static void Main(string[] args)
{
var str = "共产党是世界上最坏的政党,压迫人民,繁重的苛捐杂税,高压统治,反对人权, 自由万岁!";
Segment.Init();
var options = PanGu.Setting.PanGuSettings.Config.MatchOptions.Clone();
var parameters = PanGu.Setting.PanGuSettings.Config.Parameters.Clone();
options.FilterStopWords = false;
Segment segment = new Segment();
var words = segment.DoSegment(str, options, parameters);
StringBuilder wordsString = new StringBuilder();
foreach (WordInfo wordInfo in words)
{
if (wordInfo == null)
{
continue;
}
wordsString.AppendFormat("{0} ", wordInfo.Word);
}
str = wordsString.ToString();
Console.WriteLine(wordsString.ToString());
var zhlist = PinyinCommon.GetZhDictsList();
Console.WriteLine(PinyinCommon.GetPinyin(str, zhlist));
Console.ReadKey();
}
/// <summary>
/// 盘古分词
/// </summary>
/// <param name="str"></param>
/// <returns></returns>
public static string StrSplit(string str)
{
try
{
Stopwatch watch = new Stopwatch();
watch.Start();
Segment segment = new Segment();
ICollection<WordInfo> words = segment.DoSegment(str, _Options, _Parameters);
watch.Stop();
StringBuilder wordsString = new StringBuilder();
foreach (WordInfo wordInfo in words)
{
if (wordInfo == null || wordInfo.Word.Length <= 1)
{
continue;
}
wordsString.AppendFormat("{0},", wordInfo.Word);
}
return wordsString.ToString().TrimEnd(',');
}
catch (Exception ex)
{
return ex.Message;
}
}
private List <Fragment> GetFragments(string keywords, string content)
{
ICollection <WordInfo> keywordsWordInfos = _PanGuSegment.DoSegment(keywords);
//Make lower
foreach (WordInfo wordInfo in keywordsWordInfos)
{
if (wordInfo == null)
{
continue;
}
if (wordInfo.Word == null)
{
continue;
}
wordInfo.Word = wordInfo.Word.ToLower();
}
ICollection <WordInfo> contentWordInfos = _PanGuSegment.DoSegment(content);
//Make lower
foreach (WordInfo wordInfo in contentWordInfos)
{
if (wordInfo == null)
{
continue;
}
if (wordInfo.Word == null)
{
continue;
}
wordInfo.Word = wordInfo.Word.ToLower();
}
_Content = content;
_Selection = PickupKeywords(keywordsWordInfos, contentWordInfos);
_Selection = Optimize(_Selection);
return(GetFragments(_Selection));
}
private string DisplaySegment(string textSource)
{
//Stopwatch watch = new Stopwatch();
//watch.Start();
Segment segment = new Segment();
ICollection<WordInfo> words = segment.DoSegment(textSource);
/*watch.Stop();
labelSrcLength.Text = textBoxSource.Text.Length.ToString();
labelSegTime.Text = watch.Elapsed.ToString();
if (watch.ElapsedMilliseconds == 0)
{
labelRegRate.Text = "无穷大";
}
else
{
labelRegRate.Text = ((double)(textBoxSource.Text.Length / watch.ElapsedMilliseconds) * 1000).ToString();
}
if (checkBoxShowTimeOnly.Checked)
{
return;
}*/
StringBuilder wordsString = new StringBuilder();
foreach (WordInfo wordInfo in words)
{
if (wordInfo == null)
{
continue;
}
/* if (showPosition)
{
wordsString.AppendFormat("{0}({1},{2})/", wordInfo.Word, wordInfo.Position, wordInfo.Rank);
//if (_Options.MultiDimensionality)
//{
//}
//else
//{
// wordsString.AppendFormat("{0}({1})/", wordInfo.Word, wordInfo.Position);
//}
}*/
// else
// {
wordsString.AppendFormat("{0}/", wordInfo.Word);
// }
}
//textBoxSegwords.Text += "\n\r" + wordsString.ToString();
return wordsString.ToString();
}
protected void btnSplit_Click(object sender, EventArgs e)
{
ltlResult.Text = "";
var segment = new PanGu.Segment();
var words = segment.DoSegment(txtContent.Text);
