何谓滑动窗口分词:
比如原句:woods data hadoop
分词后为:
woods
woods data
woods data hadoop
data
data hadoop
hadoop
创建索引,自定义分词方式:
使用ik_smart中文分词,并使用shingle过滤器(滑动窗口模式)
put test
{
"settings": {
"index": {
"analysis": {
"analyzer": {
"shingle1": {
"type":"custom",
"tokenizer":"ik_smart",
"filter":["shingle-filter"]
}
},
"filter": {
"shingle-filter": {
"type":"shingle",
"min_shingle_size":2, -- 设置最小和最大的滑动窗口尺寸
"max_shingle_size":10,
"output_unigrams":true -- 是否保留原始的单个分词,默认true:保留
}
}
}
}
},
"mappings": {
"my_type_name": {
"dynamic": false,
"_all": {
"enabled": false
},
"properties": {
"country": {
"type": "keyword"
},
"age": {
"type": "integer"
}
}
}
}
}
查询:
post test/_analyze
{
"analyzer": "shingle1",
"text": "飞利浦吸尘器家用自营"
}
结果:
{
"tokens": [
{
"token": "飞利浦",
"start_offset": 0,
"end_offset": 3,
"type": "CN_WORD",
"position": 0
},
{
"token": "飞利浦 吸尘器",
"start_offset": 0,
"end_offset": 6,
"type": "shingle",
"position": 0,
"positionLength": 2
},
{
"token": "飞利浦 吸尘器 家用",
"start_offset": 0,
"end_offset": 8,
"type": "shingle",
"position": 0,
"positionLength": 3
},
{
"token": "飞利浦 吸尘器 家用 自营",
"start_offset": 0,
"end_offset": 10,
"type": "shingle",
"position": 0,
"positionLength": 4
},
{
"token": "吸尘器",
"start_offset": 3,
"end_offset": 6,
"type": "CN_WORD",
"position": 1
},
{
"token": "吸尘器 家用",
"start_offset": 3,
"end_offset": 8,
"type": "shingle",
"position": 1,
"positionLength": 2
},
{
"token": "吸尘器 家用 自营",
"start_offset": 3,
"end_offset": 10,
"type": "shingle",
"position": 1,
"positionLength": 3
},
{
"token": "家用",
"start_offset": 6,
"end_offset": 8,
"type": "CN_WORD",
"position": 2
},
{
"token": "家用 自营",
"start_offset": 6,
"end_offset": 10,
"type": "shingle",
"position": 2,
"positionLength": 2
},
{
"token": "自营",
"start_offset": 8,
"end_offset": 10,
"type": "CN_WORD",
"position": 3
}
]
}
此种分词方式适用于用户搜索词的模糊匹配,类似于hive中的rlike功能。
相较于使用match、match_phrase、regex的查询,该方式可以使用trem查询,效率大幅提高、且准确性几乎相同。