEBUG_DICT) {
|
||||
AKLOGI("Bigram: InsertAt -> %d maxBigrams: %d", insertAt, mMaxBigrams);
|
||||
AKLOGI("Bigram: InsertAt -> %d maxBigrams: %d", insertAt, maxBigrams);
|
||||
}
|
||||
if (insertAt < mMaxBigrams) {
|
||||
memmove((char*) mBigramFreq + (insertAt + 1) * sizeof(mBigramFreq[0]),
|
||||
(char*) mBigramFreq + insertAt * sizeof(mBigramFreq[0]),
|
||||
(mMaxBigrams - insertAt - 1) * sizeof(mBigramFreq[0]));
|
||||
mBigramFreq[insertAt] = frequency;
|
||||
memmove((char*) mBigramChars + (insertAt + 1) * MAX_WORD_LENGTH * sizeof(short),
|
||||
(char*) mBigramChars + (insertAt ) * MAX_WORD_LENGTH * sizeof(short),
|
||||
(mMaxBigrams - insertAt - 1) * sizeof(short) * MAX_WORD_LENGTH);
|
||||
unsigned short *dest = mBigramChars + (insertAt ) * MAX_WORD_LENGTH;
|
||||
if (insertAt < maxBigrams) {
|
||||
memmove((char*) bigramFreq + (insertAt + 1) * sizeof(bigramFreq[0]),
|
||||
(char*) bigramFreq + insertAt * sizeof(bigramFreq[0]),
|
||||
(maxBigrams - insertAt - 1) * sizeof(bigramFreq[0]));
|
||||
bigramFreq[insertAt] = frequency;
|
||||
memmove((char*) bigramChars + (insertAt + 1) * MAX_WORD_LENGTH * sizeof(short),
|
||||
(char*) bigramChars + (insertAt ) * MAX_WORD_LENGTH * sizeof(short),
|
||||
(maxBigrams - insertAt - 1) * sizeof(short) * MAX_WORD_LENGTH);
|
||||
unsigned short *dest = bigramChars + (insertAt ) * MAX_WORD_LENGTH;
|
||||
while (length--) {
|
||||
*dest++ = *word++;
|
||||
}
|
||||
@ -84,7 +84,7 @@ bool BigramDictionary::addWordBigram(unsigned short *word, int length, int frequ
|
||||
/* Parameters :
|
||||
* prevWord: the word before, the one for which we need to look up bigrams.
|
||||
* prevWordLength: its length.
|
||||
* codes: what user typed, in the same format as for UnigramDictionary::getSuggestions.
|
||||
* inputCodes: what user typed, in the same format as for UnigramDictionary::getSuggestions.
|
||||
* codesSize: the size of the codes array.
|
||||
* bigramChars: an array for output, at the same format as outwords for getSuggestions.
|
||||
* bigramFreq: an array to output frequencies.
|
||||
@ -98,15 +98,11 @@ 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(const int32_t *prevWord, int prevWordLength, int *codes,
|
||||
int BigramDictionary::getBigrams(const int32_t *prevWord, int prevWordLength, int *inputCodes,
|
||||
int codesSize, unsigned short *bigramChars, int *bigramFreq, int maxWordLength,
|
||||
int maxBigrams) {
|
||||
int maxBigrams) const {
|
||||
// TODO: remove unused arguments, and refrain from storing stuff in members of this class
|
||||
// TODO: have "in" arguments before "out" ones, and make out args explicit in the name
|
||||
mBigramFreq = bigramFreq;
|
||||
mBigramChars = bigramChars;
|
||||
mInputCodes = codes;
|
||||
mMaxBigrams = maxBigrams;
|
||||
|
||||
const uint8_t* const root = DICT;
|
||||
int pos = getBigramListPositionForWord(prevWord, prevWordLength);
|
||||
@ -124,16 +120,17 @@ int BigramDictionary::getBigrams(const int32_t *prevWord, int prevWordLength, in
|
||||
bigramBuffer, &unigramFreq);
|
||||
|
||||
// codesSize == 0 means we are trying to find bigram predictions.