foreach (var word in words)
{
ltlResult.Text += (word + "," + GetChsPos(word.Pos) + "<br />");
}
}
static Dictionary<string, int> ContetnWordSegment(string content)
{
Dictionary<string, int> ret = new Dictionary<string, int>();
Segment seg = new Segment();
ICollection<WordInfo> words = seg.DoSegment(content);
foreach (WordInfo w in words)
{
if (!ret.ContainsKey(w.Word))
{
ret[w.Word] = 1;
}
else ret[w.Word]++;
}
return ret;
}
public static List<string> panguDivide(String s)
{
Segment segment = new Segment();
ICollection<WordInfo> words = segment.DoSegment(s);
List<string> wordsResult = new List<string>();
foreach (WordInfo wordInfo in words)
{
if (wordInfo == null) continue;
wordsResult.Add(wordInfo.Word);
}
return wordsResult;
}
public static List <Tuple <string, double> > GetTermFreq(this string text)
{
Segment.Init();
var seg = new PanGu.Segment();
var rst = seg.DoSegment(text, new PanGu.Match.MatchOptions
{
FilterNumeric = true,
FrequencyFirst = true,
//EnglishSegment = true,
IgnoreCapital = true,
});
return(rst.Where(t => t.Word.Length > 1)
.Select(t => new Tuple <string, double>(t.Word, 1))
.ToList());
}
private static string SegmentKeyWord(VideoNode video)
{
var builder = new StringBuilder(50);
var segment = new Segment();
var words = segment.DoSegment(video.ChannelName);
foreach (var word in words)
{
if (word == null)
continue;
builder.AppendFormat("{0}^{1}.0 ", word.Word, (int)Math.Pow(3, word.Rank));
}
//if (!string.IsNullOrEmpty(video.Language[CustomArray.LanguageArray[0]].CatalogTags))
//{
// var catalogs = video.Language[CustomArray.LanguageArray[0]].CatalogTags.FormatStrToArray(SplitArray.DHArray);
// foreach (var catalog in catalogs)
// {
// builder.AppendFormat("{0}^{1}.0 ", catalog, 3);
// }
//}
//if(!string.IsNullOrEmpty(video.BKInfo.Actors))
//{
// var acs = video.BKInfo.Actors.FormatStrToArray(SplitArray.DHArray);
// foreach (var ac in acs)
// {
// builder.AppendFormat("{0}^{1}.0 ", ac, 2);
// }
//}
//if(!string.IsNullOrEmpty(video.Language[CustomArray.LanguageArray[0]].AreaTags))
//{
// var areas = video.Language[CustomArray.LanguageArray[0]].AreaTags.FormatStrToArray(SplitArray.DHArray);
// foreach (var area in areas)
// {
// builder.AppendFormat("{0}^{1}.0 ", area, 1);
// }
//}
return builder.ToString();
}
public PanGuTokenizer(TextReader input)
: base(input)
{
lock (_LockObj)
{
InitPanGuSegment();
}
_InputText = base.input.ReadToEnd();
if (string.IsNullOrEmpty(_InputText))
{
char[] readBuf = new char[1024];
int relCount = base.input.Read(readBuf, 0, readBuf.Length);
StringBuilder inputStr = new StringBuilder(readBuf.Length);
while (relCount > 0)
{
inputStr.Append(readBuf, 0, relCount);
relCount = input.Read(readBuf, 0, readBuf.Length);
}
if (inputStr.Length > 0)
{
_InputText = inputStr.ToString();
}
}
if (string.IsNullOrEmpty(_InputText))
{
_WordList = new WordInfo[0];
}
else
{
global::PanGu.Segment segment = new Segment();
ICollection<WordInfo> wordInfos = segment.DoSegment(_InputText);
_WordList = new WordInfo[wordInfos.Count];
wordInfos.CopyTo(_WordList, 0);
}
}
private void button2_Click(object sender, EventArgs e)
{
PanGu.Segment.Init();
PanGu.Segment segment = new PanGu.Segment();
ICollection <PanGu.WordInfo> words = segment.DoSegment("山东落花生花落东山，长春市长春花店");
foreach (var word in words)
{
Console.WriteLine(word.Word);
}
