Elasticsearch的分析器

最新推荐文章于 2024-05-22 18:27:07 发布

weixin_34353714

最新推荐文章于 2024-05-22 18:27:07 发布

阅读量85

点赞数

CC 4.0 BY-SA版权

原文链接：http://www.cnblogs.com/shaner/p/6340925.html

本文深入解析了Elasticsearch中的分词机制，包括内置分析器、分词器及过滤器的工作原理，并通过.NET Nest代码示例展示了如何自定义分词器与过滤器。

概念解释：　　

　　全文搜索引擎会用某种算法对要建索引的文档进行分析，从文档中提取出若干Token(词元)，这些算法称为Tokenizer(分词器)，这些Token会被进一步处理，比如转成小写等，这些处理算法被称为Token Filter(词元处理器), 被处理后的结果被称为Term(词)，文档中包含了几个这样的Term被称为Frequency(词频)。引擎会建立Term和原文档的Inverted Index(倒排索引)，这样就能根据Term很快到找到源文档了。文本被Tokenizer处理前可能要做一些预处理，比如去掉里面的HTML标记，这些处理的算法被称为Character Filter(字符过滤器)，这整个的分析算法被称为Analyzer(分析器)。
　　一个分析器是3个顺序执行的组件的结合（字符过滤器Character Filter（0/N个），分词器Tokenizer(1个)，词元处理器Token Filter（0/N个））：
　　　　1)Character Filters的作用就是对文本进行一个预处理，例如把文本中所有“&”换成“and”，把“?”去掉等等操作
　　　　2)Tokenizer的作用是进行分词，例如，“tom is a good doctor”,分词器Tokenizer会将这个文本分出很多词来：“tom”、“is”、“a”、“good”、“doctor”
　　　　3)Token Filter的作用就是对分词出来的词元进行处理，得到term，例如tom可能被处理成"t","o","m",最后得出来的结果集合，就是最终的集合

ES中关于内置分析器，分词器的定义在： https://www.elastic.co/guide/en/elasticsearch/reference/current/analysis.html

常用分词器介绍：http://jingyan.baidu.com/article/cbcede071e1b9e02f40b4d19.html

http://blog.youkuaiyun.com/i6448038/article/details/51614220

http://blog.youkuaiyun.com/i6448038/article/details/51509439

1 ES内置的分析器

analyzer	logical name	description
standard analyzer	standard	standard tokenizer, standard filter, lower case filter, stop filter
simple analyzer	simple	lower case tokenizer
stop analyzer	stop	lower case tokenizer, stop filter
keyword analyzer	keyword	不分词，内容整体作为一个token(not_analyzed)
pattern analyzer	whitespace	正则表达式分词，默认匹配\W+
language analyzers	lang	各种语言
snowball analyzer	snowball	standard tokenizer, standard filter, lower case filter, stop filter, snowball filter
custom analyzer	custom	一个Tokenizer, 零个或多个Token Filter, 零个或多个Char Filter

2 ES内置的分词器

tokenizer	logical name	description
standard tokenizer	standard
edge ngram tokenizer	edgeNGram
keyword tokenizer	keyword	不分词
letter analyzer	letter	按单词分
lowercase analyzer	lowercase	letter tokenizer, lower case filter
ngram analyzers	nGram
whitespace analyzer	whitespace	以空格为分隔符拆分
pattern analyzer	pattern	定义分隔符的正则表达式
uax email url analyzer	uax_url_email	不拆分url和email
path hierarchy analyzer	path_hierarchy	处理类似`/path/to/somthing`样式的字符串

ES内置的过滤器：

token filter	logical name	description
standard filter	standard
ascii folding filter	asciifolding
length filter	length	去掉太长或者太短的
lowercase filter	lowercase	转成小写
ngram filter	nGram
edge ngram filter	edgeNGram
porter stem filter	porterStem	波特词干算法
shingle filter	shingle	定义分隔符的正则表达式
stop filter	stop	移除 stop words
word delimiter filter	word_delimiter	将一个单词再拆成子分词
stemmer token filter	stemmer
stemmer override filter	stemmer_override
keyword marker filter	keyword_marker
keyword repeat filter	keyword_repeat
kstem filter	kstem
snowball filter	snowball
phonetic filter	phonetic	插件
synonym filter	synonyms	处理同义词
compound word filter	dictionary_decompounder, hyphenation_decompounder	分解复合词
reverse filter	reverse	反转字符串
elision filter	elision	去掉缩略语
truncate filter	truncate	截断字符串
unique filter	unique
pattern capture filter	pattern_capture
pattern replace filte	pattern_replace	用正则表达式替换
trim filter	trim	去掉空格
limit token count filter	limit	限制token数量
hunspell filter	hunspell	拼写检查
common grams filter	common_grams
normalization filter	arabic_normalization, persian_normalization