|
||||
if (codesSize < 1 || checkFirstCharacter(bigramBuffer)) {
|
||||
const int bigramFreq = UnigramDictionary::MASK_ATTRIBUTE_FREQUENCY & bigramFlags;
|
||||
if (codesSize < 1 || checkFirstCharacter(bigramBuffer, inputCodes)) {
|
||||
const int bigramFreqTemp = UnigramDictionary::MASK_ATTRIBUTE_FREQUENCY & bigramFlags;
|
||||
// Due to space constraints, the frequency for bigrams is approximate - the lower the
|
||||
// unigram frequency, the worse the precision. The theoritical maximum error in
|
||||
// resulting frequency is 8 - although in the practice it's never bigger than 3 or 4
|
||||
// in very bad cases. This means that sometimes, we'll see some bigrams interverted
|
||||
// here, but it can't get too bad.
|
||||
const int frequency =
|
||||
BinaryFormat::computeFrequencyForBigram(unigramFreq, bigramFreq);
|
||||
if (addWordBigram(bigramBuffer, length, frequency)) {
|
||||
BinaryFormat::computeFrequencyForBigram(unigramFreq, bigramFreqTemp);
|
||||
if (addWordBigram(
|
||||
bigramBuffer, length, frequency, maxBigrams, bigramFreq, bigramChars)) {
|
||||
++bigramCount;
|
||||
}
|
||||
}
|
||||
@ -144,7 +141,7 @@ int BigramDictionary::getBigrams(const int32_t *prevWord, int prevWordLength, in
|
||||
// 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::getBigramListPositionForWord(const int32_t *prevWord,
|
||||
const int prevWordLength) {
|
||||
const int prevWordLength) const {
|
||||
if (0 >= prevWordLength) return 0;
|
||||
const uint8_t* const root = DICT;
|
||||
int pos = BinaryFormat::getTerminalPosition(root, prevWord, prevWordLength);
|
||||
@ -164,7 +161,7 @@ int BigramDictionary::getBigramListPositionForWord(const int32_t *prevWord,
|
||||
}
|
||||
|
||||
void BigramDictionary::fillBigramAddressToFrequencyMapAndFilter(const int32_t *prevWord,
|
||||
const int prevWordLength, std::map<int, int> *map, uint8_t *filter) {
|
||||
const int prevWordLength, std::map<int, int> *map, uint8_t *filter) const {
|
||||
memset(filter, 0, BIGRAM_FILTER_BYTE_SIZE);
|
||||
const uint8_t* const root = DICT;
|
||||
int pos = getBigramListPositionForWord(prevWord, prevWordLength);
|
||||
@ -181,11 +178,10 @@ void BigramDictionary::fillBigramAddressToFrequencyMapAndFilter(const int32_t *p
|
||||
} while (0 != (UnigramDictionary::FLAG_ATTRIBUTE_HAS_NEXT & bigramFlags));
|
||||
}
|
||||
|
||||
bool BigramDictionary::checkFirstCharacter(unsigned short *word) {
|
||||
bool BigramDictionary::checkFirstCharacter(unsigned short *word, int *inputCodes) const {
|
||||
// Checks whether this word starts with same character or neighboring characters of
|
||||
// what user typed.