PanGu.HighLight.SimpleHTMLFormatter simpleHTMLFormatter = new PanGu.HighLight.SimpleHTMLFormatter("<font color=\"red\">", "</font>");
PanGu.HighLight.Highlighter highlighter = new PanGu.HighLight.Highlighter(simpleHTMLFormatter, new PanGu.Segment());
highlighter.FragmentSize = 100; // 设置每个摘要段的字符数
string keywords = "信号/道路/开通";
string content = @"高德完胜百度。我专门花了几个星期，在我所在的城市测试两个地图，高德数据不准确在少数，而百度就是家常便饭了，表现为：
已经管制一年的道路（双向变单向），百度仍然提示双向皆可走。
已经封闭数年的道路，百度仍然说是通的。
新修道路，还没有开通，百度居然让走。
有时候规划路线时明明是正确的，但是导航过程中，就出乱子，信号没问题、路线不复杂，明明是要左转，百度却叫右转。";
string abs = highlighter.GetBestFragment(keywords, content);
Console.WriteLine(abs);
}
public PanGuTokenizer(TextReader input)
: base(input)
{
termAttribute = AddAttribute<ITermAttribute>();
offsetAttribute = AddAttribute<IOffsetAttribute>();
inputText = base.input.ReadToEnd();
if (string.IsNullOrEmpty(inputText)) {
char[] readBuf = new char[1024];
int relCount = base.input.Read(readBuf, 0, readBuf.Length);
StringBuilder inputStr = new StringBuilder(readBuf.Length);
while (relCount > 0) {
inputStr.Append(readBuf, 0, relCount);
relCount = input.Read(readBuf, 0, readBuf.Length);
}
if (inputStr.Length > 0) {
inputText = inputStr.ToString();
}
}
if (string.IsNullOrEmpty(inputText)) {
words = new WordInfo[0];
}
else {
global::PanGu.Segment segment = new Segment();
ICollection<WordInfo> wordInfos = segment.DoSegment(inputText);
words = new WordInfo[wordInfos.Count];
wordInfos.CopyTo(words, 0);
}
}
private string TruncateParagraph(string paragraph) {
Segment segment = new Segment();
ICollection<WordInfo> words = segment.DoSegment(paragraph, segmentMatchOptions);
int stop = 0;
// 最短不能小于需要长度的1/2
int lowerBound = truncateLimit / 2;
foreach (WordInfo word in words) {
if (word.Position + word.Word.Length > truncateLimit) {
return stop < lowerBound ?
paragraph.Substring(0, truncateLimit) :
paragraph.Substring(0, stop);
}
stop = word.Position + word.Word.Length;
}
return paragraph;
}
static string ProcessQuery(string[] query)
{
/* 对query进行分词 */
List<string> query_words = new List<string>();
PanGu.Segment.Init();
Segment seg = new Segment();
foreach (string q in query)
{
ICollection<WordInfo> words = seg.DoSegment(q);
foreach (WordInfo wi in words)
{
if (!query_words.Contains(wi.Word))
query_words.Add(wi.Word);
}
}
/* 检索出符合条件的docIds */
MongodbAccess mongo = new MongodbAccess();
List<ObjectId> docIds = mongo.GetDocIDByQuery(query_words);
docIds = SortResult(docIds, query_words);
List<DocUrlAbstractResult> result = GetResult(docIds);
/*
foreach (DocUrlAbstractResult duar in result)
{
Console.WriteLine("{0}\n\t{1}", duar.title, duar.url);
}
* */
StringBuilder strbuilder = new StringBuilder();
foreach (DocUrlAbstractResult duar in result)
{
strbuilder.Append(String.Format(
"<li><div><span><a href='{2}' target='_blank' class='link'>{0}<a></span><br/><span class='abstract'>{1}</span><br/><span class='url'>{2}</span></div></li>",
duar.title, duar.abst, duar.url));
}
if (strbuilder.Length == 0)
{
strbuilder.Append("No pages mathch the query.");
}
return "<ol>" + strbuilder.ToString() + "</ol>";
}
/// <summary>
/// 根据文章标题智能解析关键字(或标签)
/// </summary>
/// <param name="title"></param>
/// <returns></returns>
private static ICollection<WordInfo> TitleToKeywordWordInfos(string title)
{
PanGu.Segment segment = new Segment();
PanGu.Match.MatchOptions matchOptions = new PanGu.Match.MatchOptions();
//中文人名识别
matchOptions.ChineseNameIdentify = false;
