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.cpp27
-rw-r--r--native/jni/src/bigram_dictionary.h3
-rw-r--r--native/jni/src/binary_format.h7
-rw-r--r--native/jni/src/char_utils.h7
-rw-r--r--native/jni/src/unigram_dictionary.cpp3
5 files changed, 34 insertions, 13 deletions
diff --git a/native/jni/src/bigram_dictionary.cpp b/native/jni/src/bigram_dictionary.cpp
index 144336981..3bfbfad25 100644
--- a/native/jni/src/bigram_dictionary.cpp
+++ b/native/jni/src/bigram_dictionary.cpp
@@ -105,8 +105,15 @@ int BigramDictionary::getBigrams(const int32_t *prevWord, int prevWordLength, in
// TODO: have "in" arguments before "out" ones, and make out args explicit in the name
const uint8_t* const root = DICT;
- int pos = getBigramListPositionForWord(prevWord, prevWordLength);
+ int pos = getBigramListPositionForWord(prevWord, prevWordLength,
+ false /* forceLowerCaseSearch */);
// getBigramListPositionForWord returns 0 if this word isn't in the dictionary or has no bigrams
+ if (0 == pos) {
+ // If no bigrams for this exact word, search again in lower case.
+ pos = getBigramListPositionForWord(prevWord, prevWordLength,
+ true /* forceLowerCaseSearch */);
+ }
+ // If still no bigrams, we really don't have them!
if (0 == pos) return 0;
int bigramFlags;
int bigramCount = 0;
@@ -141,10 +148,11 @@ 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 {
+ const int prevWordLength, const bool forceLowerCaseSearch) const {
if (0 >= prevWordLength) return 0;
const uint8_t* const root = DICT;
- int pos = BinaryFormat::getTerminalPosition(root, prevWord, prevWordLength);
+ int pos = BinaryFormat::getTerminalPosition(root, prevWord, prevWordLength,
+ forceLowerCaseSearch);
if (NOT_VALID_WORD == pos) return 0;
const int flags = BinaryFormat::getFlagsAndForwardPointer(root, &pos);
@@ -164,7 +172,13 @@ void BigramDictionary::fillBigramAddressToFrequencyMapAndFilter(const int32_t *p
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);
+ int pos = getBigramListPositionForWord(prevWord, prevWordLength,
+ false /* forceLowerCaseSearch */);
+ if (0 == pos) {
+ // If no bigrams for this exact string, search again in lower case.
+ pos = getBigramListPositionForWord(prevWord, prevWordLength,
+ true /* forceLowerCaseSearch */);
+ }
if (0 == pos) return;
int bigramFlags;
@@ -197,10 +211,11 @@ bool BigramDictionary::checkFirstCharacter(unsigned short *word, int *inputCodes
bool BigramDictionary::isValidBigram(const int32_t *word1, int length1, const int32_t *word2,
int length2) const {
const uint8_t* const root = DICT;
- int pos = getBigramListPositionForWord(word1, length1);
+ int pos = getBigramListPositionForWord(word1, length1, false /* forceLowerCaseSearch */);
// getBigramListPositionForWord returns 0 if this word isn't in the dictionary or has no bigrams
if (0 == pos) return false;
- int nextWordPos = BinaryFormat::getTerminalPosition(root, word2, length2);
+ int nextWordPos = BinaryFormat::getTerminalPosition(root, word2, length2,
+ false /* forceLowerCaseSearch */);
if (NOT_VALID_WORD == nextWordPos) return false;
int bigramFlags;
do {
diff --git a/native/jni/src/bigram_dictionary.h b/native/jni/src/bigram_dictionary.h
index 1ff1b2ec6..5372276cd 100644
--- a/native/jni/src/bigram_dictionary.h
+++ b/native/jni/src/bigram_dictionary.h
@@ -30,7 +30,8 @@ class BigramDictionary {
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;
+ int getBigramListPositionForWord(const int32_t *prevWord, const int prevWordLength,
+ const bool forceLowerCaseSearch) const;
void fillBigramAddressToFrequencyMapAndFilter(const int32_t *prevWord, const int prevWordLength,
std::map<int, int> *map, uint8_t *filter) const;
bool isValidBigram(const int32_t *word1, int length1, const int32_t *word2, int length2) const;
diff --git a/native/jni/src/binary_format.h b/native/jni/src/binary_format.h
index 214ecfa8d..474c854fe 100644
--- a/native/jni/src/binary_format.h
+++ b/native/jni/src/binary_format.h
@@ -19,6 +19,7 @@
#include <limits>
#include "bloom_filter.h"
+#include "char_utils.h"
#include "unigram_dictionary.h"
namespace latinime {
@@ -65,7 +66,7 @@ class BinaryFormat {
static int getAttributeAddressAndForwardPointer(const uint8_t* const dict, const uint8_t flags,
int *pos);
static int getTerminalPosition(const uint8_t* const root, const int32_t* const inWord,
- const int length);
+ const int length, const bool forceLowerCaseSearch);
static int getWordAtAddress(const uint8_t* const root, const int address, const int maxDepth,
uint16_t* outWord, int* outUnigramFrequency);
static int computeFrequencyForBigram(const int unigramFreq, const int bigramFreq);
@@ -309,7 +310,7 @@ 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 int32_t* const inWord, const int length) {
+ const int32_t* const inWord, const int length, const bool forceLowerCaseSearch) {
int pos = 0;
int wordPos = 0;
@@ -318,7 +319,7 @@ 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 int32_t wChar = inWord[wordPos];
+ const int32_t wChar = forceLowerCaseSearch ? toLowerCase(inWord[wordPos]) : 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.
diff --git a/native/jni/src/char_utils.h b/native/jni/src/char_utils.h
index 607dc5195..21dca9a0a 100644
--- a/native/jni/src/char_utils.h
+++ b/native/jni/src/char_utils.h
@@ -50,8 +50,7 @@ inline static unsigned short toBaseChar(unsigned short c) {
return c;
}
-inline static unsigned short toBaseLowerCase(unsigned short c) {
- c = toBaseChar(c);
+inline static unsigned short toLowerCase(const unsigned short c) {
if (isAsciiUpper(c)) {
return toAsciiLower(c);
} else if (isAscii(c)) {
@@ -60,6 +59,10 @@ inline static unsigned short toBaseLowerCase(unsigned short c) {
return latin_tolower(c);
}
+inline static unsigned short toBaseLowerCase(const unsigned short c) {
+ return toLowerCase(toBaseChar(c));
+}
+
} // namespace latinime
#endif // LATINIME_CHAR_UTILS_H
diff --git a/native/jni/src/unigram_dictionary.cpp b/native/jni/src/unigram_dictionary.cpp
index 3417d2ba7..22f1657ef 100644
--- a/native/jni/src/unigram_dictionary.cpp
+++ b/native/jni/src/unigram_dictionary.cpp
@@ -817,7 +817,8 @@ int UnigramDictionary::getMostFrequentWordLikeInner(const uint16_t * const inWor
int UnigramDictionary::getFrequency(const int32_t* const inWord, const int length) const {
const uint8_t* const root = DICT_ROOT;
- int pos = BinaryFormat::getTerminalPosition(root, inWord, length);
+ int pos = BinaryFormat::getTerminalPosition(root, inWord, length,
+ false /* forceLowerCaseSearch */);
if (NOT_VALID_WORD == pos) {
return NOT_A_PROBABILITY;
}
generated by cgit v1.2.3-83-g751a (git 2.47.1) at 2025-05-30 22:22:20 +0000