org.ansj.util.Graph Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of ansj_seg Show documentation
Show all versions of ansj_seg Show documentation
best java chinese word seg !
package org.ansj.util;
import java.util.List;
import java.util.Map;
import org.ansj.domain.AnsjItem;
import org.ansj.domain.Term;
import org.ansj.domain.TermNatures;
import org.ansj.library.DATDictionary;
import org.ansj.splitWord.Analysis.Merger;
import org.ansj.util.TermUtil.InsertTermType;
/**
* 最短路径
*
* @author ansj
*
*/
public class Graph {
public char[] chars = null;
public String realStr = null;
public Term[] terms = null;
protected Term end = null;
protected Term root = null;
protected static final String E = "末##末";
protected static final String B = "始##始";
// 是否有人名
public boolean hasPerson;
// 是否有数字
public boolean hasNum;
// 是否需有歧异
public Graph(String str) {
realStr = str;
this.chars = str.toCharArray();
terms = new Term[chars.length + 1];
end = new Term(E, chars.length, AnsjItem.END);
root = new Term(B, -1, AnsjItem.BEGIN);
terms[chars.length] = end;
}
/**
* 构建最优路径
*/
public List getResult(Merger merger) {
return merger.merger();
}
/**
* 增加一个词语到图中
*
* @param term
*/
public void addTerm(Term term) {
// 是否有数字
if (!hasNum && term.termNatures().numAttr.numFreq > 0) {
hasNum = true;
}
// 是否有人名
if (!hasPerson && term.termNatures().personAttr.flag) {
hasPerson = true;
}
TermUtil.insertTerm(terms, term, InsertTermType.REPLACE);
}
/**
* 取得最优路径的root Term
*
* @return
*/
protected Term optimalRoot() {
Term to = end;
to.clearScore();
Term from = null;
while ((from = to.from()) != null) {
for (int i = from.getOffe() + 1; i < to.getOffe(); i++) {
terms[i] = null;
}
if (from.getOffe() > -1) {
terms[from.getOffe()] = from;
}
// 断开横向链表.节省内存
from.setNext(null);
from.setTo(to);
from.clearScore();
to = from;
}
return root;
}
/**
* 删除最短的节点
*/
public void rmLittlePath() {
int maxTo = -1;
Term temp = null;
Term maxTerm = null;
// 是否有交叉
boolean flag = false;
int length = terms.length - 1;
for (int i = 0; i < length; i++) {
maxTerm = getMaxTerm(i);
if (maxTerm == null)
continue;
maxTo = maxTerm.toValue();
/**
* 对字数进行优化.如果一个字.就跳过..两个字.且第二个为null则.也跳过.从第二个后开始
*/
switch (maxTerm.getName().length()) {
case 1:
continue;
case 2:
if (terms[i + 1] == null) {
i = i + 1;
continue;
}
}
/**
* 判断是否有交叉
*/
for (int j = i + 1; j < maxTo; j++) {
temp = getMaxTerm(j);
if (temp == null) {
continue;
}
if (maxTo < temp.toValue()) {
maxTo = temp.toValue();
flag = true;
}
}
if (flag) {
i = maxTo - 1;
flag = false;
} else {
maxTerm.setNext(null);
terms[i] = maxTerm;
for (int j = i + 1; j < maxTo; j++) {
terms[j] = null;
}
// FIXME: 这里理论上得设置。但是跑了这么久，还不发生错误。应该是不依赖于双向链接。需要确认下。这段代码是否有用
// //将下面的to的from设置回来
// temp = terms[i+maxTerm.getName().length()] ;
// do{
// temp.setFrom(maxTerm) ;
// }while((temp=temp.next())!=null) ;
}
}
}
/**
* 得道最到本行最大term,也就是最右面的term
*
* @param i
* @return
*/
private Term getMaxTerm(int i) {
Term maxTerm = terms[i];
if (maxTerm == null) {
return null;
}
Term term = maxTerm;
while ((term = term.next()) != null) {
maxTerm = term;
}
return maxTerm;
}
/**
* 删除无意义的节点,防止viterbi太多
*/
public void rmLittleSinglePath() {
int maxTo = -1;
Term temp = null;