|
||||
|
||||
int *inputCodes = mInputCodes;
|
||||
int maxAlt = MAX_ALTERNATIVES;
|
||||
const unsigned short firstBaseChar = toBaseLowerCase(*word);
|
||||
while (maxAlt > 0) {
|
||||
@ -199,7 +195,7 @@ bool BigramDictionary::checkFirstCharacter(unsigned short *word) {
|
||||
}
|
||||
|
||||
bool BigramDictionary::isValidBigram(const int32_t *word1, int length1, const int32_t *word2,
|
||||
int length2) {
|
||||
int length2) const {
|
||||
const uint8_t* const root = DICT;
|
||||
int pos = getBigramListPositionForWord(word1, length1);
|
||||
// getBigramListPositionForWord returns 0 if this word isn't in the dictionary or has no bigrams
|
||||
|
@ -27,35 +27,29 @@ namespace latinime {
|
||||
class Dictionary;
|
||||
class BigramDictionary {
|
||||
public:
|
||||
BigramDictionary(const unsigned char *dict, int maxWordLength, Dictionary *parentDictionary);
|
||||
int getBigrams(const int32_t *word, int length, int *codes, int codesSize,
|
||||
unsigned short *outWords, int *frequencies, int maxWordLength, int maxBigrams);
|
||||
int getBigramListPositionForWord(const int32_t *prevWord, const int prevWordLength);
|
||||
BigramDictionary(const unsigned char *dict, int maxWordLength);
|
||||
int getBigrams(const int32_t *word, int length, int *inputCodes, int codesSize,
|
||||
unsigned short *outWords, int *frequencies, int maxWordLength, int maxBigrams) const;
|
||||
int getBigramListPositionForWord(const int32_t *prevWord, const int prevWordLength) const;
|
||||
void fillBigramAddressToFrequencyMapAndFilter(const int32_t *prevWord, const int prevWordLength,
|
||||
std::map<int, int> *map, uint8_t *filter);
|
||||
bool isValidBigram(const int32_t *word1, int length1, const int32_t *word2, int length2);
|
||||
std::map<int, int> *map, uint8_t *filter) const;
|
||||
bool isValidBigram(const int32_t *word1, int length1, const int32_t *word2, int length2) const;
|
||||
~BigramDictionary();
|
||||
private:
|
||||
DISALLOW_IMPLICIT_CONSTRUCTORS(BigramDictionary);
|
||||
bool addWordBigram(unsigned short *word, int length, int frequency);
|
||||
bool addWordBigram(unsigned short *word, int length, int frequency, const int maxBigrams,
|
||||
int *bigramFreq, unsigned short *bigramChars) const;
|
||||
int getBigramAddress(int *pos, bool advance);
|
||||
int getBigramFreq(int *pos);
|
||||
void searchForTerminalNode(int addressLookingFor, int frequency);
|
||||
bool getFirstBitOfByte(int *pos) { return (DICT[*pos] & 0x80) > 0; }
|
||||
bool getSecondBitOfByte(int *pos) { return (DICT[*pos] & 0x40) > 0; }
|
||||
bool checkFirstCharacter(unsigned short *word);
|
||||
bool checkFirstCharacter(unsigned short *word, int *inputCodes) const;
|
||||
|
||||
const unsigned char *DICT;
|
||||
const int MAX_WORD_LENGTH;
|
||||
// TODO: Re-implement proximity correction for bigram correction
|
||||
static const int MAX_ALTERNATIVES = 1;
|
||||
|
||||
Dictionary *mParentDictionary;
|
||||
int *mBigramFreq;
|
||||
int mMaxBigrams;
|
||||
unsigned short *mBigramChars;
|
||||
int *mInputCodes;
|
||||
int mInputLength;
|
||||
};
|
||||
|
||||
} // namespace latinime
|
||||
|
@ -38,27 +38,24 @@ Dictionary::Dictionary(void *dict, int dictSize, int mmapFd, int dictBufAdjust,
|
||||
AKLOGI("IN NATIVE SUGGEST Version: %d", (mDict[0] & 0xFF));
|
||||
}
|
||||
}
|
||||
mWordsPriorityQueuePool = new WordsPriorityQueuePool(
|
||||
maxWords, SUB_QUEUE_MAX_WORDS, maxWordLength);
|
||||
const unsigned int headerSize = BinaryFormat::getHeaderSize(mDict);
|
||||
const unsigned int options = BinaryFormat::getFlags(mDict);
