Prune out suggestions that have a very large edit distance.
If the number of keys picked from proximity is too large, prune out the subtree. Otherwise you get vastly unrelated suggestions. Fix a bug introduced with the missing_chars checkin.main
parent
c3df2d6fd2
commit
322dc3d3f3
|
@ -60,8 +60,9 @@ int Dictionary::getSuggestions(int *codes, int codesSize, unsigned short *outWor
|
|||
mMaxWords = maxWords;
|
||||
mWords = 0;
|
||||
mSkipPos = skipPos;
|
||||
mMaxEditDistance = mInputLength < 5 ? 2 : mInputLength / 2;
|
||||
|
||||
getWordsRec(0, 0, mInputLength * 3, false, 1, 0);
|
||||
getWordsRec(0, 0, mInputLength * 3, false, 1, 0, 0);
|
||||
|
||||
if (DEBUG_DICT) LOGI("Returning %d words", mWords);
|
||||
return mWords;
|
||||
|
@ -108,7 +109,11 @@ bool
|
|||
Dictionary::addWord(unsigned short *word, int length, int frequency)
|
||||
{
|
||||
word[length] = 0;
|
||||
if (DEBUG_DICT) LOGI("Found word = %s, freq = %d : \n", word, frequency);
|
||||
if (DEBUG_DICT) {
|
||||
char s[length + 1];
|
||||
for (int i = 0; i <= length; i++) s[i] = word[i];
|
||||
LOGI("Found word = %s, freq = %d : \n", s, frequency);
|
||||
}
|
||||
|
||||
// Find the right insertion point
|
||||
int insertAt = 0;
|
||||
|
@ -176,12 +181,16 @@ Dictionary::sameAsTyped(unsigned short *word, int length)
|
|||
static char QUOTE = '\'';
|
||||
|
||||
void
|
||||
Dictionary::getWordsRec(int pos, int depth, int maxDepth, bool completion, int snr, int inputIndex)
|
||||
Dictionary::getWordsRec(int pos, int depth, int maxDepth, bool completion, int snr, int inputIndex,
|
||||
int diffs)
|
||||
{
|
||||
// Optimization: Prune out words that are too long compared to how much was typed.
|
||||
if (depth > maxDepth) {
|
||||
return;
|
||||
}
|
||||
if (diffs > mMaxEditDistance) {
|
||||
return;
|
||||
}
|
||||
int count = getCount(&pos);
|
||||
int *currentChars = NULL;
|
||||
if (mInputLength <= inputIndex) {
|
||||
|
@ -205,19 +214,19 @@ Dictionary::getWordsRec(int pos, int depth, int maxDepth, bool completion, int s
|
|||
}
|
||||
if (childrenAddress != 0) {
|
||||
getWordsRec(childrenAddress, depth + 1, maxDepth,
|
||||
completion, snr, inputIndex);
|
||||
completion, snr, inputIndex, diffs);
|
||||
}
|
||||
} else if (c == QUOTE && currentChars[0] != QUOTE || mSkipPos == depth) {
|
||||
// Skip the ' or other letter and continue deeper
|
||||
mWord[depth] = c;
|
||||
if (childrenAddress != 0) {
|
||||
getWordsRec(childrenAddress, depth + 1, maxDepth, false, snr, inputIndex);
|
||||
getWordsRec(childrenAddress, depth + 1, maxDepth, false, snr, inputIndex, diffs);
|
||||
}
|
||||
} else {
|
||||
int j = 0;
|
||||
while (currentChars[j] > 0) {
|
||||
int addedWeight = j == 0 ? mTypedLetterMultiplier : 1;
|
||||
if (currentChars[j] == lowerC || currentChars[j] == c) {
|
||||
int addedWeight = j == 0 ? mTypedLetterMultiplier : 1;
|
||||
mWord[depth] = c;
|
||||
if (mInputLength == inputIndex + 1) {
|
||||
if (terminal) {
|
||||
|
@ -229,11 +238,12 @@ Dictionary::getWordsRec(int pos, int depth, int maxDepth, bool completion, int s
|
|||
}
|
||||
if (childrenAddress != 0) {
|
||||
getWordsRec(childrenAddress, depth + 1,
|
||||
maxDepth, true, snr * addedWeight, inputIndex + 1);
|
||||
maxDepth, true, snr * addedWeight, inputIndex + 1,
|
||||
diffs + (j > 0));
|
||||
}
|
||||
} else if (childrenAddress != 0) {
|
||||
getWordsRec(childrenAddress, depth + 1, maxDepth,
|
||||
false, snr * addedWeight, inputIndex + 1);
|
||||
false, snr * addedWeight, inputIndex + 1, diffs + (j > 0));
|
||||
}
|
||||
}
|
||||
j++;
|
||||
|
|
|
@ -51,7 +51,7 @@ private:
|
|||
bool addWord(unsigned short *word, int length, int frequency);
|
||||
unsigned short toLowerCase(unsigned short c, int depth);
|
||||
void getWordsRec(int pos, int depth, int maxDepth, bool completion, int frequency,
|
||||
int inputIndex);
|
||||
int inputIndex, int diffs);
|
||||
bool isValidWordRec(int pos, unsigned short *word, int offset, int length);
|
||||
|
||||
unsigned char *mDict;
|
||||
|
@ -67,6 +67,7 @@ private:
|
|||
int mMaxAlternatives;
|
||||
unsigned short mWord[128];
|
||||
int mSkipPos;
|
||||
int mMaxEditDistance;
|
||||
|
||||
int mFullWordMultiplier;
|
||||
int mTypedLetterMultiplier;
|
||||
|
|
|
@ -101,10 +101,11 @@ public class BinaryDictionary extends Dictionary {
|
|||
// completions.
|
||||
if (ENABLE_MISSED_CHARACTERS && count < 5) {
|
||||
for (int skip = 0; skip < codesSize; skip++) {
|
||||
count = getSuggestionsNative(mNativeDict, mInputCodes, codesSize,
|
||||
int tempCount = getSuggestionsNative(mNativeDict, mInputCodes, codesSize,
|
||||
mOutputChars, mFrequencies,
|
||||
MAX_WORD_LENGTH, MAX_WORDS, MAX_ALTERNATIVES, skip);
|
||||
if (count > 0) break;
|
||||
count = Math.max(count, tempCount);
|
||||
if (tempCount > 0) break;
|
||||
}
|
||||
}
|
||||
|
||||
|
|
Loading…
Reference in New Issue