Lucene4.3进阶开发之纯阳无极（十九）

原创不易，转载请务必注明，原创地址，谢谢配合!
http://qindongliang.iteye.com/blog/2164583

Lucene内置很多的分词器工具包，几乎涵盖了全球所有的国家和地区，最近散仙，在搞多语言分词的一个处理，主要国家有西班牙，葡萄牙，德语，法语，意大利，其实这些语系都与英语非常类似，都是以空格为分割的语种。


那么首先，探讨下分词器的词形还原和词干提取的对搜索的意义？在这之前，先看下两者的概念：
词形还原（lemmatization），是把一个任何形式的语言词汇还原为一般形式（能表达完整语义），而词干提取

（stemming）是抽取词的词干或词根形式（不一定能够表达完整语义）。词形还原和词干提取是词形规范化的两类
重要方式，都能够达到有效归并词形的目的，二者既有联系也有区别

详细介绍，请参考这篇文章


在电商搜索里，词干的抽取，和单复数的还原比较重要（这里主要针对名词来讲），因为这有关搜索的查准率，和查全率的命中，如果我们的分词器没有对这些词做过处理，会造成什么影响呢？那么请看如下的一个例子？

句子： i have two cats

分词器如果什么都没有做：

这时候我们搜cat，就会无命中结果，而必须搜cats才能命中到一条数据，而事实上cat和cats是同一个东西，只不过单词的形式不一样，这样以来，如果不做处理，我们的查全率和查全率都会下降，会涉及影响到我们的搜索体验，所以stemming这一步，在某些场合的分词中至关重要。

本篇，散仙，会参考源码分析一下，关于德语分词中中如何做的词干提取，先看下德语的分词声明：

	 List<String> list=new ArrayList<String>();
		list.add("player");//这里面的词，不会被做词干抽取，词形还原
		CharArraySet ar=new CharArraySet(Version.LUCENE_43,list , true);
		//分词器的第二个参数是禁用词参数，第三个参数是排除不做词形转换，或单复数的词
		GermanAnalyzer sa=new GermanAnalyzer(Version.LUCENE_43,null,ar);

接着，我们具体看下，在德语的分词器中，都经过了哪几部分的过滤处理：

  protected TokenStreamComponents createComponents(String fieldName,
      Reader reader) {
	  //标准分词器过滤
    final Tokenizer source = new StandardTokenizer(matchVersion, reader);
    TokenStream result = new StandardFilter(matchVersion, source);
	//转小写过滤
    result = new LowerCaseFilter(matchVersion, result);
	//禁用词过滤
    result = new StopFilter( matchVersion, result, stopwords);
	//排除词过滤
    result = new SetKeywordMarkerFilter(result, exclusionSet);
    if (matchVersion.onOrAfter(Version.LUCENE_36)) {
	//在lucene3.6以后的版本，采用如下filter过滤
	  //规格化，将德语中的特殊字符，映射成英语
      result = new GermanNormalizationFilter(result);
	  //stem词干抽取，词性还原
      result = new GermanLightStemFilter(result);
    } else if (matchVersion.onOrAfter(Version.LUCENE_31)) {
	//在lucene3.1至3.6的版本中，采用SnowballFilter处理
      result = new SnowballFilter(result, new German2Stemmer());
    } else {
	//在lucene3.1之前的采用兼容的GermanStemFilter处理
      result = new GermanStemFilter(result);
    }
    return new TokenStreamComponents(source, result);
  }

OK，我们从源码中得知，在Lucene4.x中对德语的分词也做了向前和向后兼容，现在我们主要关注在lucene4.x之后的版本如何的词形转换，下面分别看下
     result = new GermanNormalizationFilter(result);
      result = new GermanLightStemFilter(result);
这两个类的功能：

package org.apache.lucene.analysis.de;

/*
 * Licensed to the Apache Software Foundation (ASF) under one or more
 * contributor license agreements.  See the NOTICE file distributed with
 * this work for additional information regarding copyright ownership.
 * The ASF licenses this file to You under the Apache License, Version 2.0
 * (the "License"); you may not use this file except in compliance with
 * the License.  You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

import java.io.IOException;

import org.apache.lucene.analysis.TokenFilter;
import org.apache.lucene.analysis.TokenStream;
import org.apache.lucene.analysis.tokenattributes.CharTermAttribute;
import org.apache.lucene.analysis.util.StemmerUtil;

