测试C#分词工具jieba.NET（续1：提取关键词及并行分词）

  jieba.NET支持通过两种算法提取文本关键词：TF-IDF算法和TextRank算法，关于这两种算法的介绍详见参考文献10-11，在jieba.NET中对应的类为TfidfExtractor和TextRankExtractor，这两个分词都都支持调用ExtractTags和ExtractTagsWithWeight函数，前者默认返回前20个关键词，后者还附带每个关键词的权重。
  j以小说《诛仙》的前7章内容为例测试这两个类的效果，首先是TfidfExtractor，示例代码及运行效果如下所示：

  //不带权重
  txtResult.Text = String.Empty;
  using (TextReader tr = File.OpenText(txtPath.Text))
  {
      TfidfExtractor extractor = new TfidfExtractor();
      var keywords=extractor.ExtractTags(tr.ReadToEnd());
      txtResult.Text = string.Join("/ ", keywords);
  }
  
  //带权重
  txtResult.Text = String.Empty;
  using (TextReader tr = File.OpenText(txtPath.Text))
   {
       TfidfExtractor extractor = new TfidfExtractor();
       IEnumerable<WordWeightPair> results = extractor.ExtractTagsWithWeight(tr.ReadToEnd());
       foreach (WordWeightPair p in results)
       {
           txtResult.Text +=  p.Word + "," + p.Weight.ToString()+ "\r\n";
       }   
    }
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

  TextRankExtractor类的示例代码及运行效果如下所示，对比可以看出这两种算法的返回结果还是存在差异：

  //不带权重
  txtResult.Text = String.Empty;
  using (TextReader tr = File.OpenText(txtPath.Text))
  {
      TextRankExtractorextractor = new TextRankExtractor();
      var keywords=extractor.ExtractTags(tr.ReadToEnd());
      txtResult.Text = string.Join("/ ", keywords);
  }
  
  //带权重
  txtResult.Text = String.Empty;
  using (TextReader tr = File.OpenText(txtPath.Text))
   {
       TextRankExtractorextractor = new TextRankExtractor();
       IEnumerable<WordWeightPair> results = extractor.ExtractTagsWithWeight(tr.ReadToEnd());
       foreach (WordWeightPair p in results)
       {
           txtResult.Text +=  p.Word + "," + p.Weight.ToString()+ "\r\n";
       }   
    }
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

  JiebaSegmenter类的CutInParallel和CutForSearchInParallel函数支持并行分词，对Cut函数相比，Cut函数的输入是单个字符串，而CutInParallel函数的输入既可以是字符串集合，也可以是单个字符串（jieba.NET会自动按行将单个字符串转换为字符串集合后再调用CutInParallel）。程序代码基本没有变化，只是将函数由Cut变成了CutInParallel。测试效果如下图所示，由于测试的文件只有几M，使用Cut函数反而比CutInParallel更快，估计是CutInParallel将单个字符串按行转换为字符串集合后再分词影响了速度。还有一定记得把文本文件转换为UTF-8或者Unicode格式，ANSI格式的文本文件分词时会卡死，不知道是什么原因。

参考文献：
[1]https://github.com/linezero/jieba.NET
[2]https://www.jianshu.com/p/6f47b670fcb0
[3]https://github.com/anderscui/jieba.NET
[4]https://github.com/JimLiu/Lucene.Net.Analysis.PanGu
[5]https://github.com/apache/lucenenet
[6]https://blog.csdn.net/lijingguang/article/details/127262360
[7]https://blog.51cto.com/u_15834522/5766716
[8]https://www.cnblogs.com/dacc123/p/8431369.html
[9]https://blog.csdn.net/wangkun9999/article/details/1574114
[10]https://baijiahao.baidu.com/s?id=1744467856047121907&wfr=spider&for=pc
[11]https://blog.csdn.net/qq_57832544/article/details/127806572