|
||||
mUnigramDictionary = new UnigramDictionary(mDict + headerSize, typedLetterMultiplier,
|
||||
fullWordMultiplier, maxWordLength, maxWords, options);
|
||||
mBigramDictionary = new BigramDictionary(mDict + headerSize, maxWordLength, this);
|
||||
mBigramDictionary = new BigramDictionary(mDict + headerSize, maxWordLength);
|
||||
}
|
||||
|
||||
Dictionary::~Dictionary() {
|
||||
delete mWordsPriorityQueuePool;
|
||||
delete mUnigramDictionary;
|
||||
delete mBigramDictionary;
|
||||
}
|
||||
|
||||
int Dictionary::getFrequency(const int32_t *word, int length) {
|
||||
int Dictionary::getFrequency(const int32_t *word, int length) const {
|
||||
return mUnigramDictionary->getFrequency(word, length);
|
||||
}
|
||||
|
||||
bool Dictionary::isValidBigram(const int32_t *word1, int length1, const int32_t *word2,
|
||||
int length2) {
|
||||
int length2) const {
|
||||
return mBigramDictionary->isValidBigram(word1, length1, word2, length2);
|
||||
}
|
||||
|
||||
|
@ -35,28 +35,28 @@ class Dictionary {
|
||||
|
||||
int getSuggestions(ProximityInfo *proximityInfo, int *xcoordinates, int *ycoordinates,
|
||||
int *codes, int codesSize, const int32_t* prevWordChars, const int prevWordLength,
|
||||
bool useFullEditDistance, unsigned short *outWords, int *frequencies) {
|
||||
bool useFullEditDistance, unsigned short *outWords, int *frequencies) const {
|
||||
std::map<int, int> bigramMap;
|
||||
uint8_t bigramFilter[BIGRAM_FILTER_BYTE_SIZE];
|
||||
mBigramDictionary->fillBigramAddressToFrequencyMapAndFilter(prevWordChars,
|
||||
prevWordLength, &bigramMap, bigramFilter);
|
||||
return mUnigramDictionary->getSuggestions(proximityInfo, mWordsPriorityQueuePool,
|
||||
return mUnigramDictionary->getSuggestions(proximityInfo,
|
||||
xcoordinates, ycoordinates, codes, codesSize, &bigramMap,
|
||||
bigramFilter, useFullEditDistance, outWords, frequencies);
|
||||
}
|
||||
|
||||
int getBigrams(const int32_t *word, int length, int *codes, int codesSize,
|
||||
unsigned short *outWords, int *frequencies, int maxWordLength, int maxBigrams) {
|
||||
unsigned short *outWords, int *frequencies, int maxWordLength, int maxBigrams) const {
|
||||
return mBigramDictionary->getBigrams(word, length, codes, codesSize, outWords, frequencies,
|
||||
maxWordLength, maxBigrams);
|
||||
}
|
||||
|
||||
int getFrequency(const int32_t *word, int length);
|
||||
bool isValidBigram(const int32_t *word1, int length1, const int32_t *word2, int length2);
|
||||
void *getDict() { return (void *)mDict; }
|
||||
int getDictSize() { return mDictSize; }
|
||||
int getMmapFd() { return mMmapFd; }
|
||||
int getDictBufAdjust() { return mDictBufAdjust; }
|
||||
int getFrequency(const int32_t *word, int length) const;
|
||||
bool isValidBigram(const int32_t *word1, int length1, const int32_t *word2, int length2) const;
|
||||
void *getDict() const { return (void *)mDict; }
|
||||
int getDictSize() const { return mDictSize; }
|
||||
int getMmapFd() const { return mMmapFd; }
|
||||
int getDictBufAdjust() const { return mDictBufAdjust; }
|
||||
~Dictionary();
|
||||
|
||||
// public static utility methods
|
||||
@ -74,8 +74,7 @@ class Dictionary {
|
||||
const int mDictBufAdjust;
|
||||
|
||||
const UnigramDictionary *mUnigramDictionary;
|
||||
BigramDictionary *mBigramDictionary;
|
||||
WordsPriorityQueuePool *mWordsPriorityQueuePool;
|
||||
const BigramDictionary *mBigramDictionary;
|
||||
};
|
||||
|
||||
// public static utility methods
|
||||
|
@ -170,12 +170,13 @@ void UnigramDictionary::getWordWithDigraphSuggestionsRec(ProximityInfo *proximit
|
||||
// bigramFilter is a bloom filter for fast rejection: see functions setInFilter and isInFilter
|
||||
// in bigram_dictionary.cpp
|
||||
int UnigramDictionary::getSuggestions(ProximityInfo *proximityInfo,
|
||||
WordsPriorityQueuePool *queuePool, const int *xcoordinates,
|
||||
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 {
|
||||
|
||||
queuePool->clearAll();
|
||||
WordsPriorityQueuePool queuePool(MAX_WORDS, SUB_QUEUE_MAX_WORDS, MAX_WORD_LENGTH);
|
||||
queuePool.clearAll();
|
||||
Correction masterCorrection;
|
||||
masterCorrection.resetCorrection();
|
||||
if (BinaryFormat::REQUIRES_GERMAN_UMLAUT_PROCESSING & FLAGS)
|
||||
@ -186,7 +187,7 @@ int UnigramDictionary::getSuggestions(ProximityInfo *proximityInfo,
|
||||
getWordWithDigraphSuggestionsRec(proximityInfo, xcoordinates, ycoordinates, codesBuffer,
|
||||
xCoordinatesBuffer, yCoordinatesBuffer, codesSize, bigramMap, bigramFilter,
|
||||
useFullEditDistance, codes, codesSize, 0, codesBuffer, &masterCorrection,
|
||||
queuePool, GERMAN_UMLAUT_DIGRAPHS,
|
||||
&queuePool, GERMAN_UMLAUT_DIGRAPHS,
|
||||
sizeof(GERMAN_UMLAUT_DIGRAPHS) / sizeof(GERMAN_UMLAUT_DIGRAPHS[0]));
|
||||
} else if (BinaryFormat::REQUIRES_FRENCH_LIGATURES_PROCESSING & FLAGS) {
|
||||
int codesBuffer[getCodesBufferSize(codes, codesSize)];
|
||||
@ -195,26 +196,26 @@ int UnigramDictionary::getSuggestions(ProximityInfo *proximityInfo,
|
||||
getWordWithDigraphSuggestionsRec(proximityInfo, xcoordinates, ycoordinates, codesBuffer,
|
||||
xCoordinatesBuffer, yCoordinatesBuffer, codesSize, bigramMap, bigramFilter,
|
||||
useFullEditDistance, codes, codesSize, 0, codesBuffer, &masterCorrection,
|
||||
queuePool, FRENCH_LIGATURES_DIGRAPHS,
|
||||
&queuePool, FRENCH_LIGATURES_DIGRAPHS,
|
||||
sizeof(FRENCH_LIGATURES_DIGRAPHS) / sizeof(FRENCH_LIGATURES_DIGRAPHS[0]));
|
||||
} else { // Normal processing
|
||||
getWordSuggestions(proximityInfo, xcoordinates, ycoordinates, codes, codesSize,
|
||||
bigramMap, bigramFilter, useFullEditDistance, &masterCorrection, queuePool);
|
||||
bigramMap, bigramFilter, useFullEditDistance, &masterCorrection, &queuePool);
|
||||
}
|
||||
|
||||
PROF_START(20);
|
||||
if (DEBUG_DICT) {
|
||||
float ns = queuePool->getMasterQueue()->getHighestNormalizedScore(
|
||||
float ns = queuePool.getMasterQueue()->getHighestNormalizedScore(
|
||||
masterCorrection.getPrimaryInputWord(), codesSize, 0, 0, 0);
|
||||
ns += 0;
|
||||
AKLOGI("Max normalized score = %f", ns);
|
||||
}
|
||||
const int suggestedWordsCount =
|
||||
queuePool->getMasterQueue()->outputSuggestions(
|
||||
queuePool.getMasterQueue()->outputSuggestions(
|
||||
masterCorrection.getPrimaryInputWord(), codesSize, frequencies, outWords);
|
||||
|
||||
if (DEBUG_DICT) {
|
||||
float ns = queuePool->getMasterQueue()->getHighestNormalizedScore(
|
||||
float ns = queuePool.getMasterQueue()->getHighestNormalizedScore(
|
||||
masterCorrection.getPrimaryInputWord(), codesSize, 0, 0, 0);
|
||||
ns += 0;
|
||||
AKLOGI("Returning %d words", suggestedWordsCount);
|
||||
|
@ -77,8 +77,8 @@ class UnigramDictionary {
|
||||
int fullWordMultiplier, int maxWordLength, int maxWords, const unsigned int flags);
|
||||
int getFrequency(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,
|
||||
const int *xcoordinates, const int *ycoordinates,
|
||||
int getSuggestions(
|
||||
ProximityInfo *proximityInfo, 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;
|
||||
|
Loading…
Reference in New Issue
Block a user