/**
 * Normalizes German characters according to the heuristics
 * of the <a href="http://snowball.tartarus.org/algorithms/german2/stemmer.html">
 * German2 snowball algorithm</a>.
 * It allows for the fact that ä, ö and ü are sometimes written as ae, oe and ue.
 * 
 * [list]
 *   <li> 'ß' is replaced by 'ss'
 *   <li> 'ä', 'ö', 'ü' are replaced by 'a', 'o', 'u', respectively.
 *   <li> 'ae' and 'oe' are replaced by 'a', and 'o', respectively.
 *   <li> 'ue' is replaced by 'u', when not following a vowel or q.
 * [/list]
 * <p>
 * This is useful if you want this normalization without using
 * the German2 stemmer, or perhaps no stemming at all.
 *上面的解释说得很清楚，主要是对德文的一些特殊字母，转换成对应的英文处理
 *
 */
 
public final class GermanNormalizationFilter extends TokenFilter {
  // FSM with 3 states:
  private static final int N = 0; /* ordinary state */
  private static final int V = 1; /* stops 'u' from entering umlaut state */
  private static final int U = 2; /* umlaut state, allows e-deletion */

  private final CharTermAttribute termAtt = addAttribute(CharTermAttribute.class);
  
  public GermanNormalizationFilter(TokenStream input) {
    super(input);
  }

  @Override
  public boolean incrementToken() throws IOException {
    if (input.incrementToken()) {
      int state = N;
      char buffer[] = termAtt.buffer();
      int length = termAtt.length();
      for (int i = 0; i < length; i++) {
        final char c = buffer[i];
        switch(c) {
          case 'a':
          case 'o':
            state = U;
            break;
          case 'u':
            state = (state == N) ? U : V;
            break;
          case 'e':
            if (state == U)
              length = StemmerUtil.delete(buffer, i--, length);
            state = V;
            break;
          case 'i':
          case 'q':
          case 'y':
            state = V;
            break;
          case 'ä':
            buffer[i] = 'a';
            state = V;
            break;
          case 'ö':
            buffer[i] = 'o';
            state = V;
            break;
          case 'ü': 
            buffer[i] = 'u';
            state = V;
            break;
          case 'ß':
            buffer[i++] = 's';
            buffer = termAtt.resizeBuffer(1+length);
            if (i < length)
              System.arraycopy(buffer, i, buffer, i+1, (length-i));
            buffer[i] = 's';
            length++;
            state = N;
            break;
          default:
            state = N;
        }
      }
      termAtt.setLength(length);
      return true;
    } else {
      return false;
    }
  }
}

package org.apache.lucene.analysis.de;

/*
 * Licensed to the Apache Software Foundation (ASF) under one or more
 * contributor license agreements.  See the NOTICE file distributed with
 * this work for additional information regarding copyright ownership.
 * The ASF licenses this file to You under the Apache License, Version 2.0
 * (the "License"); you may not use this file except in compliance with
 * the License.  You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

import java.io.IOException;

import org.apache.lucene.analysis.TokenFilter;
import org.apache.lucene.analysis.TokenStream;
import org.apache.lucene.analysis.miscellaneous.SetKeywordMarkerFilter;
import org.apache.lucene.analysis.tokenattributes.CharTermAttribute;
import org.apache.lucene.analysis.tokenattributes.KeywordAttribute;

/**
 * A {@link TokenFilter} that applies {@link GermanLightStemmer} to stem German
 * words.
 * <p>
 * To prevent terms from being stemmed use an instance of
 * {@link SetKeywordMarkerFilter} or a custom {@link TokenFilter} that sets
 * the {@link KeywordAttribute} before this {@link TokenStream}.
 * 

 *
 *
 *这个类，主要做Stemmer（词干提取），而我们主要关注
 *GermanLightStemmer这个类的作用
 *
 *
 */
public final class GermanLightStemFilter extends TokenFilter {
  private final GermanLightStemmer stemmer = new GermanLightStemmer();
  private final CharTermAttribute termAtt = addAttribute(CharTermAttribute.class);
  private final KeywordAttribute keywordAttr = addAttribute(KeywordAttribute.class);

  public GermanLightStemFilter(TokenStream input) {
    super(input);
  }
  
  @Override
  public boolean incrementToken() throws IOException {
    if (input.incrementToken()) {
      if (!keywordAttr.isKeyword()) {
        final int newlen = stemmer.stem(termAtt.buffer(), termAtt.length());
        termAtt.setLength(newlen);
      }
      return true;
    } else {
      return false;
    }
  }
}

下面看下，在GermanLightStemmer中，如何做的词干提取：源码如下：

 package org.apache.lucene.analysis.de;

/*
 * Licensed to the Apache Software Foundation (ASF) under one or more
 * contributor license agreements.  See the NOTICE file distributed with
 * this work for additional information regarding copyright ownership.
 * The ASF licenses this file to You under the Apache License, Version 2.0
 * (the "License"); you may not use this file except in compliance with
 * the License.  You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

