Merge "Remove checkFirstCharacter from BigramDictionary."

main
Keisuke Kuroynagi 2013-08-05 04:56:23 +00:00 committed by Android (Google) Code Review
commit 22f951c881
5 changed files with 23 additions and 55 deletions

View File

@ -186,7 +186,7 @@ static int latinime_BinaryDictionary_getSuggestions(JNIEnv *env, jclass clazz, j
scores, spaceIndices, outputTypes);
} else {
count = dictionary->getBigrams(prevWordCodePoints, prevWordCodePointsLength,
inputCodePoints, inputSize, outputCodePoints, scores, outputTypes);
outputCodePoints, scores, outputTypes);
}
// Copy back the output values

View File

@ -87,21 +87,14 @@ void BigramDictionary::addWordBigram(int *word, int length, int probability, int
/* Parameters :
* prevWord: the word before, the one for which we need to look up bigrams.
* prevWordLength: its length.
* inputCodePoints: what user typed, in the same format as for UnigramDictionary::getSuggestions.
* inputSize: the size of the codes array.
* bigramCodePoints: an array for output, at the same format as outwords for getSuggestions.
* bigramProbability: an array to output frequencies.
* outBigramCodePoints: an array for output, at the same format as outwords for getSuggestions.
* outBigramProbability: an array to output frequencies.
* outputTypes: an array to output types.
* This method returns the number of bigrams this word has, for backward compatibility.
* Note: this is not the number of bigrams output in the array, which is the number of
* bigrams this word has WHOSE first letter also matches the letter the user typed.
* TODO: this may not be a sensible thing to do. It makes sense when the bigrams are
* used to match the first letter of the second word, but once the user has typed more
* 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::getPredictions(const int *prevWord, int prevWordLength, int *inputCodePoints,
int inputSize, int *bigramCodePoints, int *bigramProbability, int *outputTypes) const {
int BigramDictionary::getPredictions(const int *prevWord, const int prevWordLength,
int *const outBigramCodePoints, int *const outBigramProbability,
int *const outputTypes) 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
@ -126,21 +119,16 @@ int BigramDictionary::getPredictions(const int *prevWord, int prevWordLength, in
getCodePointsAndProbabilityAndReturnCodePointCount(
mBinaryDictionaryInfo, bigramsIt.getBigramPos(), MAX_WORD_LENGTH,
bigramBuffer, &unigramProbability);
// inputSize == 0 means we are trying to find bigram predictions.
if (inputSize < 1 || checkFirstCharacter(bigramBuffer, inputCodePoints)) {
const int bigramProbabilityTemp = bigramsIt.getProbability();
// Due to space constraints, the probability for bigrams is approximate - the lower the
// unigram probability, the worse the precision. The theoritical maximum error in
// resulting probability 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 probability = ProbabilityUtils::computeProbabilityForBigram(
unigramProbability, bigramProbabilityTemp);
addWordBigram(bigramBuffer, length, probability, bigramProbability, bigramCodePoints,
outputTypes);
++bigramCount;
}
// Due to space constraints, the probability for bigrams is approximate - the lower the
// unigram probability, the worse the precision. The theoritical maximum error in
// resulting probability 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 probability = ProbabilityUtils::computeProbabilityForBigram(
unigramProbability, bigramsIt.getProbability());
addWordBigram(bigramBuffer, length, probability, outBigramProbability, outBigramCodePoints,
outputTypes);
++bigramCount;
}
return min(bigramCount, MAX_RESULTS);
}
@ -157,22 +145,6 @@ int BigramDictionary::getBigramListPositionForWord(const int *prevWord, const in
mBinaryDictionaryInfo, pos);
}
bool BigramDictionary::checkFirstCharacter(int *word, int *inputCodePoints) const {
// Checks whether this word starts with same character or neighboring characters of
// what user typed.
int maxAlt = MAX_ALTERNATIVES;
const int firstBaseLowerCodePoint = CharUtils::toBaseLowerCase(*word);
while (maxAlt > 0) {
if (CharUtils::toBaseLowerCase(*inputCodePoints) == firstBaseLowerCodePoint) {
return true;
}
inputCodePoints++;
maxAlt--;
}
return false;
}
bool BigramDictionary::isValidBigram(const int *word0, int length0, const int *word1,
int length1) const {
int pos = getBigramListPositionForWord(word0, length0, false /* forceLowerCaseSearch */);

View File

@ -27,8 +27,8 @@ class BigramDictionary {
public:
BigramDictionary(const BinaryDictionaryInfo *const binaryDictionaryInfo);
int getPredictions(const int *word, int length, int *inputCodePoints, int inputSize,
int *outWords, int *frequencies, int *outputTypes) const;
int getPredictions(const int *word, int length, int *outBigramCodePoints,
int *outBigramProbability, int *outputTypes) const;
bool isValidBigram(const int *word1, int length1, const int *word2, int length2) const;
~BigramDictionary();
@ -37,13 +37,10 @@ class BigramDictionary {
void addWordBigram(int *word, int length, int probability, int *bigramProbability,
int *bigramCodePoints, int *outputTypes) const;
bool checkFirstCharacter(int *word, int *inputCodePoints) const;
int getBigramListPositionForWord(const int *prevWord, const int prevWordLength,
const bool forceLowerCaseSearch) const;
const BinaryDictionaryInfo *const mBinaryDictionaryInfo;
// TODO: Re-implement proximity correction for bigram correction
static const int MAX_ALTERNATIVES = 1;
};
} // namespace latinime
#endif // LATINIME_BIGRAM_DICTIONARY_H

View File

@ -76,11 +76,10 @@ int Dictionary::getSuggestions(ProximityInfo *proximityInfo, DicTraverseSession
}
}
int Dictionary::getBigrams(const int *word, int length, int *inputCodePoints, int inputSize,
int *outWords, int *frequencies, int *outputTypes) const {
int Dictionary::getBigrams(const int *word, int length, int *outWords, int *frequencies,
int *outputTypes) const {
if (length <= 0) return 0;
return mBigramDictionary->getPredictions(word, length, inputCodePoints, inputSize, outWords,
frequencies, outputTypes);
return mBigramDictionary->getPredictions(word, length, outWords, frequencies, outputTypes);
}
int Dictionary::getProbability(const int *word, int length) const {

View File

@ -62,8 +62,8 @@ class Dictionary {
const SuggestOptions *const suggestOptions, int *outWords, int *frequencies,
int *spaceIndices, int *outputTypes) const;
int getBigrams(const int *word, int length, int *inputCodePoints, int inputSize, int *outWords,
int *frequencies, int *outputTypes) const;
int getBigrams(const int *word, int length, int *outWords, int *frequencies,
int *outputTypes) const;
int getProbability(const int *word, int length) const;