Terry.Li-彬

虚其心，可解天下之问；专其心，可治天下之学；静其心，可悟天下之理；恒其心，可成天下之业。
  BlogJava :: 首页 :: 新随笔 :: 联系 :: 聚合  :: 管理 ::
  143 随笔 :: 344 文章 :: 130 评论 :: 0 Trackbacks
基于词典的正向最大匹配中文分词算法，能实现中英文数字混合分词
基于词典的正向最大匹配中文分词算法，能实现中英文数字混合分词。比如能分出这样的词：bb霜、3室、乐phone、touch4、mp3、T恤

第一次写中文分词程序，欢迎拍砖。

public class MM2 
{
    private static final Log log = LogFactory.getLog(MM2.class);
    
    private static HashMap<String, Integer> dictionary = null; 
    private static final int WORD_MAX_LENGTH = 9;
    private Reader reader;
    
    static
    {
        loadDictionary();
    }
    
    public MM2(Reader reader) 
    { 
        this.reader = reader; 
    } 
    
    //切分出由中文、字母、数字组成的句子
    public ArrayList<Sentence> getSentence() throws IOException
    {   
        ArrayList<Sentence> list=new ArrayList<Sentence>();
        StringBuffer cb=new StringBuffer();
        int d=reader.read();
        int offset=0;
        boolean b=false;
        while(d>-1)
        {
            int type=Character.getType(d);
            if(type==2 || type==9 || type==5)
            {
                d=toAscii(d);
                cb.append((char)d);
            }
            else
            {
                b=true;
            }
            d=reader.read();
            if(d==-1 || b)
            {
                if(d==-1) offset++;
                b=false;
                char[] ioBuffer = new char[cb.length()];
                cb.getChars(0, cb.length(), ioBuffer, 0);
                Sentence sen=new Sentence(ioBuffer,offset-cb.length());
                list.add(sen);
                cb.setLength(0);
            }
            offset++;
        }
        return list;
    }
    
    //将句子切分出词
    public ArrayList<Token> getToken(ArrayList<Sentence> list) throws IOException
    {
        ArrayList<Token> tokenlist=new ArrayList<Token>();
        for(Sentence sen:list)
        {
            StringBuffer word = new StringBuffer();
            int offset=sen.getStartOffset();
            int bufferIndex = 0;
            char c;
            boolean b=false;
            while(bufferIndex<sen.getText().length)
            {
                offset++;
                c=sen.getText()[bufferIndex++];
                if(word.length()==0)
                    word.append(c);
                else
                {
                    String temp = (word.toString() + c).intern();
                    if(dictionary.containsKey(temp) && dictionary.get(temp)==1)
                        word.append(c);
                    else if(dictionary.containsKey(temp) && bufferIndex<sen.getText().length)
                        word.append(c);
                    else
                    {
                        bufferIndex--;
                        offset--;
                        while(word.length()>1 && dictionary.get(word.toString())!=null && dictionary.get(word.toString())==2)
                        {
                            word.deleteCharAt(word.length()-1);
                            bufferIndex--;
                            offset--;
                        }
                        b=true;
                    }
                }
                if(b || bufferIndex==sen.getText().length)
                {
                    Token token = new Token(word.toString(),offset-word.length(),offset,"word");
                    word.setLength(0);
                    tokenlist.add(token);
                    b=false;
                }
            }
        }
        return tokenlist;
    }
    
    //将相连的单个英文或数字组合成词
    public ArrayList<Token> getNewToken(ArrayList<Token> list) throws IOException
    {
        ArrayList<Token> tokenlist=new ArrayList<Token>();
        Token word=null;
        for(int i=0;i<list.size();i++)
        {
            Token t=list.get(i);
            if(t.getWord().length()==1 && Character.getType((int)t.getWord().charAt(0))!=5)
            {
                if(word==null)
                    word=t;
                else if(word.getEnd()==t.getStart())
                {
                    word.setEnd(t.getEnd());
                    word.setWord(word.getWord()+t.getWord());
                }
                else
                {
                    tokenlist.add(word);
                    word=t;
                }
            }
            else if(word!=null)
            {
                tokenlist.add(word);
                word=null;
                tokenlist.add(t);
            }
            else
                tokenlist.add(t);
        }
        if(word!=null)
            tokenlist.add(word);
        return tokenlist;
    }
    
    //双角转单角
    public static int toAscii(int codePoint) 
    {
        if((codePoint>=65296 && codePoint<=65305)    //０-９
                || (codePoint>=65313 && codePoint<=65338)    //Ａ-Ｚ
                || (codePoint>=65345 && codePoint<=65370)    //ａ-ｚ
                )
        {    
            codePoint -= 65248;
        }
        return codePoint;
    }
    
    //加载词典
    public static void loadDictionary() 
    {  
        if (dictionary == null) 
        {    
            dictionary = new HashMap<String, Integer>();    
            InputStream is = null;    
            BufferedReader br = null;            
            try
            {
                is = new FileInputStream(new File(MM2.class.getClassLoader().getResource("dictionary.txt").toURI()));
                br = new BufferedReader(new InputStreamReader(is, "UTF-8"));
                String word = null;
                while ((word = br.readLine()) != null) 
                {
                    word=word.toLowerCase();
                    if ((word.indexOf("#") == -1) && (word.length() <= WORD_MAX_LENGTH))
                    {
                        dictionary.put(word.intern(), 1);    
                        int i = word.length()-1; 
                        while(i >= 2)
                        {
                            String temp = word.substring(0, i).intern(); 
                            if (!dictionary.containsKey(temp))
                                dictionary.put(temp,2); 
                            i--;
                        }
                    }
                }
            }
            catch (Exception e) 
            {      
                log.info(e);
            }
            finally
            {
                try 
                {      
                    if(br!=null)
                        br.close();   
                    if(is!=null)
                        is.close();  
                }
                catch (IOException e)
                {     
                    log.info(e);
                }            
            } 
        } 
    }
    
    public static String[] segWords(Reader input)
    {
        ArrayList<String> list=new ArrayList<String>();
        try
        {
            MM2 f=new MM2(input);
            ArrayList<Token> tlist= f.getNewToken(f.getToken(f.getSentence()));
            for(Token t:tlist)
            {
                list.add(t.getWord());
            }
        }
        catch(IOException e)
        {
            log.info(e);
        }
        return (String[])list.toArray(new String[0]);
    }
    
    public static void main(String[] args) 
    {
        String[] cc=MM2.segWords(new StringReader("ibm商务机t60p".toLowerCase()));
        for(String c:cc)
        {
            System.out.println(c);
        }
    }
}
posted on 2011-08-05 08:34 礼物 阅读(2088) 评论(2)  编辑  收藏

评论

# re: 基于词典的正向最大匹配中文分词算法，能实现中英文数字混合分词 2013-07-25 22:09 yi
这是全的么，楼主？我导入到MyEclipse里好多错误呀，除了import包之外还有好多错，看不懂。。。  回复  更多评论
  

# re: 基于词典的正向最大匹配中文分词算法，能实现中英文数字混合分词 2013-08-22 20:01 love code
麻烦 博主把dictionary.txt发给我吧，让我学习学习
1182787467@qq.com
谢谢  回复  更多评论
  


只有注册用户登录后才能发表评论。
网站导航: