/**
* 从一个词的词性到另一个词的词的分数
*
* @param form 前面的词
* @param to 后面的词
* @return 分数
*/
public static double compuScore(Term from, Term to) {
double frequency = from.termNatures().allFreq + 1;
if (frequency < 0) {
double score = from.score() + MAX_FREQUENCE;
from.score(score);
return score;
}
int nTwoWordsFreq = NgramLibrary.getTwoWordFreq(from, to);
double value =
-Math.log(
dSmoothingPara * frequency / (MAX_FREQUENCE + 80000)
+ (1 - dSmoothingPara) * ((1 - dTemp) * nTwoWordsFreq / frequency + dTemp));
if (value < 0) {
value += frequency;
}
return from.score() + value;
}
/**
* 人名识别
*
* @param term
* @param offe
* @param freq
*/
private Term nameFind(int offe, int beginFreq, int size) {
// TODO Auto-generated method stub
StringBuilder sb = new StringBuilder();
int undefinite = 0;
skip = false;
PersonNatureAttr pna = null;
int index = 0;
int freq = 0;
double allFreq = 0;
Term term = null;
int i = offe;
for (; i < terms.length; i++) {
// 走到结尾处识别出来一个名字.
if (terms[i] == null) {
continue;
}
term = terms[i];
pna = term.getTermNatures().personAttr;
// 在这个长度的这个位置的词频,如果没有可能就干掉,跳出循环
if ((freq = pna.getFreq(size, index)) == 0) {
return null;
}
if (pna.allFreq > 0) {
undefinite++;
}
sb.append(term.getName());
allFreq += Math.log(term.getTermNatures().allFreq + 1);
allFreq += -Math.log((freq));
index++;
if (index == size + 2) {
break;
}
}
double score = -Math.log(FACTORY[size]);
score += allFreq;
double endFreq = 0;
// 开始寻找结尾词
boolean flag = true;
while (flag) {
i++;
if (i >= terms.length) {
endFreq = 10;
flag = false;
} else if (terms[i] != null) {
int twoWordFreq = NgramLibrary.getTwoWordFreq(term, terms[i]);
if (twoWordFreq > 3) {
return null;
}
endFreq = terms[i].getTermNatures().personAttr.end + 1;
flag = false;
}
}
score -= Math.log(endFreq);
score -= Math.log(beginFreq);
if (score > -3) {
return null;
}
if (allFreq > 0 && undefinite > 0) {
return null;
}
skip = undefinite == 0;
term = new Term(sb.toString(), offe, TermNatures.NR);
term.selfScore = score;
return term;
}