ES内置的字符过滤器：

character filter	logical name	description
mapping char filter	mapping	根据配置的映射关系替换字符
html strip char filter	html_strip	去掉HTML元素
pattern replace char filter	pattern_replace	用正则表达式处理字符串

自定义一个拼音分析器的命令：

/test  --建索引命令  post
/test/_settings   --修改索引setting的命令  put
{    --设置参数    
  "index": {
    "analysis": {
      "analyzer": {   --自定义分析器
        "pinyinanalyzer": {
          "tokenizer": "lishuai_pinyin", --使用下面自定义的分词器
          "filter": [  --使用过滤器
            "lowercase",
        "mynGramFilter" --自定义的过滤器
          ]
        }
      },
      "tokenizer": {  --自定义分词器
        "lishuai_pinyin": {
          "type": "pinyin", --对应plugin中分词器的名称
          "first_letter": "prefix", --前缀分词器
          "padding_char": "" 
        }
      },
      "filter": { --自定义过滤器
        "mynGramFilter": { --如果词的长度大于最短词长度则分词，则依次分成最小长度递进到最大长度的词。
          "type": "nGram",
          "min_gram": "2", --词语2个以上才会分词
          "max_gram": "5"  --词语最多分成长度为5个的词语
        }
      }
    }
  }
}
"padding_char": " "  --分此后 每个词元以指定符号区分：
"first_letter": "prefix"  --是否分出每个字的首字母 ，如果配置了，词语刘德华会被分成ldh liu de hua

mappings: post
/test/product/_mapping  --为索引设置type 指定mapping
{
  "product": {
    "properties": {
      "name": {
        "type": "string",
        "store": "yes",
        "index": "analyzed",
        "term_vector": "with_positions_offsets"  --表示额外索引的当前和结束位置 
        "analyzer": "pinyinanalyzer"
      }
    }
  }
}

ES命令

.NET nest实现代码：

1）es的mapping模型和自定义的分词插件

 1     [ElasticsearchType(Name = "associationtype")]
 2     public class AssociationInfo
 3     {
 4          [String(Store = true, Index = FieldIndexOption.Analyzed, TermVector = TermVectorOption.WithPositionsOffsets, Analyzer = "mypinyinanalyzer")]
 5         public string keywords { get; set; }
 6          [String(Store = true, Index = FieldIndexOption.NotAnalyzed)]
 7         public string webSite { get; set; }
 8     }
 9 
10     /// <summary>
11     /// 自定义拼音分词器
12     /// </summary>
13     public class PinYinTokenizer : ITokenizer
14     {
15 
16 
17         public string Type
18         {
19             get { return "pinyin"; }
20         }
21 
22         public string Version
23         {
24             get
25             {
26                 return string.Empty;
27             }
28             set
29             {
30                 throw new NotImplementedException();
31             }
32         }
33 
34         public string first_letter
35         {
36             get { return "prefix"; } //开启前缀匹配
37         }
38 
39         public string padding_char {
40             get { return " "; }
41         }
42 
43         public bool keep_full_pinyin
44         {
45             get { return true; }
46 
47         }
48 
49     }
50     /// <summary>
51     /// 自定义拼音词元过滤器
52     /// </summary>
53     public class PinYinTF : ITokenFilter
54     {
55         /// <summary>
56         /// 连词过滤器
57         /// </summary>
58         public string Type
59         {
60             get { return "nGram"; }
61         }
62 
63         public string Version
64         {
65             get
66             {
67                 return "";
68             }
69             set
70             {
71                 throw new NotImplementedException();
72             }
73         }
74 
75         public int min_gram
76         {
77             get
78             {
79                 return 2;
80             }
81         }
82         public int max_gram
83         {
84             get
85             {
86                 return 5;
87             }
88         }
89     }