for (int i = 0; i < terms.length; i++) {
if (terms[i] == null)
continue;
maxTo = terms[i].toValue();
if (maxTo - i == 1 || i + 1 == terms.length)
continue;
for (int j = i; j < maxTo; j++) {
temp = terms[j];
if (temp != null && temp.toValue() <= maxTo && temp.getName().length() == 1) {
terms[j] = null;
}
}
}
}
/**
* 删除小节点。保证被删除的小节点的单个分数小于等于大节点的分数
*/
public void rmLittlePathByScore() {
int maxTo = -1;
Term temp = null;
for (int i = 0; i < terms.length; i++) {
if (terms[i] == null) {
continue;
}
Term maxTerm = null;
double maxScore = 0;
Term term = terms[i];
// 找到自身分数对大最长的
do {
if (maxTerm == null || maxScore > term.score()) {
maxTerm = term;
} else if (maxScore == term.score() && maxTerm.getName().length() < term.getName().length()) {
maxTerm = term;
}
} while ((term = term.next()) != null);
term = maxTerm;
do {
maxTo = term.toValue();
maxScore = term.score();
if (maxTo - i == 1 || i + 1 == terms.length)
continue;
boolean flag = true;// 可以删除
out: for (int j = i; j < maxTo; j++) {
temp = terms[j];
if (temp == null) {
continue;
}
do {
if (temp.toValue() > maxTo || temp.score() < maxScore) {
flag = false;
break out;
}
} while ((temp = temp.next()) != null);
}
// 验证通过可以删除了
if (flag) {
for (int j = i + 1; j < maxTo; j++) {
terms[j] = null;
}
}
} while ((term = term.next()) != null);
}
}
public void walkPathByScore() {
Term term = null;
// BEGIN先行打分
mergerByScore(root, 0);
// 从第一个词开始往后打分
for (int i = 0; i < terms.length; i++) {
term = terms[i];
while (term != null && term.from() != null && term != end) {
int to = term.toValue();
mergerByScore(term, to);
term = term.next();
}
}
optimalRoot();
}
public void walkPath() {
walkPath(null);
}
/**
* 干涉性增加相对权重
*
* @param relationMap
*/
public void walkPath(Map relationMap) {
Term term = null;
// BEGIN先行打分
merger(root, 0, relationMap);
// 从第一个词开始往后打分
for (int i = 0; i < terms.length; i++) {
term = terms[i];
while (term != null && term.from() != null && term != end) {
int to = term.toValue();
merger(term, to, relationMap);
term = term.next();
}
}
optimalRoot();
}
/**
* 具体的遍历打分方法
*
* @param i
* 起始位置
* @param j
* 起始属性
* @param to
*/
private void merger(Term fromTerm, int to, Map relationMap) {
Term term = null;
if (terms[to] != null) {
term = terms[to];
while (term != null) {
// 关系式to.set(from)
term.setPathScore(fromTerm, relationMap);
term = term.next();
}
} else {
char c = chars[to];
TermNatures tn = DATDictionary.getItem(c).termNatures;
if (tn == null || tn == TermNatures.NULL) {
tn = TermNatures.NULL;
}
terms[to] = new Term(String.valueOf(c), to, tn);
terms[to].setPathScore(fromTerm, relationMap);
}
}
/**
* 根据分数
*
* @param i
* 起始位置
* @param j
* 起始属性
* @param to
*/
private void mergerByScore(Term fromTerm, int to) {
Term term = null;
if (terms[to] != null) {
term = terms[to];
while (term != null) {
// 关系式to.set(from)
term.setPathSelfScore(fromTerm);
term = term.next();
}
}
}
/**
* 对graph进行调试用的
*/
public void printGraph() {
for (Term term : terms) {
if (term == null) {
continue;
}
System.out.print(term.getName() + "\t" + term.score() + " ,");
while ((term = term.next()) != null) {
System.out.print(term + "\t" + term.score() + " ,");
}
System.out.println();
}
}
}