diff options
Diffstat (limited to 'native')
29 files changed, 229 insertions, 181 deletions
diff --git a/native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp b/native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp index 76c7fdd6f..f8dadb488 100644 --- a/native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp +++ b/native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp @@ -28,7 +28,7 @@ #include "suggest/core/dictionary/property/unigram_property.h" #include "suggest/core/dictionary/property/word_property.h" #include "suggest/core/result/suggestion_results.h" -#include "suggest/core/session/prev_words_info.h" +#include "suggest/core/session/ngram_context.h" #include "suggest/core/suggest_options.h" #include "suggest/policyimpl/dictionary/structure/dictionary_structure_with_buffer_policy_factory.h" #include "utils/char_utils.h" @@ -242,15 +242,15 @@ static void latinime_BinaryDictionary_getSuggestions(JNIEnv *env, jclass clazz, env->GetFloatArrayRegion(inOutWeightOfLangModelVsSpatialModel, 0, 1 /* len */, &weightOfLangModelVsSpatialModel); SuggestionResults suggestionResults(MAX_RESULTS); - const PrevWordsInfo prevWordsInfo = JniDataUtils::constructPrevWordsInfo(env, + const NgramContext ngramContext = JniDataUtils::constructNgramContext(env, prevWordCodePointArrays, isBeginningOfSentenceArray, prevWordCount); if (givenSuggestOptions.isGesture() || inputSize > 0) { // TODO: Use SuggestionResults to return suggestions. dictionary->getSuggestions(pInfo, traverseSession, xCoordinates, yCoordinates, - times, pointerIds, inputCodePoints, inputSize, &prevWordsInfo, + times, pointerIds, inputCodePoints, inputSize, &ngramContext, &givenSuggestOptions, weightOfLangModelVsSpatialModel, &suggestionResults); } else { - dictionary->getPredictions(&prevWordsInfo, &suggestionResults); + dictionary->getPredictions(&ngramContext, &suggestionResults); } if (DEBUG_DICT) { suggestionResults.dumpSuggestions(); @@ -289,10 +289,10 @@ static jint latinime_BinaryDictionary_getNgramProbability(JNIEnv *env, jclass cl const jsize wordLength = env->GetArrayLength(word); int wordCodePoints[wordLength]; env->GetIntArrayRegion(word, 0, wordLength, wordCodePoints); - const PrevWordsInfo prevWordsInfo = JniDataUtils::constructPrevWordsInfo(env, + const NgramContext ngramContext = JniDataUtils::constructNgramContext(env, prevWordCodePointArrays, isBeginningOfSentenceArray, env->GetArrayLength(prevWordCodePointArrays)); - return dictionary->getNgramProbability(&prevWordsInfo, + return dictionary->getNgramProbability(&ngramContext, CodePointArrayView(wordCodePoints, wordLength)); } @@ -327,8 +327,9 @@ static jint latinime_BinaryDictionary_getNextWord(JNIEnv *env, jclass clazz, static void latinime_BinaryDictionary_getWordProperty(JNIEnv *env, jclass clazz, jlong dict, jintArray word, jboolean isBeginningOfSentence, jintArray outCodePoints, - jbooleanArray outFlags, jintArray outProbabilityInfo, jobject outBigramTargets, - jobject outBigramProbabilityInfo, jobject outShortcutTargets, + jbooleanArray outFlags, jintArray outProbabilityInfo, jobject /* outNgramPrevWordsArray */, + jobject /* outNgramPrevWordIsBeginningOfSentenceArray */, jobject outNgramTargets, + jobject outNgramProbabilityInfo, jobject outShortcutTargets, jobject outShortcutProbabilities) { Dictionary *dictionary = reinterpret_cast<Dictionary *>(dict); if (!dictionary) return; @@ -351,7 +352,7 @@ static void latinime_BinaryDictionary_getWordProperty(JNIEnv *env, jclass clazz, const WordProperty wordProperty = dictionary->getWordProperty( CodePointArrayView(wordCodePoints, codePointCount)); wordProperty.outputProperties(env, outCodePoints, outFlags, outProbabilityInfo, - outBigramTargets, outBigramProbabilityInfo, outShortcutTargets, + outNgramTargets, outNgramProbabilityInfo, outShortcutTargets, outShortcutProbabilities); } @@ -401,7 +402,7 @@ static bool latinime_BinaryDictionary_addNgramEntry(JNIEnv *env, jclass clazz, j if (!dictionary) { return false; } - const PrevWordsInfo prevWordsInfo = JniDataUtils::constructPrevWordsInfo(env, + const NgramContext ngramContext = JniDataUtils::constructNgramContext(env, prevWordCodePointArrays, isBeginningOfSentenceArray, env->GetArrayLength(prevWordCodePointArrays)); jsize wordLength = env->GetArrayLength(word); @@ -410,7 +411,7 @@ static bool latinime_BinaryDictionary_addNgramEntry(JNIEnv *env, jclass clazz, j // Use 1 for count to indicate the ngram has inputted. const NgramProperty ngramProperty(CodePointArrayView(wordCodePoints, wordLength).toVector(), probability, HistoricalInfo(timestamp, 0 /* level */, 1 /* count */)); - return dictionary->addNgramEntry(&prevWordsInfo, &ngramProperty); + return dictionary->addNgramEntry(&ngramContext, &ngramProperty); } static bool latinime_BinaryDictionary_removeNgramEntry(JNIEnv *env, jclass clazz, jlong dict, @@ -420,31 +421,32 @@ static bool latinime_BinaryDictionary_removeNgramEntry(JNIEnv *env, jclass clazz if (!dictionary) { return false; } - const PrevWordsInfo prevWordsInfo = JniDataUtils::constructPrevWordsInfo(env, + const NgramContext ngramContext = JniDataUtils::constructNgramContext(env, prevWordCodePointArrays, isBeginningOfSentenceArray, env->GetArrayLength(prevWordCodePointArrays)); jsize codePointCount = env->GetArrayLength(word); int wordCodePoints[codePointCount]; env->GetIntArrayRegion(word, 0, codePointCount, wordCodePoints); - return dictionary->removeNgramEntry(&prevWordsInfo, + return dictionary->removeNgramEntry(&ngramContext, CodePointArrayView(wordCodePoints, codePointCount)); } -static bool latinime_BinaryDictionary_updateCounter(JNIEnv *env, jclass clazz, jlong dict, - jobjectArray prevWordCodePointArrays, jbooleanArray isBeginningOfSentenceArray, - jintArray word, jboolean isValidWord, jint count, jint timestamp) { +static bool latinime_BinaryDictionary_updateEntriesForWordWithNgramContext(JNIEnv *env, + jclass clazz, jlong dict, jobjectArray prevWordCodePointArrays, + jbooleanArray isBeginningOfSentenceArray, jintArray word, jboolean isValidWord, jint count, + jint timestamp) { Dictionary *dictionary = reinterpret_cast<Dictionary *>(dict); if (!dictionary) { return false; } - const PrevWordsInfo prevWordsInfo = JniDataUtils::constructPrevWordsInfo(env, + const NgramContext ngramContext = JniDataUtils::constructNgramContext(env, prevWordCodePointArrays, isBeginningOfSentenceArray, env->GetArrayLength(prevWordCodePointArrays)); jsize codePointCount = env->GetArrayLength(word); int wordCodePoints[codePointCount]; env->GetIntArrayRegion(word, 0, codePointCount, wordCodePoints); const HistoricalInfo historicalInfo(timestamp, 0 /* level */, count); - return dictionary->updateCounter(&prevWordsInfo, + return dictionary->updateEntriesForWordWithNgramContext(&ngramContext, CodePointArrayView(wordCodePoints, codePointCount), isValidWord == JNI_TRUE, historicalInfo); } @@ -527,9 +529,9 @@ static int latinime_BinaryDictionary_addMultipleDictionaryEntries(JNIEnv *env, j const NgramProperty ngramProperty( CodePointArrayView(word1CodePoints, word1Length).toVector(), bigramProbability, HistoricalInfo(timestamp, 0 /* level */, 1 /* count */)); - const PrevWordsInfo prevWordsInfo(word0CodePoints, word0Length, + const NgramContext ngramContext(word0CodePoints, word0Length, false /* isBeginningOfSentence */); - dictionary->addNgramEntry(&prevWordsInfo, &ngramProperty); + dictionary->addNgramEntry(&ngramContext, &ngramProperty); } if (dictionary->needsToRunGC(true /* mindsBlockByGC */)) { return i + 1; @@ -639,10 +641,10 @@ static bool latinime_BinaryDictionary_migrateNative(JNIEnv *env, jclass clazz, j return false; } } - const PrevWordsInfo prevWordsInfo(wordCodePoints, wordCodePointCount, + const NgramContext ngramContext(wordCodePoints, wordCodePointCount, wordProperty.getUnigramProperty()->representsBeginningOfSentence()); for (const NgramProperty &ngramProperty : *wordProperty.getNgramProperties()) { - if (!dictionaryStructureWithBufferPolicy->addNgramEntry(&prevWordsInfo, + if (!dictionaryStructureWithBufferPolicy->addNgramEntry(&ngramContext, &ngramProperty)) { LogUtils::logToJava(env, "Cannot add ngram to the new dict."); return false; @@ -718,7 +720,8 @@ static const