//词频优先
matchOptions.FrequencyFirst = false;
//多元分词
matchOptions.MultiDimensionality = false;
//英文多元分词，这个开关，会将英文中的字母和数字分开
matchOptions.EnglishMultiDimensionality = false;
//过滤停用词
matchOptions.FilterStopWords = true;
//忽略空格、回车、Tab
matchOptions.IgnoreSpace = true;
//强制一元分词
matchOptions.ForceSingleWord = false;
//繁体中文开关
matchOptions.TraditionalChineseEnabled = false;
//同时输出简体和繁体
matchOptions.OutputSimplifiedTraditional = false;
//未登录词识别
matchOptions.UnknownWordIdentify = false;
//过滤英文，这个选项只有在过滤停用词选项生效时才有效
matchOptions.FilterEnglish = true;
//过滤数字，这个选项只有在过滤停用词选项生效时才有效
matchOptions.FilterNumeric = true;
//忽略英文大小写
matchOptions.IgnoreCapital = false;
//英文分词
matchOptions.EnglishSegment = false;
//同义词输出 （同义词输出功能一般用于对搜索字符串的分词，不建议在索引时使用）
matchOptions.SynonymOutput = false;
//通配符匹配输出 （）
matchOptions.WildcardOutput = false;
//对通配符匹配的结果分词
matchOptions.WildcardSegment = false;
PanGu.Match.MatchParameter matchParameter = new PanGu.Match.MatchParameter();
//未登录词权值
matchParameter.UnknowRank = 1;
//最匹配词权值
matchParameter.BestRank = 5;
//次匹配词权值
matchParameter.SecRank = 3;
//再次匹配词权值
matchParameter.ThirdRank = 2;
//强行输出的单字的权值
matchParameter.SingleRank = 1;
//数字的权值
matchParameter.NumericRank = 1;
//英文词汇权值
matchParameter.EnglishRank = 5;
//英文词汇小写的权值
matchParameter.EnglishLowerRank = 3;
//英文词汇词根的权值
matchParameter.EnglishStemRank = 2;
//符号的权值
matchParameter.SymbolRank = 1;
//强制同时输出简繁汉字时，非原来文本的汉字输出权值。 比如原来文本是简体，这里就是输出的繁体字的权值，反之亦然。
matchParameter.SimplifiedTraditionalRank = 1;
//同义词权值
matchParameter.SynonymRank = 1;
//通配符匹配结果的权值
matchParameter.WildcardRank = 1;
//过滤英文选项生效时，过滤大于这个长度的英文
matchParameter.FilterEnglishLength = 0;
//过滤数字选项生效时，过滤大于这个长度的数字
matchParameter.FilterNumericLength = 0;
//用户自定义规则的配件文件名
matchParameter.CustomRuleAssemblyFileName = string.Empty;
//用户自定义规则的类的完整名，即带名字空间的名称
matchParameter.CustomRuleFullClassName = string.Empty;
//冗余度
matchParameter.Redundancy = 0;
return segment.DoSegment(title, matchOptions, matchParameter);
}
public ICollection<WordInfo> SegmentToWordInfos(string str)
{
if (string.IsNullOrEmpty(str)) {
return new LinkedList<WordInfo>();
}
Segment segment = new Segment();
return segment.DoSegment(str);
}
private List<string> GetSegmentWords(List<string> content)
{
Segment segment = new Segment();
List<string> segmentWords = new List<string>();
foreach (string s in content)
{
ICollection<WordInfo> words = segment.DoSegment(s);
List<string> sWords = new List<string>();
foreach (WordInfo wordInfo in words)
{
if (wordInfo == null)
{
continue;
}
sWords.Add(wordInfo.Word);
}
segmentWords.AddRange(sWords);
}
return segmentWords;
}
private static string segment(string s)
{
Segment segment = new Segment();
ICollection<WordInfo> words = segment.DoSegment(s);
StringBuilder wordsString = new StringBuilder();
foreach (WordInfo wordInfo in words)
{
if (wordInfo == null)
{
continue;
}
wordsString.AppendFormat("{0}\t", wordInfo.Word);
}
return wordsString.ToString();
}
public PanGuTokenizer(TextReader input)
: base(input)
{
_inputText = base.input.ReadToEnd();
if (string.IsNullOrEmpty(_inputText))
{
char[] readBuf = new char[1024];
int relCount = base.input.Read(readBuf, 0, readBuf.Length);
StringBuilder inputStr = new StringBuilder(readBuf.Length);
while (relCount > 0)
{
inputStr.Append(readBuf, 0, relCount);
relCount = input.Read(readBuf, 0, readBuf.Length);
}
if (inputStr.Length > 0)
{
_inputText = inputStr.ToString();
