/// <summary>
/// 预分词
/// </summary>
/// <param name="str">要分词的句子</param>
/// <returns>预分词后的字符串输出</returns>
private List <String> PreSegment(String str)
{
ArrayList initSeg = new ArrayList();
if (!CRegex.GetSingleMatchStrings(str, PATTERNS, true, ref initSeg))
{
return(new List <String>());
}
List <String> retWords = new List <String>();
int i = 0;
_ExtractWords.MatchDirection = MatchDirection;
while (i < initSeg.Count)
{
String word = (String)initSeg[i];
if (word == "")
{
word = " ";
}
if (i < initSeg.Count - 1)
{
bool mergeOk = false;
if (((word[0] >= '0' && word[0] <= '9') || (word[0] >= '０' && word[0] <= '９')) &&
((word[word.Length - 1] >= '0' && word[word.Length - 1] <= '9') ||
(word[word.Length - 1] >= '０' && word[word.Length - 1] <= '９'))
)
{
//合并浮点数
word = MergeFloat(initSeg, i, ref i);
mergeOk = true;
}
else if ((word[0] >= 'a' && word[0] <= 'z') ||
(word[0] >= 'A' && word[0] <= 'Z')
)
{
//合并成英文专业名词
String specialEnglish = MergeEnglishSpecialWord(_ExtractWords, initSeg, i, ref i);
if (specialEnglish != null)
{
InsertWordToArray(specialEnglish, retWords);
continue;
}
//合并邮件地址
if ((String)initSeg[i + 1] != "")
{
if (((String)initSeg[i + 1])[0] == '@')
{
word = MergeEmail(initSeg, i, ref i);
mergeOk = true;
}
}
}
if (mergeOk)
{
InsertWordToArray(word, retWords);
continue;
}
}
if (word[0] < 0x4e00 || word[0] > 0x9fa5)
{
//英文或符号，直接加入
InsertWordToArray(word, retWords);
}
else
{
List <T_WordInfo> words = _ExtractWords.ExtractFullTextMaxMatch(word);
int lastPos = 0;
bool lstIsName = false; //前一个词是人名
foreach (T_WordInfo wordInfo in words)
{
if (lastPos < wordInfo.Position)
{
/*
* String unMatchWord = word.Substring(lastPos, wordInfo.Position - lastPos);
*
* InsertWordToArray(unMatchWord, retWords);
*/
//中间有未匹配词，将单个字逐个加入
for (int j = lastPos; j < wordInfo.Position; j++)
{
InsertWordToArray(word[j].ToString(), retWords);
}
}
lastPos = wordInfo.Position + wordInfo.Word.Length;
//统计中文姓名的后缀
if (AutoStudy && lstIsName)
{
T_DictStruct wordDict = (T_DictStruct)wordInfo.Tag;
if ((wordDict.Pos & (int)T_POS.POS_A_NR) == 0)
{
_MatchNameRule.AddBefore(wordInfo.Word);
}
lstIsName = false;
}
//统计中文姓名的前缀
//如总统，主席等
if ((((T_DictStruct)wordInfo.Tag).Pos & (int)T_POS.POS_A_NR) != 0)
{
if (wordInfo.Word.Length > 1 && wordInfo.Word.Length <= 4 && retWords.Count > 0 && AutoStudy && !lstIsName)
{
T_DictStruct wordDict = (T_DictStruct)wordInfo.Tag;
_MatchNameRule.AddBefore(retWords[retWords.Count - 1]);
}
lstIsName = true;
}
InsertWordToArray(wordInfo.Word, retWords);
}
if (lastPos < word.Length)
{
//尾部有未匹配词，将单个字逐个加入
for (int j = lastPos; j < word.Length; j++)
{
InsertWordToArray(word[j].ToString(), retWords);
}
//InsertWordToArray(word.Substring(lastPos, word.Length - lastPos), retWords);
}
}
i++;
}
return(retWords);
}