JNINativeMethod sMethods[] = { { const_cast<char *>("getWordPropertyNative"), const_cast<char *>("(J[IZ[I[Z[ILjava/util/ArrayList;Ljava/util/ArrayList;" - "Ljava/util/ArrayList;Ljava/util/ArrayList;)V"), + "Ljava/util/ArrayList;Ljava/util/ArrayList;Ljava/util/ArrayList;" + "Ljava/util/ArrayList;)V"), reinterpret_cast<void *>(latinime_BinaryDictionary_getWordProperty) }, { @@ -747,9 +750,9 @@ static const JNINativeMethod sMethods[] = { reinterpret_cast<void *>(latinime_BinaryDictionary_removeNgramEntry) }, { - const_cast<char *>("updateCounterNative"), + const_cast<char *>("updateEntriesForWordWithNgramContextNative"), const_cast<char *>("(J[[I[Z[IZII)Z"), - reinterpret_cast<void *>(latinime_BinaryDictionary_updateCounter) + reinterpret_cast<void *>(latinime_BinaryDictionary_updateEntriesForWordWithNgramContext) }, { const_cast<char *>("addMultipleDictionaryEntriesNative"), diff --git a/native/jni/com_android_inputmethod_latin_DicTraverseSession.cpp b/native/jni/com_android_inputmethod_latin_DicTraverseSession.cpp index 766064153..3c6bff3b6 100644 --- a/native/jni/com_android_inputmethod_latin_DicTraverseSession.cpp +++ b/native/jni/com_android_inputmethod_latin_DicTraverseSession.cpp @@ -22,7 +22,7 @@ #include "jni.h" #include "jni_common.h" #include "suggest/core/session/dic_traverse_session.h" -#include "suggest/core/session/prev_words_info.h" +#include "suggest/core/session/ngram_context.h" namespace latinime { class Dictionary; @@ -40,14 +40,14 @@ static void latinime_initDicTraverseSession(JNIEnv *env, jclass clazz, jlong tra } Dictionary *dict = reinterpret_cast<Dictionary *>(dictionary); if (!previousWord) { - PrevWordsInfo prevWordsInfo; - ts->init(dict, &prevWordsInfo, 0 /* suggestOptions */); + NgramContext emptyNgramContext; + ts->init(dict, &emptyNgramContext, 0 /* suggestOptions */); return; } int prevWord[previousWordLength]; env->GetIntArrayRegion(previousWord, 0, previousWordLength, prevWord); - PrevWordsInfo prevWordsInfo(prevWord, previousWordLength, false /* isStartOfSentence */); - ts->init(dict, &prevWordsInfo, 0 /* suggestOptions */); + NgramContext ngramContext(prevWord, previousWordLength, false /* isStartOfSentence */); + ts->init(dict, &ngramContext, 0 /* suggestOptions */); } static void latinime_releaseDicTraverseSession(JNIEnv *env, jclass clazz, jlong traverseSession) { diff --git a/native/jni/src/suggest/core/dictionary/dictionary.cpp b/native/jni/src/suggest/core/dictionary/dictionary.cpp index 7a69d3ceb..697e99ffb 100644 --- a/native/jni/src/suggest/core/dictionary/dictionary.cpp +++ b/native/jni/src/suggest/core/dictionary/dictionary.cpp @@ -23,7 +23,7 @@ #include "suggest/core/policy/dictionary_header_structure_policy.h" #include "suggest/core/result/suggestion_results.h" #include "suggest/core/session/dic_traverse_session.h" -#include "suggest/core/session/prev_words_info.h" +#include "suggest/core/session/ngram_context.h" #include "suggest/core/suggest.h" #include "suggest/core/suggest_options.h" #include "suggest/policyimpl/gesture/gesture_suggest_policy_factory.h" @@ -46,11 +46,11 @@ Dictionary::Dictionary(JNIEnv *env, DictionaryStructureWithBufferPolicy::Structu void Dictionary::getSuggestions(ProximityInfo *proximityInfo, DicTraverseSession *traverseSession, int *xcoordinates, int *ycoordinates, int *times, int *pointerIds, int *inputCodePoints, - int inputSize, const PrevWordsInfo *const prevWordsInfo, + int inputSize, const NgramContext *const ngramContext, const SuggestOptions *const suggestOptions, const float weightOfLangModelVsSpatialModel, SuggestionResults *const outSuggestionResults) const { TimeKeeper::setCurrentTime(); - traverseSession->init(this, prevWordsInfo, suggestOptions); + traverseSession->init(this, ngramContext, suggestOptions); const auto &suggest = suggestOptions->isGesture() ? mGestureSuggest : mTypingSuggest; suggest->getSuggestions(proximityInfo, traverseSession, xcoordinates, ycoordinates, times, pointerIds, inputCodePoints, inputSize, @@ -58,10 +58,10 @@ void Dictionary::getSuggestions(ProximityInfo *proximityInfo, DicTraverseSession } Dictionary::NgramListenerForPrediction::NgramListenerForPrediction( - const PrevWordsInfo *const prevWordsInfo, const WordIdArrayView prevWordIds, + const NgramContext *const ngramContext, const WordIdArrayView prevWordIds, SuggestionResults *const suggestionResults, const DictionaryStructureWithBufferPolicy *const dictStructurePolicy) - : mPrevWordsInfo(prevWordsInfo), mPrevWordIds(prevWordIds), + : mNgramContext(ngramContext), mPrevWordIds(prevWordIds), mSuggestionResults(suggestionResults), mDictStructurePolicy(dictStructurePolicy) {} void Dictionary::NgramListenerForPrediction::onVisitEntry(const int ngramProbability, @@ -69,7 +69,7 @@ void Dictionary::NgramListenerForPrediction::onVisitEntry(const int ngramProbabi if (targetWordId == NOT_A_WORD_ID) { return; } - if (mPrevWordsInfo->isNthPrevWordBeginningOfSentence(1 /* n */) + if (mNgramContext->isNthPrevWordBeginningOfSentence(1 /* n */) && ngramProbability == NOT_A_PROBABILITY) { return; } @@ -85,20 +85,20 @@ void Dictionary::NgramListenerForPrediction::onVisitEntry(const int ngramProbabi wordAttributes.getProbability()); } -void Dictionary::getPredictions(const PrevWordsInfo *const prevWordsInfo, +void Dictionary::getPredictions(const NgramContext *const ngramContext, SuggestionResults *const outSuggestionResults) const { TimeKeeper::setCurrentTime(); WordIdArray<MAX_PREV_WORD_COUNT_FOR_N_GRAM> prevWordIdArray; - const WordIdArrayView prevWordIds = prevWordsInfo->getPrevWordIds( + const WordIdArrayView prevWordIds = ngramContext->getPrevWordIds( mDictionaryStructureWithBufferPolicy.get(), &prevWordIdArray, true /* tryLowerCaseSearch */); - NgramListenerForPrediction listener(prevWordsInfo, prevWordIds, outSuggestionResults, + NgramListenerForPrediction listener(ngramContext, prevWordIds, outSuggestionResults, mDictionaryStructureWithBufferPolicy.get()); mDictionaryStructureWithBufferPolicy->iterateNgramEntries(prevWordIds, &listener); } int Dictionary::getProbability(const CodePointArrayView codePoints) const { - return getNgramProbability(nullptr /* prevWordsInfo */, codePoints); + return getNgramProbability(nullptr /* ngramContext */, codePoints); } int Dictionary::getMaxProbabilityOfExactMatches(const CodePointArrayView codePoints) const { @@ -107,18 +107,18 @@ int Dictionary::getMaxProbabilityOfExactMatches(const CodePointArrayView codePoi mDictionaryStructureWithBufferPolicy.get(), codePoints); } -int Dictionary::getNgramProbability(const PrevWordsInfo *const prevWordsInfo, +int Dictionary::getNgramProbability(const NgramContext *const ngramContext, const CodePointArrayView codePoints) const { TimeKeeper::setCurrentTime(); const int wordId = mDictionaryStructureWithBufferPolicy->getWordId(codePoints, false /* forceLowerCaseSearch */); if (wordId == NOT_A_WORD_ID) return NOT_A_PROBABILITY; - if (!prevWordsInfo) { + if (!ngramContext) { return getDictionaryStructurePolicy()->getProbabilityOfWord(WordIdArrayView(), wordId); } WordIdArray<MAX_PREV_WORD_COUNT_FOR_N_GRAM> prevWordIdArray; - const WordIdArrayView prevWordIds = prevWordsInfo->getPrevWordIds - (mDictionaryStructureWithBufferPolicy.get(), &prevWordIdArray, + const WordIdArrayView prevWordIds = ngramContext->getPrevWordIds( + mDictionaryStructureWithBufferPolicy.get(), &prevWordIdArray, true /* tryLowerCaseSearch */); return getDictionaryStructurePolicy()->getProbabilityOfWord(prevWordIds, wordId); } @@ -140,24 +140,24 @@ bool Dictionary::removeUnigramEntry(const CodePointArrayView codePoints) { return mDictionaryStructureWithBufferPolicy->removeUnigramEntry(codePoints); } -bool Dictionary::addNgramEntry(const PrevWordsInfo *const prevWordsInfo, +bool Dictionary::addNgramEntry(const NgramContext *const ngramContext, const NgramProperty *const ngramProperty) { TimeKeeper::setCurrentTime(); - return mDictionaryStructureWithBufferPolicy->addNgramEntry(prevWordsInfo, ngramProperty); + return mDictionaryStructureWithBufferPolicy->addNgramEntry(ngramContext, ngramProperty); } -bool Dictionary::removeNgramEntry(const PrevWordsInfo *const prevWordsInfo, +bool Dictionary::removeNgramEntry(const NgramContext *const ngramContext, const CodePointArrayView codePoints) { TimeKeeper::setCurrentTime(); - return mDictionaryStructureWithBufferPolicy->removeNgramEntry(prevWordsInfo, codePoints); + return mDictionaryStructureWithBufferPolicy->removeNgramEntry(ngramContext, codePoints); } -bool Dictionary::updateCounter(const PrevWordsInfo *const prevWordsInfo, +bool Dictionary::updateEntriesForWordWithNgramContext(const NgramContext *const ngramContext, const CodePointArrayView codePoints, const bool isValidWord, const HistoricalInfo historicalInfo) { TimeKeeper::setCurrentTime(); - return mDictionaryStructureWithBufferPolicy->updateCounter(prevWordsInfo, codePoints, - isValidWord, historicalInfo); + return mDictionaryStructureWithBufferPolicy->updateEntriesForWordWithNgramContext(ngramContext, + codePoints, isValidWord, historicalInfo); } bool Dictionary::flush(const char *const filePath) { diff --git a/native/jni/src/suggest/core/dictionary/dictionary.h b/native/jni/src/suggest/core/dictionary/dictionary.h index a58dbfbd7..843aec473 100644 --- a/native/jni/src/suggest/core/dictionary/dictionary.h +++ b/native/jni/src/suggest/core/dictionary/dictionary.h @@ -33,7 +33,7 @@ namespace latinime { class DictionaryStructureWithBufferPolicy; class DicTraverseSession; -class PrevWordsInfo; +class NgramContext; class ProximityInfo; class SuggestionResults; class SuggestOptions; @@ -66,18 +66,18 @@ class Dictionary { void getSuggestions(ProximityInfo *proximityInfo, DicTraverseSession *traverseSession, int *xcoordinates, int *ycoordinates, int *times, int *pointerIds, int *inputCodePoints, - int inputSize, const PrevWordsInfo *const prevWordsInfo, + int inputSize, const NgramContext *const ngramContext, const SuggestOptions *const suggestOptions, const float weightOfLangModelVsSpatialModel, SuggestionResults *const outSuggestionResults) const; - void getPredictions(const PrevWordsInfo *const prevWordsInfo, + void getPredictions(const NgramContext *const ngramContext, SuggestionResults *const outSuggestionResults) const; int getProbability(const CodePointArrayView codePoints) const; int getMaxProbabilityOfExactMatches(const CodePointArrayView codePoints) const; - int getNgramProbability(const PrevWordsInfo *const prevWordsInfo, + int getNgramProbability(const NgramContext *const ngramContext, const CodePointArrayView codePoints) const; bool addUnigramEntry(const CodePointArrayView codePoints, @@ -85,13 +85,13 @@ class Dictionary { bool removeUnigramEntry(const CodePointArrayView codePoints); - bool addNgramEntry(const PrevWordsInfo *const prevWordsInfo, + bool addNgramEntry(const NgramContext *const ngramContext, const NgramProperty *const ngramProperty); - bool removeNgramEntry(const PrevWordsInfo *const prevWordsInfo, + bool removeNgramEntry(const NgramContext *const ngramContext, const CodePointArrayView codePoints); - bool updateCounter(const PrevWordsInfo *const prevWordsInfo, + bool updateEntriesForWordWithNgramContext(const NgramContext *const ngramContext, const CodePointArrayView codePoints, const bool isValidWord, const HistoricalInfo historicalInfo); @@ -123,7 +123,7 @@ class Dictionary { class NgramListenerForPrediction : public NgramListener { public: - NgramListenerForPrediction(const PrevWordsInfo *const prevWordsInfo, + NgramListenerForPrediction(const NgramContext *const ngramContext, const WordIdArrayView prevWordIds, SuggestionResults *const suggestionResults, const DictionaryStructureWithBufferPolicy *const dictStructurePolicy); virtual void onVisitEntry(const int ngramProbability, const int targetWordId); @@ -131,7 +131,7 @@ class Dictionary { private: DISALLOW_IMPLICIT_CONSTRUCTORS(NgramListenerForPrediction); - const PrevWordsInfo *const mPrevWordsInfo; + const NgramContext *const mNgramContext; const WordIdArrayView mPrevWordIds; SuggestionResults *const mSuggestionResults; const DictionaryStructureWithBufferPolicy *const mDictStructurePolicy; diff --git a/native/jni/src/suggest/core/dictionary/dictionary_utils.cpp b/native/jni/src/suggest/core/dictionary/dictionary_utils.cpp index b85f3622a..9573c37bc 100644 --- a/native/jni/src/suggest/core/dictionary/dictionary_utils.cpp +++ b/native/jni/src/suggest/core/dictionary/dictionary_utils.cpp @@ -21,7 +21,7 @@ #include "suggest/core/dicnode/dic_node_vector.h" #include "suggest/core/dictionary/dictionary.h" #include "suggest/core/dictionary/digraph_utils.h" -#include "suggest/core/session/prev_words_info.h" +#include "suggest/core/session/ngram_context.h" #include "suggest/core/policy/dictionary_structure_with_buffer_policy.h" #include "utils/int_array_view.h" @@ -33,10 +33,10 @@ namespace latinime { std::vector<DicNode> current; std::vector<DicNode> next; - // No prev words information. - PrevWordsInfo emptyPrevWordsInfo; + // No ngram context. + NgramContext emptyNgramContext; WordIdArray<MAX_PREV_WORD_COUNT_FOR_N_GRAM> prevWordIdArray; - const WordIdArrayView prevWordIds = emptyPrevWordsInfo.getPrevWordIds( + const WordIdArrayView prevWordIds = emptyNgramContext.getPrevWordIds( dictionaryStructurePolicy, &prevWordIdArray, false /* tryLowerCaseSearch */); current.emplace_back(); DicNodeUtils::initAsRoot(dictionaryStructurePolicy, prevWordIds, ¤t.front()); diff --git a/native/jni/src/suggest/core/dictionary/property/historical_info.h b/native/jni/src/suggest/core/dictionary/property/historical_info.h index 5ed9ebfca..f9bd6fd8c 100644 --- a/native/jni/src/suggest/core/dictionary/property/historical_info.h +++ b/native/jni/src/suggest/core/dictionary/property/historical_info.h @@ -47,12 +47,12 @@ class HistoricalInfo { } private: - // Default copy constructor and assign operator are used for using in std::vector. + // Default copy constructor is used for using in std::vector. + DISALLOW_ASSIGNMENT_OPERATOR(HistoricalInfo); - // TODO: Make members const. - int mTimestamp; - int mLevel; - int mCount; + const int mTimestamp; + const int mLevel; + const int mCount; }; } // namespace latinime #endif /* LATINIME_HISTORICAL_INFO_H */ diff --git a/native/jni/src/suggest/core/dictionary/property/ngram_property.h b/native/jni/src/suggest/core/dictionary/property/ngram_property.h index dce460099..8709799f9 100644 --- a/native/jni/src/suggest/core/dictionary/property/ngram_property.h +++ b/native/jni/src/suggest/core/dictionary/property/ngram_property.h @@ -44,13 +44,13 @@ class NgramProperty { } private: - // Default copy constructor and assign operator are used for using in std::vector. + // Default copy constructor is used for using in std::vector. DISALLOW_DEFAULT_CONSTRUCTOR(NgramProperty); + DISALLOW_ASSIGNMENT_OPERATOR(NgramProperty); - // TODO: Make members const. - std::vector<int> mTargetCodePoints; - int mProbability; - HistoricalInfo mHistoricalInfo; + const std::vector<int> mTargetCodePoints; + const int mProbability; + const HistoricalInfo mHistoricalInfo; }; } // namespace latinime #endif // LATINIME_NGRAM_PROPERTY_H diff --git a/native/jni/src/suggest/core/dictionary/property/unigram_property.h b/native/jni/src/suggest/core/dictionary/property/unigram_property.h index d1f0ab4ca..5ed2e2602 100644 --- a/native/jni/src/suggest/core/dictionary/property/unigram_property.h +++ b/native/jni/src/suggest/core/dictionary/property/unigram_property.h @@ -41,12 +41,11 @@ class UnigramProperty { } private: - // Default copy constructor and assign operator are used for using in std::vector. + // Default copy constructor is used for using in std::vector. DISALLOW_DEFAULT_CONSTRUCTOR(ShortcutProperty); - // TODO: Make members const. - std::vector<int> mTargetCodePoints; - int mProbability; + const std::vector<int> mTargetCodePoints; + const int mProbability; }; UnigramProperty() @@ -104,13 +103,12 @@ class UnigramProperty { // Default copy constructor is used for using as a return value. DISALLOW_ASSIGNMENT_OPERATOR(UnigramProperty); - // TODO: Make members const. - bool mRepresentsBeginningOfSentence; - bool mIsNotAWord; - bool mIsBlacklisted; - int mProbability; - HistoricalInfo mHistoricalInfo; - std::vector<ShortcutProperty> mShortcuts; + const bool mRepresentsBeginningOfSentence; + const bool mIsNotAWord; + const bool mIsBlacklisted; + const int mProbability; + const HistoricalInfo