View Code

2）创建索引

 1         /// <summary>
 2         /// 创建索引
 3         /// </summary>
 4         public void CreateIndex()
 5         {
 6             //使用es默认的过滤器
 7             //List<string> filters = new List<string>() { "word_delimiter" };    // word_delimiter 将一个单词再拆成子分词,nGram 连词过滤器
 8             //自定义的分词器和过滤器
 9             var pinYinTokenizer = new PinYinTokenizer();
10             var pinYinTF = new PinYinTF();
11             //一个自定义分析器需要设置1个分词器，0/n个Token Filters和0到多个Char Filters
12             var create = new CreateIndexDescriptor("shuaiindex").Settings(s => s.Analysis(a => //自定义一个分析器
13                 a.Tokenizers(ts => ts.UserDefined("pinyintoken", pinYinTokenizer))  //1 为分析器创建一个用户自定义的pinyin分词器,当然还可以创建自定义TokenFilters表征过滤器，自定义CharFilters字符过滤器
14                  .TokenFilters(tf => tf.UserDefined("pinyintf", pinYinTF))
15                  .Analyzers(c => c.Custom("mypinyinanalyzer", f =>  //2设置自定义分析器的名称
16                      f.Tokenizer("pinyintoken").Filters("pinyintf"))) //3 为分析器设置1个刚才自定义的pinyin分词器和一个自定义的词元过滤器
17                 ))
18                 //设置mapping
19                 .Mappings(map => map.Map<AssociationInfo>(m => m.AutoMap()));
20 
21             var client = ElasticSearchCommon.GetInstance().GetElasticClient();
22             var rs = client.CreateIndex(create);
23             client.IndexMany(CreateAssociationInfo(), "shuaiindex", "associationtype");
24         }
25         public List<AssociationInfo> CreateAssociationInfo()
26         {
27             List<AssociationInfo> AssociationInfos = new List<AssociationInfo>();
28             AssociationInfos.Add(new AssociationInfo() { keywords = "牛奶牛肉", webSite = "1" });
29             AssociationInfos.Add(new AssociationInfo() { keywords = "小牛", webSite = "1" });
30             AssociationInfos.Add(new AssociationInfo() { keywords = "果苹", webSite = "1" });
31             AssociationInfos.Add(new AssociationInfo() { keywords = "牛肉", webSite = "1" });
32             AssociationInfos.Add(new AssociationInfo() { keywords = "牛肉干", webSite = "1" });
33             AssociationInfos.Add(new AssociationInfo() { keywords = "牛奶", webSite = "1" });
34             AssociationInfos.Add(new AssociationInfo() { keywords = "肥牛", webSite = "1" }); 
35             AssociationInfos.Add(new AssociationInfo() { keywords = "牛头", webSite = "1" });
36             AssociationInfos.Add(new AssociationInfo() { keywords = "苹果", webSite = "1" }); 
37             AssociationInfos.Add(new AssociationInfo() { keywords = "车厘子", webSite = "1" });
38             return AssociationInfos;
39         
40         }

View Code

3）查询

 1         public List<AssociationInfo> query(string key)
 2         {
 3             QueryContainer query = new QueryContainer();
 4             //query=Query<AssociationInfo>.Prefix(s => s.Field(f => f.keywords).Value(key));
 5             query = Query<AssociationInfo>.Term("keywords", key);
 6             //match 会把词语拆开，匹配每一项
 7             //query = Query<AssociationInfo>.Match(m => m
 8             //        .Field(p => p.keywords)
 9             //        .Query(key)
10             //        );
11             var client = ElasticSearchCommon.GetInstance().GetElasticClient();
12 
13             SearchRequest request = new SearchRequest("shuaiindex", "associationtype");
14             request.Query = query;
15             //request.Analyzer = "mypinyinanalyzer";
16             ISearchResponse<AssociationInfo> response = client.Search<AssociationInfo>(request);
17             List<AssociationInfo> re = new List<AssociationInfo>();
18 
19             if (response.Documents.Count() > 0)
20             {
21                 re = response.Documents.ToList();
22             }
23             return re;             
24         }