private List <string> WordSplitResult(string strWords)
{
List <string> result = new List <string>();
IEnumerable <string> segments;
switch (comboBoxCutMode.SelectedIndex)
{
case 0:
segments = segmenter.Cut(strWords);
break;
case 1:
segments = segmenter.CutForSearch(strWords);
break;
case 2:
var idf = new TfidfExtractor();
segments = idf.ExtractTags(strWords, 20, Constants.NounAndVerbPos);
break;
default:
var textRank = new TextRankExtractor();
segments = textRank.ExtractTags(strWords, 20, Constants.NounAndVerbPos);
break;
}
foreach (string str in string.Join(" ", segments).Split(' '))
{
if (!stopwordsList.Contains(str))
{
result.Add(str);
}
}
return(result);
}
public void TestTextRankExtractorWithWeight()
{
var s = "此外，公司拟对全资子公司吉林欧亚置业有限公司增资4.3亿元，增资后，吉林欧亚置业注册资本由7000万元增加到5亿元。吉林欧亚置业主要经营范围为房地产开发及百货零售等业务。目前在建吉林欧亚城市商业综合体项目 2013年，实现营业收入0万元，实现净利润-139.13万元。";
var extractor = new TextRankExtractor();
var result = extractor.ExtractTagsWithWeight(s);
foreach (var tag in result)
{
Console.WriteLine("({0}, {1})", tag.Word, tag.Weight);
}
}
//计算权值 并生成js文件 用于生成词云
static public void WordWeightCal()
{
string wordcloud = @"D:\data\wordcloud\";
DirectoryInfo theFolder = new DirectoryInfo(@"D:\data\news\");
if (!Directory.Exists(wordcloud))
{
Directory.CreateDirectory(wordcloud);
//Console.WriteLine("create");
}
DirectoryInfo[] dirInfo = theFolder.GetDirectories();
var threads = new List <Thread>();
foreach (var item in dirInfo)
{
var thread = new Thread(() =>
{
//Console.WriteLine(item.Name);
FileInfo[] fileInfo = item.GetFiles("*.txt", SearchOption.AllDirectories);
string text = "";
foreach (var file in fileInfo)
{
using (StreamReader sr = new StreamReader(file.FullName))
{
text += sr.ReadToEnd();
}
}
//var extractor = new TfidfExtractor();
var extractor = new TextRankExtractor();
var a = extractor.ExtractTagsWithWeight(text);
using (StreamWriter sw = new StreamWriter(wordcloud + item.Name + ".js", false))
{
sw.WriteLine("words = [");
foreach (var result in a)
{
sw.WriteLine("{ text: '" + result.Word + "', size: " + result.Weight + ", href: 'https://www.baidu.com/s?wd=" + result.Word + "' },");
}
sw.WriteLine("];");
}
lock (locker)
{
c++;
}
//Console.WriteLine(c);
});
threads.Add(thread);
thread.Start();
}
foreach (var t in threads)
{
t.Join();
}
Console.WriteLine("分词完成！");
}
public void TestTextRankExtractorWithoutWeights()
{
var s = "此外，公司拟对全资子公司吉林欧亚置业有限公司增资4.3亿元，增资后，吉林欧亚置业注册资本由7000万元增加到5亿元。吉林欧亚置业主要经营范围为房地产开发及百货零售等业务。目前在建吉林欧亚城市商业综合体项目 2013年，实现营业收入0万元，实现净利润-139.13万元。";
var extractor = new TextRankExtractor();
var result = extractor.ExtractTags(s);
Assert.That(result, Contains.Item("吉林"));
result = extractor.ExtractTags(s, allowPos: new [] { "n" });
Assert.That(result, Is.Not.Contains("吉林"));
Assert.That(result, Is.Not.Contains("实现"));
}
void init()
{
if (Extractor == null)
{
TextRankExtractor = new TextRankExtractor();
TfidfExtractor = new TfidfExtractor();
}
if (Algorithm == ExtractAlgorithm.TextRank)
{
Extractor = TextRankExtractor;
}
if (Algorithm == ExtractAlgorithm.TF_IDF)
{
Extractor = TfidfExtractor;
}
}
/// <summary>
/// 获取num个核心句
/// </summary>
/// <param name="text">文本</param>
/// <param name="num">核心句数</param>
/// <param name="type">抽取类型</param>
public void GetList(string text, int num, int type)
{
keywordList.Clear();
//获取核心关键词列表
switch (type)
{
case 1:
{
TfidfExtractor te = new TfidfExtractor();
keywordList = te.ExtractTags(text, num).ToList();
}
break;
case 2:
{
TextRankExtractor te = new TextRankExtractor();
keywordList = te.ExtractTags(text, num).ToList();
}
break;
}
AllsentenceList.Clear();
keySentenceList.Clear();
//将文章拆为句子列表，并分词
text = text.Replace(Environment.NewLine.ToString(), " 。");
//text = text.Replace(" ", "");
AllsentenceList = text.Split('。', '？').Where(x => !string.IsNullOrEmpty(x) && x != "undefined").Select(x => x.Trim()).ToList();
List <Sentence> temp = new List <Sentence>();
for (int i = 0; i < AllsentenceList.Count; i++)
{
AllsentenceList[i] = AllsentenceList[i] + "。";
var sentence = segmenter.Cut(AllsentenceList[i]);
Sentence v = new Sentence();
v.Sen = string.Join(" ", sentence);
v.Index = i;
temp.Add(v);
}
GetSentenceList(keywordList, temp);
}