mHistoricalInfo; + const std::vector<ShortcutProperty> mShortcuts; }; } // namespace latinime #endif // LATINIME_UNIGRAM_PROPERTY_H diff --git a/native/jni/src/suggest/core/policy/dictionary_structure_with_buffer_policy.h b/native/jni/src/suggest/core/policy/dictionary_structure_with_buffer_policy.h index 6624b7921..ceda5c03f 100644 --- a/native/jni/src/suggest/core/policy/dictionary_structure_with_buffer_policy.h +++ b/native/jni/src/suggest/core/policy/dictionary_structure_with_buffer_policy.h @@ -33,7 +33,7 @@ class DicNodeVector; class DictionaryHeaderStructurePolicy; class MultiBigramMap; class NgramListener; -class PrevWordsInfo; +class NgramContext; class UnigramProperty; /* @@ -81,15 +81,15 @@ class DictionaryStructureWithBufferPolicy { virtual bool removeUnigramEntry(const CodePointArrayView wordCodePoints) = 0; // Returns whether the update was success or not. - virtual bool addNgramEntry(const PrevWordsInfo *const prevWordsInfo, + virtual bool addNgramEntry(const NgramContext *const ngramContext, const NgramProperty *const ngramProperty) = 0; // Returns whether the update was success or not. - virtual bool removeNgramEntry(const PrevWordsInfo *const prevWordsInfo, + virtual bool removeNgramEntry(const NgramContext *const ngramContext, const CodePointArrayView wordCodePoints) = 0; // Returns whether the update was success or not. - virtual bool updateCounter(const PrevWordsInfo *const prevWordsInfo, + virtual bool updateEntriesForWordWithNgramContext(const NgramContext *const ngramContext, const CodePointArrayView wordCodePoints, const bool isValidWord, const HistoricalInfo historicalInfo) = 0; diff --git a/native/jni/src/suggest/core/policy/traversal.h b/native/jni/src/suggest/core/policy/traversal.h index 6dfa7e314..5b6616d9a 100644 --- a/native/jni/src/suggest/core/policy/traversal.h +++ b/native/jni/src/suggest/core/policy/traversal.h @@ -44,7 +44,7 @@ class Traversal { virtual bool needsToTraverseAllUserInput() const = 0; virtual float getMaxSpatialDistance() const = 0; virtual int getDefaultExpandDicNodeSize() const = 0; - virtual int getMaxCacheSize(const int inputSize) const = 0; + virtual int getMaxCacheSize(const int inputSize, const float weightForLocale) const = 0; virtual int getTerminalCacheSize() const = 0; virtual bool isPossibleOmissionChildNode(const DicTraverseSession *const traverseSession, const DicNode *const parentDicNode, const DicNode *const dicNode) const = 0; diff --git a/native/jni/src/suggest/core/session/dic_traverse_session.cpp b/native/jni/src/suggest/core/session/dic_traverse_session.cpp index b4d01d0f0..52dc2f86c 100644 --- a/native/jni/src/suggest/core/session/dic_traverse_session.cpp +++ b/native/jni/src/suggest/core/session/dic_traverse_session.cpp @@ -20,7 +20,7 @@ #include "suggest/core/dictionary/dictionary.h" #include "suggest/core/policy/dictionary_header_structure_policy.h" #include "suggest/core/policy/dictionary_structure_with_buffer_policy.h" -#include "suggest/core/session/prev_words_info.h" +#include "suggest/core/session/ngram_context.h" namespace latinime { @@ -30,12 +30,12 @@ const int DicTraverseSession::DICTIONARY_SIZE_THRESHOLD_TO_USE_LARGE_CACHE_FOR_S 256 * 1024; void DicTraverseSession::init(const Dictionary *const dictionary, - const PrevWordsInfo *const prevWordsInfo, const SuggestOptions *const suggestOptions) { + const NgramContext *const ngramContext, const SuggestOptions *const suggestOptions) { mDictionary = dictionary; mMultiWordCostMultiplier = getDictionaryStructurePolicy()->getHeaderStructurePolicy() ->getMultiWordCostMultiplier(); mSuggestOptions = suggestOptions; - mPrevWordIdCount = prevWordsInfo->getPrevWordIds(getDictionaryStructurePolicy(), + mPrevWordIdCount = ngramContext->getPrevWordIds(getDictionaryStructurePolicy(), &mPrevWordIdArray, true /* tryLowerCaseSearch */).size(); } diff --git a/native/jni/src/suggest/core/session/dic_traverse_session.h b/native/jni/src/suggest/core/session/dic_traverse_session.h index 9f841aa3c..bc53167f0 100644 --- a/native/jni/src/suggest/core/session/dic_traverse_session.h +++ b/native/jni/src/suggest/core/session/dic_traverse_session.h @@ -30,7 +30,7 @@ namespace latinime { class Dictionary; class DictionaryStructureWithBufferPolicy; -class PrevWordsInfo; +class NgramContext; class ProximityInfo; class SuggestOptions; @@ -61,7 +61,7 @@ class DicTraverseSession { // Non virtual inline destructor -- never inherit this class AK_FORCE_INLINE ~DicTraverseSession() {} - void init(const Dictionary *dictionary, const PrevWordsInfo *const prevWordsInfo, + void init(const Dictionary *dictionary, const NgramContext *const ngramContext, const SuggestOptions *const suggestOptions); // TODO: Remove and merge into init void setupForGetSuggestions(const ProximityInfo *pInfo, const int *inputCodePoints, diff --git a/native/jni/src/suggest/core/session/prev_words_info.h b/native/jni/src/suggest/core/session/ngram_context.h index 553d5ad07..64c71410f 100644 --- a/native/jni/src/suggest/core/session/prev_words_info.h +++ b/native/jni/src/suggest/core/session/ngram_context.h @@ -14,8 +14,8 @@ * limitations under the License. */ -#ifndef LATINIME_PREV_WORDS_INFO_H -#define LATINIME_PREV_WORDS_INFO_H +#ifndef LATINIME_NGRAM_CONTEXT_H +#define LATINIME_NGRAM_CONTEXT_H #include <array> @@ -26,25 +26,26 @@ namespace latinime { -class PrevWordsInfo { +// Rename to NgramContext. +class NgramContext { public: // No prev word information. - PrevWordsInfo() : mPrevWordCount(0) { + NgramContext() : mPrevWordCount(0) { clear(); } - PrevWordsInfo(const PrevWordsInfo &prevWordsInfo) - : mPrevWordCount(prevWordsInfo.mPrevWordCount) { + NgramContext(const NgramContext &ngramContext) + : mPrevWordCount(ngramContext.mPrevWordCount) { for (size_t i = 0; i < mPrevWordCount; ++i) { - mPrevWordCodePointCount[i] = prevWordsInfo.mPrevWordCodePointCount[i]; - memmove(mPrevWordCodePoints[i], prevWordsInfo.mPrevWordCodePoints[i], + mPrevWordCodePointCount[i] = ngramContext.mPrevWordCodePointCount[i]; + memmove(mPrevWordCodePoints[i], ngramContext.mPrevWordCodePoints[i], sizeof(mPrevWordCodePoints[i][0]) * mPrevWordCodePointCount[i]); - mIsBeginningOfSentence[i] = prevWordsInfo.mIsBeginningOfSentence[i]; + mIsBeginningOfSentence[i] = ngramContext.mIsBeginningOfSentence[i]; } } // Construct from previous words. - PrevWordsInfo(const int prevWordCodePoints[][MAX_WORD_LENGTH], + NgramContext(const int prevWordCodePoints[][MAX_WORD_LENGTH], const int *const prevWordCodePointCount, const bool *const isBeginningOfSentence, const size_t prevWordCount) : mPrevWordCount(std::min(NELEMS(mPrevWordCodePoints), prevWordCount)) { @@ -61,7 +62,7 @@ class PrevWordsInfo { } // Construct from a previous word. - PrevWordsInfo(const int *const prevWordCodePoints, const int prevWordCodePointCount, + NgramContext(const int *const prevWordCodePoints, const int prevWordCodePointCount, const bool isBeginningOfSentence) : mPrevWordCount(1) { clear(); if (prevWordCodePointCount > MAX_WORD_LENGTH || !prevWordCodePoints) { @@ -78,8 +79,8 @@ class PrevWordsInfo { } // TODO: Remove. - const PrevWordsInfo getTrimmedPrevWordsInfo(const size_t maxPrevWordCount) const { - return PrevWordsInfo(mPrevWordCodePoints, mPrevWordCodePointCount, mIsBeginningOfSentence, + const NgramContext getTrimmedNgramContext(const size_t maxPrevWordCount) const { + return NgramContext(mPrevWordCodePoints, mPrevWordCodePointCount, mIsBeginningOfSentence, std::min(mPrevWordCount, maxPrevWordCount)); } @@ -122,7 +123,7 @@ class PrevWordsInfo { } private: - DISALLOW_ASSIGNMENT_OPERATOR(PrevWordsInfo); + DISALLOW_ASSIGNMENT_OPERATOR(NgramContext); static int getWordId(const DictionaryStructureWithBufferPolicy *const dictStructurePolicy, const int *const wordCodePoints, const int wordCodePointCount, @@ -165,4 +166,4 @@ class PrevWordsInfo { bool mIsBeginningOfSentence[MAX_PREV_WORD_COUNT_FOR_N_GRAM]; }; } // namespace latinime -#endif // LATINIME_PREV_WORDS_INFO_H +#endif // LATINIME_NGRAM_CONTEXT_H diff --git a/native/jni/src/suggest/core/suggest.cpp b/native/jni/src/suggest/core/suggest.cpp index 457414f2b..cf2df86a6 100644 --- a/native/jni/src/suggest/core/suggest.cpp +++ b/native/jni/src/suggest/core/suggest.cpp @@ -28,6 +28,7 @@ #include "suggest/core/policy/weighting.h" #include "suggest/core/result/suggestions_output_utils.h" #include "suggest/core/session/dic_traverse_session.h" +#include "suggest/core/suggest_options.h" namespace latinime { @@ -88,7 +89,8 @@ void Suggest::initializeSearch(DicTraverseSession *traverseSession) const { traverseSession->getDicTraverseCache()->continueSearch(); } else { // Restart recognition at the root. - traverseSession->resetCache(TRAVERSAL->getMaxCacheSize(traverseSession->getInputSize()), + traverseSession->resetCache(TRAVERSAL->getMaxCacheSize(traverseSession->getInputSize(), + traverseSession->getSuggestOptions()->weightForLocale()), TRAVERSAL->getTerminalCacheSize()); // Create a new dic node here DicNode rootNode; diff --git a/native/jni/src/suggest/core/suggest_options.h b/native/jni/src/suggest/core/suggest_options.h index d456680dd..4d331292b 100644 --- a/native/jni/src/suggest/core/suggest_options.h +++ b/native/jni/src/suggest/core/suggest_options.h @@ -42,6 +42,12 @@ class SuggestOptions{ return getBoolOption(SPACE_AWARE_GESTURE_ENABLED); } + AK_FORCE_INLINE float weightForLocale() const { + // The weight is in thousands and we want the real value, so we divide by 1000. + // NativeSuggestOptions#setWeightForLocale does the opposite processing in Java. + return static_cast<float>(getIntOption(WEIGHT_FOR_LOCALE_IN_THOUSANDS)) / 1000.0f; + } + AK_FORCE_INLINE bool getAdditionalFeaturesBoolOption(const int key) const { return getBoolOption(key + ADDITIONAL_FEATURES_OPTIONS); } @@ -55,9 +61,10 @@ class SuggestOptions{ static const int USE_FULL_EDIT_DISTANCE = 1; static const int BLOCK_OFFENSIVE_WORDS = 2; static const int SPACE_AWARE_GESTURE_ENABLED = 3; + static const int WEIGHT_FOR_LOCALE_IN_THOUSANDS = 4; // Additional features options are stored after the other options and used as setting values of // experimental features. - static const int ADDITIONAL_FEATURES_OPTIONS = 4; + static const int ADDITIONAL_FEATURES_OPTIONS = 5; const int *const mOptions; const int mLength; diff --git a/native/jni/src/suggest/policyimpl/dictionary/structure/backward/v402/ver4_patricia_trie_node_writer.cpp b/native/jni/src/suggest/policyimpl/dictionary/structure/backward/v402/ver4_patricia_trie_node_writer.cpp index 8d169743c..6243f14cc 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/structure/backward/v402/ver4_patricia_trie_node_writer.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/backward/v402/ver4_patricia_trie_node_writer.cpp @@ -310,7 +310,7 @@ bool Ver4PatriciaTrieNodeWriter::addShortcutTarget(const PtNodeParams *const ptN const int shortcutProbability) { if (!mShortcutPolicy->addNewShortcut(ptNodeParams->getTerminalId(), targetCodePoints, targetCodePointCount, shortcutProbability)) { - AKLOGE("Cannot add new shortuct entry. terminalId: %d", ptNodeParams->getTerminalId()); + AKLOGE("Cannot add new shortcut entry. terminalId: %d", ptNodeParams->getTerminalId()); return false; } if (!ptNodeParams->hasShortcutTargets()) { diff --git a/native/jni/src/suggest/policyimpl/dictionary/structure/backward/v402/ver4_patricia_trie_policy.cpp b/native/jni/src/suggest/policyimpl/dictionary/structure/backward/v402/ver4_patricia_trie_policy.cpp index 36eafa1e9..0eae934ae 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/structure/backward/v402/ver4_patricia_trie_policy.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/backward/v402/ver4_patricia_trie_policy.cpp @@ -33,7 +33,7 @@ #include "suggest/core/dictionary/property/ngram_property.h" #include "suggest/core/dictionary/property/unigram_property.h" #include "suggest/core/dictionary/property/word_property.h" -#include "suggest/core/session/prev_words_info.h" +#include "suggest/core/session/ngram_context.h" #include "suggest/policyimpl/dictionary/structure/pt_common/dynamic_pt_reading_helper.h" #include "suggest/policyimpl/dictionary/structure/backward/v402/ver4_patricia_trie_node_reader.h" #include "suggest/policyimpl/dictionary/utils/forgetting_curve_utils.h" @@ -186,7 +186,9 @@ int Ver4PatriciaTriePolicy::getProbabilityOfWord(const WordIdArrayView prevWordI if (bigramsIt.getBigramPos() == ptNodePos && bigramsIt.getProbability() != NOT_A_PROBABILITY) { const int bigramConditionalProbability = getBigramConditionalProbability( - prevWordPtNodeParams.getProbability(), bigramsIt.getProbability()); + prevWordPtNodeParams.getProbability(), + prevWordPtNodeParams.representsBeginningOfSentence(), + bigramsIt.getProbability()); return getProbability(ptNodeParams.getProbability(), bigramConditionalProbability); } } @@ -209,15 +211,19 @@ void Ver4PatriciaTriePolicy::iterateNgramEntries(const WordIdArrayView prevWordI while (bigramsIt.hasNext()) { bigramsIt.next(); const int bigramConditionalProbability = getBigramConditionalProbability( - prevWordPtNodeParams.getProbability(), bigramsIt.getProbability()); + prevWordPtNodeParams.getProbability(), + prevWordPtNodeParams.representsBeginningOfSentence(), bigramsIt.getProbability()); listener->onVisitEntry(bigramConditionalProbability, getWordIdFromTerminalPtNodePos(bigramsIt.getBigramPos())); } } int Ver4PatriciaTriePolicy::getBigramConditionalProbability(const int prevWordUnigramProbability, - const int bigramProbability) const { + const bool isInBeginningOfSentenceContext, const int bigramProbability) const { if (mHeaderPolicy->hasHistoricalInfoOfWords()) { + if (isInBeginningOfSentenceContext) { + return bigramProbability; + } // Calculate conditional probability. return std::min(MAX_PROBABILITY - prevWordUnigramProbability + bigramProbability, MAX_PROBABILITY); @@ -338,7 +344,7 @@ bool Ver4PatriciaTriePolicy::removeUnigramEntry(const CodePointArrayView wordCod return mNodeWriter.suppressUnigramEntry(&ptNodeParams); } -bool Ver4PatriciaTriePolicy::addNgramEntry(const PrevWordsInfo *const prevWordsInfo, +bool Ver4PatriciaTriePolicy::addNgramEntry(const NgramContext *const ngramContext, const NgramProperty *const ngramProperty) { if (!mBuffers->isUpdatable()) { AKLOGI("Warning: addNgramEntry() is called for non-updatable dictionary."); @@ -349,8 +355,8 @@ bool Ver4PatriciaTriePolicy::addNgramEntry(const PrevWordsInfo *const prevWordsI mDictBuffer->getTailPosition()); return false; } - if (!prevWordsInfo->isValid()) { - AKLOGE("prev words info is not valid for adding n-gram entry to the dictionary."); + if (!ngramContext->isValid()) { + AKLOGE("Ngram context is not valid for adding n-gram entry to the dictionary."); return false; } if (ngramProperty->getTargetCodePoints()->size() > MAX_WORD_LENGTH) { @@ -359,23 +365,23 @@ bool Ver4PatriciaTriePolicy::addNgramEntry(const PrevWordsInfo *const prevWordsI return false; } WordIdArray<MAX_PREV_WORD_COUNT_FOR_N_GRAM> prevWordIdArray; - const WordIdArrayView prevWordIds = prevWordsInfo->getPrevWordIds(this, &prevWordIdArray, + const WordIdArrayView prevWordIds = ngramContext->getPrevWordIds(this, &prevWordIdArray, false /* tryLowerCaseSearch */); if (prevWordIds.empty()) { return false; } if (prevWordIds[0] == NOT_A_WORD_ID) { - if (prevWordsInfo->isNthPrevWordBeginningOfSentence(1 /* n */)) { + if (ngramContext->isNthPrevWordBeginningOfSentence(1 /* n */)) { const UnigramProperty beginningOfSentenceUnigramProperty( true /* representsBeginningOfSentence */, true /* isNotAWord */, false /* isBlacklisted */, MAX_PROBABILITY /* probability */, HistoricalInfo()); - if (!addUnigramEntry(prevWordsInfo->getNthPrevWordCodePoints(1 /* n */), + if (!addUnigramEntry(ngramContext->getNthPrevWordCodePoints(1 /* n */), &beginningOfSentenceUnigramProperty)) { AKLOGE("Cannot add unigram entry for the beginning-of-sentence."); return false; } // Refresh word ids. - prevWordsInfo->getPrevWordIds(this, &prevWordIdArray, false /* tryLowerCaseSearch */); + ngramContext->getPrevWordIds(this, &prevWordIdArray, false /* tryLowerCaseSearch */); } else { return false; } @@ -399,7 +405,7 @@ bool Ver4PatriciaTriePolicy::addNgramEntry(const PrevWordsInfo *const prevWordsI } } -bool Ver4PatriciaTriePolicy::removeNgramEntry(const PrevWordsInfo *const prevWordsInfo, +bool Ver4PatriciaTriePolicy::removeNgramEntry(const NgramContext *const ngramContext, const CodePointArrayView