/* 
 * This algorithm is updated based on code located at:
 * http://members.unine.ch/jacques.savoy/clef/
 * 
 * Full copyright for that code follows:
 */

/*
 * Copyright (c) 2005, Jacques Savoy
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without 
 * modification, are permitted provided that the following conditions are met:
 *
 * Redistributions of source code must retain the above copyright notice, this 
 * list of conditions and the following disclaimer. Redistributions in binary 
 * form must reproduce the above copyright notice, this list of conditions and
 * the following disclaimer in the documentation and/or other materials 
 * provided with the distribution. Neither the name of the author nor the names 
 * of its contributors may be used to endorse or promote products derived from 
 * this software without specific prior written permission.
 * 
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE 
 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 
 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 
 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 
 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 
 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 
 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
 * POSSIBILITY OF SUCH DAMAGE.
 */

/**
 * Light Stemmer for German.
 * <p>
 * This stemmer implements the "UniNE" algorithm in:
 * <i>Light Stemming Approaches for the French, Portuguese, German and Hungarian Languages</i>
 * Jacques Savoy
 */
public class GermanLightStemmer {
  
  //处理特殊字符映射
  public int stem(char s[], int len) {   
    for (int i = 0; i < len; i++)
      switch(s[i]) {
        case 'ä':
        case 'à':
        case 'á':
        case 'â': s[i] = 'a'; break;
        case 'ö':
        case 'ò':
        case 'ó':
        case 'ô': s[i] = 'o'; break;
        case 'ï':
        case 'ì':
        case 'í':
        case 'î': s[i] = 'i'; break;
        case 'ü': 
        case 'ù': 
        case 'ú':
        case 'û': s[i] = 'u'; break;
      }
    
    len = step1(s, len);
    return step2(s, len);
  }
  
  
  private boolean stEnding(char ch) {
    switch(ch) {
      case 'b':
      case 'd':
      case 'f':
      case 'g':
      case 'h':
      case 'k':
      case 'l':
      case 'm':
      case 'n':
      case 't': return true;
      default: return false;
    }
  }
  //处理基于以下规则的词干抽取和缩减
  private int step1(char s[], int len) {
    if (len > 5 && s[len-3] == 'e' && s[len-2] == 'r' && s[len-1] == 'n')
      return len - 3;
    
    if (len > 4 && s[len-2] == 'e')
      switch(s[len-1]) {
        case 'm':
        case 'n':
        case 'r':
        case 's': return len - 2;
      }
    
    if (len > 3 && s[len-1] == 'e')
      return len - 1;
    
    if (len > 3 && s[len-1] == 's' && stEnding(s[len-2]))
      return len - 1;
    
    return len;
  }
  //处理基于以下规则est，er，en等的词干抽取和缩减
  private int step2(char s[], int len) {
    if (len > 5 && s[len-3] == 'e' && s[len-2] == 's' && s[len-1] == 't')
      return len - 3;
    
    if (len > 4 && s[len-2] == 'e' && (s[len-1] == 'r' || s[len-1] == 'n'))
      return len - 2;
    
    if (len > 4 && s[len-2] == 's' && s[len-1] == 't' && stEnding(s[len-3]))
      return len - 2;
    
    return len;
  }
}

具体的分析结果如下：

搜索技术交流群：324714439
大数据hadoop交流群：376932160

0，将一些德语特殊字符，替换成对应的英文表示
1，将所有词干元音还原 a ，o，i，u
ste（2）（按先后顺序，符合以下任意一项，就完成一次校验（return））
2，单词长度大于5的词，以ern结尾的，直接去掉
3，单词长度大于4的词，以em，en，es，er结尾的，直接去掉
4，单词长度大于3的词，以e结尾的直接去掉
5，单词长度大于3的词，以bs，ds，fs，gs，hs，ks，ls，ms，ns，ts结尾的，直接去掉s
step（3）（按先后顺序，符合以下任意一项，就完成一次校验（return））
6，单词长度大于5的词，以est结尾的，直接去掉
7，单词长度大于4的词，以er或en结尾的直接去掉
8，单词长度大于4的词，bst，dst，fst，gst，hst，kst，lst，mst，nst，tst，直接去掉后两位字母st

最后，结合网上资料分析，基于er，en，e，s结尾的是做单复数转换的，其他的几条规则主要是对非名词的单词，做词干抽取。


原创不易，转载请务必注明，原创地址，谢谢配合!
http://qindongliang.iteye.com/blog/2164583