import java.io.BufferedReader;
import java.io.File;
import java.io.FileInputStream;
import java.io.IOException;
import java.io.InputStreamReader;
import java.io.RandomAccessFile;
import java.util.StringTokenizer;

/*
 * 文本格式：已分词的中文文本，空格分割。有若干行，每行为一个段落。
 * 功能：遍历文档，逐个返回词语。
 * 两种模式：
 * 1 到文档末尾后，结束
 * 2 到文档末尾后，从头再读。
/
public class WordReader 
{
	static final int normalMode = 0;
	static final int againMode = 1;
	
	int currentMode = 0;
	//BufferedReader br=null;
	RandomAccessFile raf= null;
	StringTokenizer tokenizer = null;
	String nextWord=null;
	int currentLine = 0;
	int allCounts = 0;
	public  WordReader(String fileName) throws IOException
	{
		File file=new File(fileName);
		//br=new BufferedReader(new InputStreamReader(new FileInputStream(file),"utf-8"));
		raf = new RandomAccessFile(file,"r") ;
	}	
	
	private boolean hasNextWord() throws IOException
	{
		if( tokenizer!=null && tokenizer.hasMoreTokens())
		{
			nextWord = tokenizer.nextToken();
			return true;
		}
		else 
		{
			String line=raf.readLine();		
			if(line == null)
			{
				if(currentMode == normalMode)
					return false;
				else //从头再来
				{
					raf.seek(0);
					return hasNextWord();//递归
				}
			}	
			tokenizer = null;		
			line = new String(line.getBytes("iso8859-1"),"utf-8");
			tokenizer= new StringTokenizer(line," ");
			return hasNextWord();//递归			
		}
	}
	
	private String getNextWord() throws IOException
	{
		if(nextWord != null)
		{
			
			String word = nextWord;
			nextWord = null;
			allCounts ++;
			return word;
		}
		else if(hasNextWord())
		{
			return getNextWord();
		}
		else return null;
	}
		
	public static void main(String[] args) throws IOException 
	{
		// TODO Auto-generated method stub
		WordReader wordReader = new WordReader("/home/linger/sources/ParaModel/electronic_seg.txt");
		wordReader.currentMode = WordReader.againMode;
		//while(wordReader.hasNextWord())//共10329309个词
		for(int i=0;i<10329319;i++)//文本从头读
		{
			System.out.println(wordReader.getNextWord());
		}
		System.out.println(wordReader.allCounts);

	}

}

用randomaccessfile类很容易操作文件指针。

但是遇到中文乱码问题，参考了这里http://blog.chinaunix.net/uid-15490606-id-211958.html，解决了。

line = new String(line.getBytes("iso8859-1"),"utf-8");

对编码不是很精通，有时见看看这个http://blog.sina.com.cn/s/blog_673c81990100t1lc.html。

java读取中文分词工具