wordCodePoints) { if (!mBuffers->isUpdatable()) { AKLOGI("Warning: removeNgramEntry() is called for non-updatable dictionary."); @@ -410,8 +416,8 @@ bool Ver4PatriciaTriePolicy::removeNgramEntry(const PrevWordsInfo *const prevWor mDictBuffer->getTailPosition()); return false; } - if (!prevWordsInfo->isValid()) { - AKLOGE("prev words info is not valid for removing n-gram entry form the dictionary."); + if (!ngramContext->isValid()) { + AKLOGE("Ngram context is not valid for removing n-gram entry form the dictionary."); return false; } if (wordCodePoints.size() > MAX_WORD_LENGTH) { @@ -419,7 +425,7 @@ bool Ver4PatriciaTriePolicy::removeNgramEntry(const PrevWordsInfo *const prevWor wordCodePoints.size()); } WordIdArray<MAX_PREV_WORD_COUNT_FOR_N_GRAM> prevWordIdArray; - const WordIdArrayView prevWordIds = prevWordsInfo->getPrevWordIds(this, &prevWordIdArray, + const WordIdArrayView prevWordIds = ngramContext->getPrevWordIds(this, &prevWordIdArray, false /* tryLowerCaseSerch */); if (prevWordIds.firstOrDefault(NOT_A_WORD_ID) == NOT_A_WORD_ID) { return false; @@ -440,26 +446,27 @@ bool Ver4PatriciaTriePolicy::removeNgramEntry(const PrevWordsInfo *const prevWor } -bool Ver4PatriciaTriePolicy::updateCounter(const PrevWordsInfo *const prevWordsInfo, - const CodePointArrayView wordCodePoints, const bool isValidWord, - const HistoricalInfo historicalInfo) { +bool Ver4PatriciaTriePolicy::updateEntriesForWordWithNgramContext( + const NgramContext *const ngramContext, const CodePointArrayView wordCodePoints, + const bool isValidWord, const HistoricalInfo historicalInfo) { if (!mBuffers->isUpdatable()) { - AKLOGI("Warning: updateCounter() is called for non-updatable dictionary."); + AKLOGI("Warning: updateEntriesForWordWithNgramContext() is called for non-updatable " + "dictionary."); return false; } const int probability = isValidWord ? DUMMY_PROBABILITY_FOR_VALID_WORDS : NOT_A_PROBABILITY; const UnigramProperty unigramProperty(false /* representsBeginningOfSentence */, false /* isNotAWord */, false /*isBlacklisted*/, probability, historicalInfo); if (!addUnigramEntry(wordCodePoints, &unigramProperty)) { - AKLOGE("Cannot update unigarm entry in updateCounter()."); + AKLOGE("Cannot update unigarm entry in updateEntriesForWordWithNgramContext()."); return false; } - const int probabilityForNgram = prevWordsInfo->isNthPrevWordBeginningOfSentence(1 /* n */) + const int probabilityForNgram = ngramContext->isNthPrevWordBeginningOfSentence(1 /* n */) ? NOT_A_PROBABILITY : probability; const NgramProperty ngramProperty(wordCodePoints.toVector(), probabilityForNgram, historicalInfo); - if (!addNgramEntry(prevWordsInfo, &ngramProperty)) { - AKLOGE("Cannot update unigarm entry in updateCounter()."); + if (!addNgramEntry(ngramContext, &ngramProperty)) { + AKLOGE("Cannot update unigarm entry in updateEntriesForWordWithNgramContext()."); return false; } return true; diff --git a/native/jni/src/suggest/policyimpl/dictionary/structure/backward/v402/ver4_patricia_trie_policy.h b/native/jni/src/suggest/policyimpl/dictionary/structure/backward/v402/ver4_patricia_trie_policy.h index b82563e61..1ad5e7e36 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/structure/backward/v402/ver4_patricia_trie_policy.h +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/backward/v402/ver4_patricia_trie_policy.h @@ -112,13 +112,13 @@ class Ver4PatriciaTriePolicy : public DictionaryStructureWithBufferPolicy { bool removeUnigramEntry(const CodePointArrayView wordCodePoints); - bool addNgramEntry(const PrevWordsInfo *const prevWordsInfo, + bool addNgramEntry(const NgramContext *const ngramContext, const NgramProperty *const ngramProperty); - bool removeNgramEntry(const PrevWordsInfo *const prevWordsInfo, + bool removeNgramEntry(const NgramContext *const ngramContext, const CodePointArrayView wordCodePoints); - bool updateCounter(const PrevWordsInfo *const prevWordsInfo, + bool updateEntriesForWordWithNgramContext(const NgramContext *const ngramContext, const CodePointArrayView wordCodePoints, const bool isValidWord, const HistoricalInfo historicalInfo); @@ -175,7 +175,7 @@ class Ver4PatriciaTriePolicy : public DictionaryStructureWithBufferPolicy { const WordAttributes getWordAttributes(const int probability, const PtNodeParams &ptNodeParams) const; int getBigramConditionalProbability(const int prevWordUnigramProbability, - const int bigramProbability) const; + const bool isInBeginningOfSentenceContext, const int bigramProbability) const; }; } // namespace v402 } // namespace backward diff --git a/native/jni/src/suggest/policyimpl/dictionary/structure/v2/patricia_trie_policy.cpp b/native/jni/src/suggest/policyimpl/dictionary/structure/v2/patricia_trie_policy.cpp index d3d684bfa..b7f1199c5 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/structure/v2/patricia_trie_policy.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v2/patricia_trie_policy.cpp @@ -23,7 +23,7 @@ #include "suggest/core/dictionary/binary_dictionary_bigrams_iterator.h" #include "suggest/core/dictionary/multi_bigram_map.h" #include "suggest/core/dictionary/ngram_listener.h" -#include "suggest/core/session/prev_words_info.h" +#include "suggest/core/session/ngram_context.h" #include "suggest/policyimpl/dictionary/structure/pt_common/dynamic_pt_reading_helper.h" #include "suggest/policyimpl/dictionary/structure/pt_common/patricia_trie_reading_utils.h" #include "suggest/policyimpl/dictionary/utils/probability_utils.h" diff --git a/native/jni/src/suggest/policyimpl/dictionary/structure/v2/patricia_trie_policy.h b/native/jni/src/suggest/policyimpl/dictionary/structure/v2/patricia_trie_policy.h index 32a95bb6c..b17681388 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/structure/v2/patricia_trie_policy.h +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v2/patricia_trie_policy.h @@ -93,25 +93,26 @@ class PatriciaTriePolicy : public DictionaryStructureWithBufferPolicy { return false; } - bool addNgramEntry(const PrevWordsInfo *const prevWordsInfo, + bool addNgramEntry(const NgramContext *const ngramContext, const NgramProperty *const ngramProperty) { // This method should not be called for non-updatable dictionary. AKLOGI("Warning: addNgramEntry() is called for non-updatable dictionary."); return false; } - bool removeNgramEntry(const PrevWordsInfo *const prevWordsInfo, + bool removeNgramEntry(const NgramContext *const ngramContext, const CodePointArrayView wordCodePoints) { // This method should not be called for non-updatable dictionary. AKLOGI("Warning: removeNgramEntry() is called for non-updatable dictionary."); return false; } - bool updateCounter(const PrevWordsInfo *const prevWordsInfo, + bool updateEntriesForWordWithNgramContext(const NgramContext *const ngramContext, const CodePointArrayView wordCodePoints, const bool isValidWord, const HistoricalInfo historicalInfo) { // This method should not be called for non-updatable dictionary. - AKLOGI("Warning: updateCounter() is called for non-updatable dictionary."); + AKLOGI("Warning: updateEntriesForWordWithNgramContext() is called for non-updatable " + "dictionary."); return false; } diff --git a/native/jni/src/suggest/policyimpl/dictionary/structure/v2/ver2_patricia_trie_node_reader.cpp b/native/jni/src/suggest/policyimpl/dictionary/structure/v2/ver2_patricia_trie_node_reader.cpp index dc0ed96d0..90d4687dd 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/structure/v2/ver2_patricia_trie_node_reader.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v2/ver2_patricia_trie_node_reader.cpp @@ -37,7 +37,7 @@ const PtNodeParams Ver2ParticiaTrieNodeReader::fetchPtNodeParamsInBufferFromPtNo int shortcutPos = NOT_A_DICT_POS; int bigramPos = NOT_A_DICT_POS; int siblingPos = NOT_A_DICT_POS; - PatriciaTrieReadingUtils::readPtNodeInfo(mBuffer.data(), ptNodePos, mShortuctPolicy, + PatriciaTrieReadingUtils::readPtNodeInfo(mBuffer.data(), ptNodePos, mShortcutPolicy, mBigramPolicy, mCodePointTable, &flags, &mergedNodeCodePointCount, mergedNodeCodePoints, &probability, &childrenPos, &shortcutPos, &bigramPos, &siblingPos); if (mergedNodeCodePointCount <= 0) { diff --git a/native/jni/src/suggest/policyimpl/dictionary/structure/v2/ver2_patricia_trie_node_reader.h b/native/jni/src/suggest/policyimpl/dictionary/structure/v2/ver2_patricia_trie_node_reader.h index 24ec5bcca..838d37314 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/structure/v2/ver2_patricia_trie_node_reader.h +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v2/ver2_patricia_trie_node_reader.h @@ -35,7 +35,7 @@ class Ver2ParticiaTrieNodeReader : public PtNodeReader { const DictionaryBigramsStructurePolicy *const bigramPolicy, const DictionaryShortcutsStructurePolicy *const shortcutPolicy, const int *const codePointTable) - : mBuffer(buffer), mBigramPolicy(bigramPolicy), mShortuctPolicy(shortcutPolicy), + : mBuffer(buffer), mBigramPolicy(bigramPolicy), mShortcutPolicy(shortcutPolicy), mCodePointTable(codePointTable) {} virtual const PtNodeParams fetchPtNodeParamsInBufferFromPtNodePos(const int ptNodePos) const; @@ -45,7 +45,7 @@ class Ver2ParticiaTrieNodeReader : public PtNodeReader { const ReadOnlyByteArrayView mBuffer; const DictionaryBigramsStructurePolicy *const mBigramPolicy; - const DictionaryShortcutsStructurePolicy *const mShortuctPolicy; + const DictionaryShortcutsStructurePolicy *const mShortcutPolicy; const int *const mCodePointTable; }; } // namespace latinime diff --git a/native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_node_writer.cpp b/native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_node_writer.cpp index f13512d5a..d28006ae9 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_node_writer.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_node_writer.cpp @@ -285,7 +285,7 @@ bool Ver4PatriciaTrieNodeWriter::addShortcutTarget(const PtNodeParams *const ptN const int shortcutProbability) { if (!mShortcutPolicy->addNewShortcut(ptNodeParams->getTerminalId(), targetCodePoints, targetCodePointCount, shortcutProbability)) { - AKLOGE("Cannot add new shortuct entry. terminalId: %d", ptNodeParams->getTerminalId()); + AKLOGE("Cannot add new shortcut entry. terminalId: %d", ptNodeParams->getTerminalId()); return false; } return true; diff --git a/native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_policy.cpp b/native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_policy.cpp index 036197c41..445bbe07e 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_policy.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_policy.cpp @@ -26,7 +26,7 @@ #include "suggest/core/dictionary/property/ngram_property.h" #include "suggest/core/dictionary/property/unigram_property.h" #include "suggest/core/dictionary/property/word_property.h" -#include "suggest/core/session/prev_words_info.h" +#include "suggest/core/session/ngram_context.h" #include "suggest/policyimpl/dictionary/structure/pt_common/dynamic_pt_reading_helper.h" #include "suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_node_reader.h" #include "suggest/policyimpl/dictionary/utils/forgetting_curve_utils.h" @@ -266,7 +266,7 @@ bool Ver4PatriciaTriePolicy::removeUnigramEntry(const CodePointArrayView wordCod return true; } -bool Ver4PatriciaTriePolicy::addNgramEntry(const PrevWordsInfo *const prevWordsInfo, +bool Ver4PatriciaTriePolicy::addNgramEntry(const NgramContext *const ngramContext, const NgramProperty *const ngramProperty) { if (!mBuffers->isUpdatable()) { AKLOGI("Warning: addNgramEntry() is called for non-updatable dictionary."); @@ -277,8 +277,8 @@ bool Ver4PatriciaTriePolicy::addNgramEntry(const PrevWordsInfo *const prevWordsI mDictBuffer->getTailPosition()); return false; } - if (!prevWordsInfo->isValid()) { - AKLOGE("prev words info is not valid for adding n-gram entry to the dictionary."); + if (!ngramContext->isValid()) { + AKLOGE("Ngram context is not valid for adding n-gram entry to the dictionary."); return false; } if (ngramProperty->getTargetCodePoints()->size() > MAX_WORD_LENGTH) { @@ -287,7 +287,7 @@ bool Ver4PatriciaTriePolicy::addNgramEntry(const PrevWordsInfo *const prevWordsI return false; } WordIdArray<MAX_PREV_WORD_COUNT_FOR_N_GRAM> prevWordIdArray; - const WordIdArrayView prevWordIds = prevWordsInfo->getPrevWordIds(this, &prevWordIdArray, + const WordIdArrayView prevWordIds = ngramContext->getPrevWordIds(this, &prevWordIdArray, false /* tryLowerCaseSearch */); if (prevWordIds.empty()) { return false; @@ -296,19 +296,19 @@ bool Ver4PatriciaTriePolicy::addNgramEntry(const PrevWordsInfo *const prevWordsI if (prevWordIds[i] != NOT_A_WORD_ID) { continue; } - if (!prevWordsInfo->isNthPrevWordBeginningOfSentence(i + 1 /* n */)) { + if (!ngramContext->isNthPrevWordBeginningOfSentence(i + 1 /* n */)) { return false; } const UnigramProperty beginningOfSentenceUnigramProperty( true /* representsBeginningOfSentence */, true /* isNotAWord */, false /* isBlacklisted */, MAX_PROBABILITY /* probability */, HistoricalInfo()); - if (!addUnigramEntry(prevWordsInfo->getNthPrevWordCodePoints(1 /* n */), + if (!addUnigramEntry(ngramContext->getNthPrevWordCodePoints(1 /* n */), &beginningOfSentenceUnigramProperty)) { AKLOGE("Cannot add unigram entry for the beginning-of-sentence."); return false; } // Refresh word ids. - prevWordsInfo->getPrevWordIds(this, &prevWordIdArray, false /* tryLowerCaseSearch */); + ngramContext->getPrevWordIds(this, &prevWordIdArray, false /* tryLowerCaseSearch */); } const int wordId = getWordId(CodePointArrayView(*ngramProperty->getTargetCodePoints()), false /* forceLowerCaseSearch */); @@ -326,7 +326,7 @@ bool Ver4PatriciaTriePolicy::addNgramEntry(const PrevWordsInfo *const prevWordsI } } -bool Ver4PatriciaTriePolicy::removeNgramEntry(const PrevWordsInfo *const prevWordsInfo, +bool Ver4PatriciaTriePolicy::removeNgramEntry(const NgramContext *const ngramContext, const CodePointArrayView wordCodePoints) { if (!mBuffers->isUpdatable()) { AKLOGI("Warning: removeNgramEntry() is called for non-updatable dictionary."); @@ -337,8 +337,8 @@ bool Ver4PatriciaTriePolicy::removeNgramEntry(const PrevWordsInfo *const prevWor mDictBuffer->getTailPosition()); return false; } - if (!prevWordsInfo->isValid()) { - AKLOGE("prev words info is not valid for removing n-gram entry form the dictionary."); + if (!ngramContext->isValid()) { + AKLOGE("Ngram context is not valid for removing n-gram entry form the dictionary."); return false; } if (wordCodePoints.size() > MAX_WORD_LENGTH) { @@ -346,7 +346,7 @@ bool Ver4PatriciaTriePolicy::removeNgramEntry(const PrevWordsInfo *const prevWor wordCodePoints.size()); } WordIdArray<MAX_PREV_WORD_COUNT_FOR_N_GRAM> prevWordIdArray; - const WordIdArrayView prevWordIds = prevWordsInfo->getPrevWordIds(this, &prevWordIdArray, + const WordIdArrayView prevWordIds = ngramContext->getPrevWordIds(this, &prevWordIdArray, false /* tryLowerCaseSerch */); if (prevWordIds.empty() || prevWordIds.contains(NOT_A_WORD_ID)) { return false; @@ -363,29 +363,30 @@ bool Ver4PatriciaTriePolicy::removeNgramEntry(const PrevWordsInfo *const prevWor } } -bool Ver4PatriciaTriePolicy::updateCounter(const PrevWordsInfo *const prevWordsInfo, - const CodePointArrayView wordCodePoints, const bool isValidWord, - const HistoricalInfo historicalInfo) { +bool Ver4PatriciaTriePolicy::updateEntriesForWordWithNgramContext( + const NgramContext *const ngramContext, const CodePointArrayView wordCodePoints, + const bool isValidWord, const HistoricalInfo historicalInfo) { if (!mBuffers->isUpdatable()) { - AKLOGI("Warning: updateCounter() is called for non-updatable dictionary."); + AKLOGI("Warning: updateEntriesForWordWithNgramContext() is called for non-updatable " + "dictionary."); return false; } // TODO: Have count up method in language model dict content. const int probability = isValidWord ? DUMMY_PROBABILITY_FOR_VALID_WORDS : NOT_A_PROBABILITY; const UnigramProperty unigramProperty(false /* representsBeginningOfSentence */, - false /* isNotAWord */, false /*isBlacklisted*/, probability, historicalInfo); + false /* isNotAWord */, false /* isBlacklisted */, probability, historicalInfo); if (!addUnigramEntry(wordCodePoints, &unigramProperty)) { - AKLOGE("Cannot update unigarm entry in updateCounter()."); + AKLOGE("Cannot update unigarm entry in updateEntriesForWordWithNgramContext()."); return false; } - const int probabilityForNgram = prevWordsInfo->isNthPrevWordBeginningOfSentence(1 /* n */) + const