}
}
_isFlag = AnalyzInput();
if (!_isFlag) //盘古分词
{
global::PanGu.Segment segment = new Segment();
var wordInfos = segment.DoSegment(_inputText);
foreach (var wi in wordInfos)
{
var list = SpellUtils.GetSpellSegment(wi.Word);
if (list != null)
_segmentList.AddRange(list);
}
_wordList = new List<WordInfo>(wordInfos);
}
else
{
if (_inputText.EndsWith(_indexFlag)) //逗号空格分词
{
string[] sources = _inputText.Replace(_indexFlag, "").Split(_splitFlag, StringSplitOptions.RemoveEmptyEntries);
foreach (string source in sources)
{
_segmentList.Add(source);
var spells = SpellUtils.GetSpellSegment(source);
if (spells != null)
_segmentList.AddRange(spells);
}
}
else if (_inputText.EndsWith(_indexCnName)) //拼音首字母分词
{
string source = _inputText.Replace(_indexCnName, "");
_wordList = SpellUtils.GetCnSegment(source);
}
else //flag分词
{
string source = _inputText.Replace("|", "");
for (int i = 0; i < source.Length; i++)
{
string first = source[i].ToString();
_segmentList.Add(first);
int f = 1;
for (int j = i + 1; j < source.Length; j++)
{
string s = string.Format("{0}|{1}", first, source[j]);
_segmentList.Add(s);
first = s;
if (j == source.Length - 1)
{
f++;
j = i + f - 1;
first = source[i].ToString();
}
}
}
}
}
}
/// <summary>
/// 处理客户端输入
/// </summary>
public static string SegmentKeyWord(string keyword)
{
#region Deleted
//var cnbuilder = new StringBuilder();
//var enbuilder = new StringBuilder ();
//foreach (char k in keyword)
//{
// if (SpellUtils.IsChinese(k.ToString()))
// {
// cnbuilder.Append(k);
// enbuilder.Append(" ");
// }
// else
// {
// enbuilder.Append(k);
// }
//}
//var rbuilder = new StringBuilder();
//if (enbuilder.Length > 0)
//{
// var enkeys = enbuilder.ToString().Split(_splitWord, StringSplitOptions.RemoveEmptyEntries);
// foreach (string enkey in enkeys)
// {
// rbuilder.AppendFormat("{0}^{1}.0 ", enkey, 1);
// //rbuilder.AppendFormat("{0}*^{1}.0", enkey, 1);
// }
//}
//if (cnbuilder.Length > 0)
//{
// var segment = new Segment();
// var words = segment.DoSegment(cnbuilder.ToString(), _option);
// foreach (var word in words)
// {
// if (word == null)
// continue;
// rbuilder.AppendFormat("{0}^{1}.0 ", word.Word, (int)Math.Pow(3, word.Rank));
// }
//}
//return rbuilder.ToString().Trim();
#endregion
var builder = new StringBuilder(20);
var segment = new Segment();
var words = segment.DoSegment(keyword, _option);
foreach (var word in words)
{
if (word == null)
continue;
builder.AppendFormat("{0}^{1}.0 ", word.Word, (int)Math.Pow(3, word.Rank));
}
return builder.ToString();
}
public void SpellTest()
{
LuceneEngine engine = new LuceneEngine();
engine.Init();
var firsts = SpellUtils.GetCnSegment("NBA常规赛-快船vs凯尔特人");
var spells = SpellUtils.GetSpellSegment("战重警和");
var segment = new Segment();
var collection = segment.DoSegment("国际足球100509K联赛釜山-大田");
var list = new List<string>();
foreach (WordInfo word in collection)
{
if (word == null)
continue;
list.Add(word.Word);
}
//var list = SpellUtils.GetSpellSegment("战警");
string sss = "san国yan义";
string ssss = Synacast.LuceneNetSearcher.Searcher.Searcher.SegmentKeyWord(sss);
string text = "重庆";
TextReader tr = new StringReader(text);
PanGuTokenizer ct = new PanGuTokenizer(tr);
int end = 0;
Lucene.Net.Analysis.Token t;
string ss = string.Empty;
while (end < text.Length)
{
t = ct.Next();
end = t.EndOffset();
ss = ss + t.TermText() + "/ ";
}
}