AI 143659: am: CL 143472 Reduce dictionary size.
Changed the tree structure to have variable length nodes to save an average of 21% on the dictionary size. Created a shortened English dictionary for Dream - 50K words. Added a shortened Spanish dictionary for Dream - 32K words. Original author: yamasani Merged from: //branches/cupcake/... Automated import of CL 143659main
parent
152d8a7957
commit
cc3e5c7c10
|
@ -85,10 +85,14 @@ int
|
||||||
Dictionary::getAddress(int *pos)
|
Dictionary::getAddress(int *pos)
|
||||||
{
|
{
|
||||||
int address = 0;
|
int address = 0;
|
||||||
address += (mDict[*pos] & 0x7F) << 16;
|
if ((mDict[*pos] & FLAG_ADDRESS_MASK) == 0) {
|
||||||
address += (mDict[*pos + 1] & 0xFF) << 8;
|
*pos += 1;
|
||||||
address += (mDict[*pos + 2] & 0xFF);
|
} else {
|
||||||
*pos += 3;
|
address += (mDict[*pos] & (ADDRESS_MASK >> 16)) << 16;
|
||||||
|
address += (mDict[*pos + 1] & 0xFF) << 8;
|
||||||
|
address += (mDict[*pos + 2] & 0xFF);
|
||||||
|
*pos += 3;
|
||||||
|
}
|
||||||
return address;
|
return address;
|
||||||
}
|
}
|
||||||
|
|
||||||
|
@ -193,7 +197,8 @@ Dictionary::getWordsRec(int pos, int depth, int maxDepth, bool completion, int s
|
||||||
unsigned short lowerC = toLowerCase(c, depth);
|
unsigned short lowerC = toLowerCase(c, depth);
|
||||||
bool terminal = getTerminal(&pos);
|
bool terminal = getTerminal(&pos);
|
||||||
int childrenAddress = getAddress(&pos);
|
int childrenAddress = getAddress(&pos);
|
||||||
int freq = getFreq(&pos);
|
int freq = 1;
|
||||||
|
if (terminal) freq = getFreq(&pos);
|
||||||
// If we are only doing completions, no need to look at the typed characters.
|
// If we are only doing completions, no need to look at the typed characters.
|
||||||
if (completion) {
|
if (completion) {
|
||||||
mWord[depth] = c;
|
mWord[depth] = c;
|
||||||
|
@ -266,7 +271,9 @@ Dictionary::isValidWordRec(int pos, unsigned short *word, int offset, int length
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
getFreq(&pos);
|
if (terminal) {
|
||||||
|
getFreq(&pos);
|
||||||
|
}
|
||||||
// There could be two instances of each alphabet - upper and lower case. So continue
|
// There could be two instances of each alphabet - upper and lower case. So continue
|
||||||
// looking ...
|
// looking ...
|
||||||
}
|
}
|
||||||
|
|
|
@ -19,35 +19,44 @@
|
||||||
|
|
||||||
namespace latinime {
|
namespace latinime {
|
||||||
|
|
||||||
|
// 22-bit address = ~4MB dictionary size limit, which on average would be about 200k-300k words
|
||||||
|
#define ADDRESS_MASK 0x3FFFFF
|
||||||
|
|
||||||
|
// The bit that decides if an address follows in the next 22 bits
|
||||||
|
#define FLAG_ADDRESS_MASK 0x40
|
||||||
|
// The bit that decides if this is a terminal node for a word. The node could still have children,
|
||||||
|
// if the word has other endings.
|
||||||
|
#define FLAG_TERMINAL_MASK 0x80
|
||||||
|
|
||||||
class Dictionary {
|
class Dictionary {
|
||||||
public:
|
public:
|
||||||
Dictionary(void *dict, int typedLetterMultipler, int fullWordMultiplier);
|
Dictionary(void *dict, int typedLetterMultipler, int fullWordMultiplier);
|
||||||
int getSuggestions(int *codes, int codesSize, unsigned short *outWords, int *frequencies,
|
int getSuggestions(int *codes, int codesSize, unsigned short *outWords, int *frequencies,
|
||||||
int maxWordLength, int maxWords, int maxAlternatives);
|
int maxWordLength, int maxWords, int maxAlternatives);
|
||||||
bool isValidWord(unsigned short *word, int length);
|
bool isValidWord(unsigned short *word, int length);
|
||||||
void setAsset(void *asset) { mAsset = asset; }
|
void setAsset(void *asset) { mAsset = asset; }
|
||||||
void *getAsset() { return mAsset; }
|
void *getAsset() { return mAsset; }
|
||||||
~Dictionary();
|
~Dictionary();
|
||||||
|
|
||||||
private:
|
private:
|
||||||
|
|
||||||
int getAddress(int *pos);
|
int getAddress(int *pos);
|
||||||
bool getTerminal(int *pos) { return (mDict[*pos] & 0x80) > 0; }
|
bool getTerminal(int *pos) { return (mDict[*pos] & FLAG_TERMINAL_MASK) > 0; }
|
||||||
int getFreq(int *pos) { return mDict[(*pos)++] & 0xFF; }
|
int getFreq(int *pos) { return mDict[(*pos)++] & 0xFF; }
|
||||||
int getCount(int *pos) { return mDict[(*pos)++] & 0xFF; }
|
int getCount(int *pos) { return mDict[(*pos)++] & 0xFF; }
|
||||||
unsigned short getChar(int *pos);
|
unsigned short getChar(int *pos);
|
||||||
int wideStrLen(unsigned short *str);
|
int wideStrLen(unsigned short *str);
|
||||||
|
|
||||||
bool sameAsTyped(unsigned short *word, int length);
|
bool sameAsTyped(unsigned short *word, int length);
|
||||||
bool addWord(unsigned short *word, int length, int frequency);
|
bool addWord(unsigned short *word, int length, int frequency);
|
||||||
unsigned short toLowerCase(unsigned short c, int depth);
|
unsigned short toLowerCase(unsigned short c, int depth);
|
||||||
void getWordsRec(int pos, int depth, int maxDepth, bool completion, int frequency,
|
void getWordsRec(int pos, int depth, int maxDepth, bool completion, int frequency,
|
||||||
int inputIndex);
|
int inputIndex);
|
||||||
bool isValidWordRec(int pos, unsigned short *word, int offset, int length);
|
bool isValidWordRec(int pos, unsigned short *word, int offset, int length);
|
||||||
|
|
||||||
unsigned char *mDict;
|
unsigned char *mDict;
|
||||||
void *mAsset;
|
void *mAsset;
|
||||||
|
|
||||||
int *mFrequencies;
|
int *mFrequencies;
|
||||||
int mMaxWords;
|
int mMaxWords;
|
||||||
int mMaxWordLength;
|
int mMaxWordLength;
|
||||||
|
@ -57,7 +66,7 @@ private:
|
||||||
int mInputLength;
|
int mInputLength;
|
||||||
int mMaxAlternatives;
|
int mMaxAlternatives;
|
||||||
unsigned short mWord[128];
|
unsigned short mWord[128];
|
||||||
|
|
||||||
int mFullWordMultiplier;
|
int mFullWordMultiplier;
|
||||||
int mTypedLetterMultiplier;
|
int mTypedLetterMultiplier;
|
||||||
};
|
};
|
||||||
|
|
Loading…
Reference in New Issue