int probabilityForNgram = ngramContext->isNthPrevWordBeginningOfSentence(1 /* n */) ? NOT_A_PROBABILITY : probability; const NgramProperty ngramProperty(wordCodePoints.toVector(), probabilityForNgram, historicalInfo); - for (size_t i = 1; i <= prevWordsInfo->getPrevWordCount(); ++i) { - const PrevWordsInfo trimmedPrevWordsInfo(prevWordsInfo->getTrimmedPrevWordsInfo(i)); - if (!addNgramEntry(&trimmedPrevWordsInfo, &ngramProperty)) { - AKLOGE("Cannot update ngram entry in updateCounter()."); + for (size_t i = 1; i <= ngramContext->getPrevWordCount(); ++i) { + const NgramContext trimmedNgramContext(ngramContext->getTrimmedNgramContext(i)); + if (!addNgramEntry(&trimmedNgramContext, &ngramProperty)) { + AKLOGE("Cannot update ngram entry in updateEntriesForWordWithNgramContext()."); return false; } } diff --git a/native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_policy.h b/native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_policy.h index 662bb8d4b..60e30f209 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_policy.h +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_policy.h @@ -92,13 +92,13 @@ class Ver4PatriciaTriePolicy : public DictionaryStructureWithBufferPolicy { bool removeUnigramEntry(const CodePointArrayView wordCodePoints); - bool addNgramEntry(const PrevWordsInfo *const prevWordsInfo, + bool addNgramEntry(const NgramContext *const ngramContext, const NgramProperty *const ngramProperty); - bool removeNgramEntry(const PrevWordsInfo *const prevWordsInfo, + bool removeNgramEntry(const NgramContext *const ngramContext, const CodePointArrayView wordCodePoints); - bool updateCounter(const PrevWordsInfo *const prevWordsInfo, + bool updateEntriesForWordWithNgramContext(const NgramContext *const ngramContext, const CodePointArrayView wordCodePoints, const bool isValidWord, const HistoricalInfo historicalInfo); diff --git a/native/jni/src/suggest/policyimpl/typing/scoring_params.cpp b/native/jni/src/suggest/policyimpl/typing/scoring_params.cpp index 3fc566e7a..b621eef22 100644 --- a/native/jni/src/suggest/policyimpl/typing/scoring_params.cpp +++ b/native/jni/src/suggest/policyimpl/typing/scoring_params.cpp @@ -31,6 +31,7 @@ const float ScoringParams::DIGRAPH_PENALTY_FOR_EXACT_MATCH = 0.03f; // TODO: Unlimit max cache dic node size const int ScoringParams::MAX_CACHE_DIC_NODE_SIZE = 170; const int ScoringParams::MAX_CACHE_DIC_NODE_SIZE_FOR_SINGLE_POINT = 310; +const int ScoringParams::MAX_CACHE_DIC_NODE_SIZE_FOR_LOW_PROBABILITY_LOCALE = 50; const int ScoringParams::THRESHOLD_SHORT_WORD_LENGTH = 4; const float ScoringParams::DISTANCE_WEIGHT_LENGTH = 0.1524f; @@ -61,4 +62,7 @@ const float ScoringParams::HAS_MULTI_WORD_TERMINAL_COST = 0.4182f; const float ScoringParams::TYPING_BASE_OUTPUT_SCORE = 1.0f; const float ScoringParams::TYPING_MAX_OUTPUT_SCORE_PER_INPUT = 0.1f; const float ScoringParams::NORMALIZED_SPATIAL_DISTANCE_THRESHOLD_FOR_EDIT = 0.095f; +const float ScoringParams::LOCALE_WEIGHT_THRESHOLD_FOR_SPACE_SUBSTITUTION = 0.99f; +const float ScoringParams::LOCALE_WEIGHT_THRESHOLD_FOR_SPACE_OMISSION = 0.99f; +const float ScoringParams::LOCALE_WEIGHT_THRESHOLD_FOR_SMALL_CACHE_SIZE = 0.99f; } // namespace latinime diff --git a/native/jni/src/suggest/policyimpl/typing/scoring_params.h b/native/jni/src/suggest/policyimpl/typing/scoring_params.h index b12de6d87..731424f3d 100644 --- a/native/jni/src/suggest/policyimpl/typing/scoring_params.h +++ b/native/jni/src/suggest/policyimpl/typing/scoring_params.h @@ -30,6 +30,7 @@ class ScoringParams { static const float AUTOCORRECT_OUTPUT_THRESHOLD; static const int MAX_CACHE_DIC_NODE_SIZE; static const int MAX_CACHE_DIC_NODE_SIZE_FOR_SINGLE_POINT; + static const int MAX_CACHE_DIC_NODE_SIZE_FOR_LOW_PROBABILITY_LOCALE; static const int THRESHOLD_SHORT_WORD_LENGTH; static const float EXACT_MATCH_PROMOTION; @@ -68,6 +69,9 @@ class ScoringParams { static const float TYPING_BASE_OUTPUT_SCORE; static const float TYPING_MAX_OUTPUT_SCORE_PER_INPUT; static const float NORMALIZED_SPATIAL_DISTANCE_THRESHOLD_FOR_EDIT; + static const float LOCALE_WEIGHT_THRESHOLD_FOR_SPACE_SUBSTITUTION; + static const float LOCALE_WEIGHT_THRESHOLD_FOR_SPACE_OMISSION; + static const float LOCALE_WEIGHT_THRESHOLD_FOR_SMALL_CACHE_SIZE; private: DISALLOW_IMPLICIT_CONSTRUCTORS(ScoringParams); diff --git a/native/jni/src/suggest/policyimpl/typing/typing_traversal.h b/native/jni/src/suggest/policyimpl/typing/typing_traversal.h index b64ee8be4..b9b6314ae 100644 --- a/native/jni/src/suggest/policyimpl/typing/typing_traversal.h +++ b/native/jni/src/suggest/policyimpl/typing/typing_traversal.h @@ -26,6 +26,7 @@ #include "suggest/core/layout/proximity_info_utils.h" #include "suggest/core/policy/traversal.h" #include "suggest/core/session/dic_traverse_session.h" +#include "suggest/core/suggest_options.h" #include "suggest/policyimpl/typing/scoring_params.h" #include "utils/char_utils.h" @@ -77,6 +78,13 @@ class TypingTraversal : public Traversal { if (!CORRECT_NEW_WORD_SPACE_SUBSTITUTION) { return false; } + if (traverseSession->getSuggestOptions()->weightForLocale() + < ScoringParams::LOCALE_WEIGHT_THRESHOLD_FOR_SPACE_SUBSTITUTION) { + // Space substitution is heavy, so we skip doing it if the weight for this language + // is low because we anticipate the suggestions out of this dictionary are not for + // the language the user intends to type in. + return false; + } if (!canDoLookAheadCorrection(traverseSession, dicNode)) { return false; } @@ -91,6 +99,13 @@ class TypingTraversal : public Traversal { if (!CORRECT_NEW_WORD_SPACE_OMISSION) { return false; } + if (traverseSession->getSuggestOptions()->weightForLocale() + < ScoringParams::LOCALE_WEIGHT_THRESHOLD_FOR_SPACE_OMISSION) { + // Space omission is heavy, so we skip doing it if the weight for this language + // is low because we anticipate the suggestions out of this dictionary are not for + // the language the user intends to type in. + return false; + } const int inputSize = traverseSession->getInputSize(); // TODO: Don't refer to isCompletion? if (dicNode->isCompletion(inputSize)) { @@ -141,9 +156,14 @@ class TypingTraversal : public Traversal { return DicNodeVector::DEFAULT_NODES_SIZE_FOR_OPTIMIZATION; } - AK_FORCE_INLINE int getMaxCacheSize(const int inputSize) const { - return (inputSize <= 1) ? ScoringParams::MAX_CACHE_DIC_NODE_SIZE_FOR_SINGLE_POINT - : ScoringParams::MAX_CACHE_DIC_NODE_SIZE; + AK_FORCE_INLINE int getMaxCacheSize(const int inputSize, const float weightForLocale) const { + if (inputSize <= 1) { + return ScoringParams::MAX_CACHE_DIC_NODE_SIZE_FOR_SINGLE_POINT; + } + if (weightForLocale < ScoringParams::LOCALE_WEIGHT_THRESHOLD_FOR_SMALL_CACHE_SIZE) { + return ScoringParams::MAX_CACHE_DIC_NODE_SIZE_FOR_LOW_PROBABILITY_LOCALE; + } + return ScoringParams::MAX_CACHE_DIC_NODE_SIZE; } AK_FORCE_INLINE int getTerminalCacheSize() const { diff --git a/native/jni/src/utils/jni_data_utils.h b/native/jni/src/utils/jni_data_utils.h index 235a03bba..25cc41742 100644 --- a/native/jni/src/utils/jni_data_utils.h +++ b/native/jni/src/utils/jni_data_utils.h @@ -21,7 +21,7 @@ #include "defines.h" #include "jni.h" -#include "suggest/core/session/prev_words_info.h" +#include "suggest/core/session/ngram_context.h" #include "suggest/core/policy/dictionary_header_structure_policy.h" #include "suggest/policyimpl/dictionary/header/header_read_write_utils.h" #include "utils/char_utils.h" @@ -96,7 +96,7 @@ class JniDataUtils { } } - static PrevWordsInfo constructPrevWordsInfo(JNIEnv *env, jobjectArray prevWordCodePointArrays, + static NgramContext constructNgramContext(JNIEnv *env, jobjectArray prevWordCodePointArrays, jbooleanArray isBeginningOfSentenceArray, const size_t prevWordCount) { int prevWordCodePoints[MAX_PREV_WORD_COUNT_FOR_N_GRAM][MAX_WORD_LENGTH]; int prevWordCodePointCount[MAX_PREV_WORD_COUNT_FOR_N_GRAM]; @@ -119,7 +119,7 @@ class JniDataUtils { &isBeginningOfSentenceBoolean); isBeginningOfSentence[i] = isBeginningOfSentenceBoolean == JNI_TRUE; } - return PrevWordsInfo(prevWordCodePoints, prevWordCodePointCount, isBeginningOfSentence, + return NgramContext(prevWordCodePoints, prevWordCodePointCount, isBeginningOfSentence, prevWordCount); } |