python 计算两篇文章的相似度算法simhash见:
https://blog.csdn.net/u013421629/article/details/85052915
对长文本 是比较合适的(超过500字以上)
下面贴上java 版本实现:
pom.xml 加入依赖
<dependency>
<groupId>org.jsoup</groupId>
<artifactId>jsoup</artifactId>
<version>1.8.1</version>
</dependency>
创建文件
MySimHash.java
/*计算两篇文章相似度*/
import com.hankcs.hanlp.seg.common.Term;
import com.hankcs.hanlp.tokenizer.StandardTokenizer;
import org.apache.commons.lang3.StringUtils;
import org.jsoup.Jsoup;
import org.jsoup.safety.Whitelist;
import java.math.BigInteger;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
public class MySimHash {
private String tokens; //字符串
private BigInteger strSimHash;//字符产的hash值
private int hashbits = 64; // 分词后的hash数;
public MySimHash(String tokens) {
this.tokens = tokens;
this.strSimHash = this.simHash();
}
private MySimHash(String tokens, int hashbits) {
this.tokens = tokens;
this.hashbits = hashbits;
this.strSimHash = this.simHash();
}
/**
* 清除html标签
* @param content
* @return
*/
private String cleanResume(String content) {
// 若输入为HTML,下面会过滤掉所有的HTML的tag
content = Jsoup.clean(content, Whitelist.none());
content = StringUtils.lowerCase(content);
String[] strings = {" ", "\n", "\r", "\t", "\\r", "\\n", "\\t", " "};
for (String s : strings) {
content = content.replaceAll(s, "");
}
return content;
}
/**
* 这个是对整个字符串进行hash计算
* @return
*/
private BigInteger simHash() {
tokens = cleanResume(tokens); // cleanResume 删除一些特殊字符
int[] v = new int[this.hashbits];
List<Term> termList = StandardTokenizer.segment(this.tokens); // 对字符串进行分词
//对分词的一些特殊处理 : 比如: 根据词性添加权重 , 过滤掉标点符号 , 过滤超频词汇等;
Map<String, Integer> weightOfNature = new HashMap<String, Integer>(); // 词性的权重
weightOfNature.put("n", 2); //给名词的权重是2;
Map<String, String> stopNatures = new HashMap<String, String>();//停用的词性 如一些标点符号之类的;
stopNatures.put("w", ""); //
int overCount = 5; //设定超频词汇的界限 ;
Map<String, Integer> wordCount = new HashMap<String, Integer>();
for (Term term : termList) {
String word = term.word; //分词字符串
String nature = term.nature.toString(); // 分词属性;
// 过滤超频词
if (wordCount.containsKey(word)) {
int count = wordCount.get(word);
if (count > overCount) {
continue;
}
wordCount.put(word, count + 1);
} else {
wordCount.put(word, 1);
}
// 过滤停用词性
if (stopNatures.containsKey(nature)) {
continue;
}
// 2、将每一个分词hash为一组固定长度的数列.比如 64bit 的一个整数.
BigInteger t = this.hash(word);
for (int i = 0; i < this.hashbits; i++) {
BigInteger bitmask = new BigInteger("1").shiftLeft(i);
// 3、建立一个长度为64的整数数组(假设要生成64位的数字指纹,也可以是其它数字),
// 对每一个分词hash后的数列进行判断,如果是1000...1,那么数组的第一位和末尾一位加1,
// 中间的62位减一,也就是说,逢1加1,逢0减1.一直到把所有的分词hash数列全部判断完毕.
int weight = 1; //添加权重
if (weightOfNature.containsKey(nature)) {
weight = weightOfNature.get(nature);
}
if (t.and(bitmask).signum() != 0) {
// 这里是计算整个文档的所有特征的向量和
v[i] += weight;
} else {
v[i] -= weight;
}
}
}
BigInteger fingerprint = new BigInteger("0");
for (int i = 0; i < this.hashbits; i++) {
if (v[i] >= 0) {
fingerprint = fingerprint.add(new BigInteger("1").shiftLeft(i));
}
}
return fingerprint;
}
/**
* 对单个的分词进行hash计算;
* @param source
* @return
*/
private BigInteger hash(String source) {
if (source == null || source.length() == 0) {
return new BigInteger("0");
} else {
/**
* 当sourece 的长度过短,会导致hash算法失效,因此需要对过短的词补偿
*/
while (source.length() < 3) {
source = source + source.charAt(0);
}
char[] sourceArray = source.toCharArray();
BigInteger x = BigInteger.valueOf(((long) sourceArray[0]) << 7);
BigInteger m = new BigInteger("1000003");
BigInteger mask = new BigInteger("2").pow(this.hashbits).subtract(new BigInteger("1"));
for (char item : sourceArray) {
BigInteger temp = BigInteger.valueOf((long) item);
x = x.multiply(m).xor(temp).and(mask);
}
x = x.xor(new BigInteger(String.valueOf(source.length())));
if (x.equals(new BigInteger("-1"))) {
x = new BigInteger("-2");
}
return x;
}
}
/**
* 计算海明距离,海明距离越小说明越相似;
* @param other
* @return
*/
private int hammingDistance(MySimHash other) {
BigInteger m = new BigInteger("1").shiftLeft(this.hashbits).subtract(
new BigInteger("1"));
BigInteger x = this.strSimHash.xor(other.strSimHash).and(m);
int tot = 0;
while (x.signum() != 0) {
tot += 1;
x = x.and(x.subtract(new BigInteger("1")));
}
return tot;
}
public double getSemblance(MySimHash s2 ){
double i = (double) this.hammingDistance(s2);
return 1 - i/this.hashbits ;
}
public static void main(String[] args) {
String s1="simhash算法的主要思想是降维,将高维的特征向量映射成一个低维的特征向量,通过两个向量的Hamming Distance来确定文章是否重复或者高度近似。";
String s2="simhash算法的主要思想是降维,将高维的特征向量映射成一个低维的特征向量,通过两个向量的Hamming Distance来确定文章是否重复或者高度近似。";
String s3="SimHash算法是Google公司进行海量网页去重的高效算法,它通过将原始的文本映射为64位的二进制数字串,然后通过比较二进制数字串的差异进而来表示原始文本内容的差异。";
long l3 = System.currentTimeMillis();
MySimHash hash1 = new MySimHash(s1, 64);
MySimHash hash2 = new MySimHash(s2, 64);
MySimHash hash3 = new MySimHash(s3, 64);
System.out.println("======================================");
System.out.println( hash1.hammingDistance(hash2) );
System.out.println( hash2.hammingDistance(hash3) );
System.out.println( hash1.getSemblance(hash3) );
System.out.println( hash2.getSemblance(hash3) );
long l4 = System.currentTimeMillis();
System.out.println("总共耗时:"+(l4-l3)+"毫秒");
System.out.println("======================================");
}
}
运行结果:
======================================
0
20
0.6875
0.6875
总共耗时:429毫秒
======================================
Process finished with exit code 0