不引入ES，使用MySQL實(shí)現(xiàn)分詞模糊匹配
來源：juejin.cn/post/7340574992256466953
推薦：https://t.zsxq.com/18Mq6lus4
業(yè)務(wù)場景概述
目標(biāo)是實(shí)現(xiàn)一個(gè)公司的申請審批流程，整個(gè)業(yè)務(wù)流程涉及到兩種角色，分別為商務(wù)角色與管理員角色。整個(gè)流程如下圖所示：
核心流程總結(jié)為一句話：商務(wù)角色申請?zhí)砑庸竞笥晒芾韱T進(jìn)行審批。
商務(wù)在添加公司時(shí)，可能為了方便，直接填寫公司簡稱，而公司全稱可能之前已經(jīng)被添加過了，為了防止添加重復(fù)的公司，所以管理員在針對(duì)公司信息審批之前，需要查看以往添加的公司信息里有無同一個(gè)公司。
實(shí)現(xiàn)思路
以上是一個(gè)業(yè)務(wù)場景的大概介紹。從技術(shù)層面需要考慮實(shí)現(xiàn)的功能點(diǎn)：
-
分詞 -
與庫里已有數(shù)據(jù)進(jìn)行匹配 -
按照匹配度對(duì)結(jié)果進(jìn)行排序
分詞功能有現(xiàn)成的分詞器，所以整個(gè)需求的核心重點(diǎn)在于如何與數(shù)據(jù)庫中的數(shù)據(jù)匹配并按照匹配度排序。
模糊匹配技術(shù)選型
-
方案一：引入ES -
方案二：利用MySQL實(shí)現(xiàn)
本系統(tǒng)規(guī)模較小，單純?yōu)榱藢?shí)現(xiàn)這個(gè)功能引入ES成本較大，還要涉及到數(shù)據(jù)同步等問題，系統(tǒng)復(fù)雜性會(huì)提高，所以盡量使用MySQL已有的功能進(jìn)行實(shí)現(xiàn)。
MySQL提供了以下三種模糊搜索的方式：
-
like匹配：要求模式串與整個(gè)目標(biāo)字段完全匹配； -
RegExp正則匹配：要求目標(biāo)字段包含模式串即可； -
Fulltext全文索引：在字段類型為CHAR,VARCHAR，TEXT的列上創(chuàng)建全文索引，執(zhí)行SQL進(jìn)行查詢。
針對(duì)于上述業(yè)務(wù)場景，對(duì)相關(guān)技術(shù)進(jìn)行優(yōu)劣分析：
-
like匹配，無法滿足需求，所以
pass； -
全文索引：可定制性差，不支持任意匹配查詢，
pass； -
正則匹配：可實(shí)現(xiàn)任意模式匹配，缺點(diǎn)在于執(zhí)行效率不如全文索引。
針對(duì)于這個(gè)場景，記錄數(shù)目相對(duì)來說沒有那么多，所以對(duì)于效率稍低的結(jié)果可以接受，因此技術(shù)選型方面采用RegExp正則匹配來實(shí)現(xiàn)模糊匹配的需求。
實(shí)現(xiàn)效果展示
核心代碼
整個(gè)邏輯基于 提取公司名稱關(guān)鍵信息 -->分詞 --> 匹配 三個(gè)核心步驟。
提取公司關(guān)鍵信息
對(duì)輸入的公司名稱去除無用信息，保留關(guān)鍵信息。這里的無用信息指的是地名，圓括號(hào)，以及集團(tuán)，股份，有限等。
-
匹配前處理公司名稱
/**
* 匹配前去除公司名稱的無意義信息
* @param targetCompanyName
* @return
*/
private String formatCompanyName(String targetCompanyName){
String regex = "(?<province>[^省]+自治區(qū)|.*?省|.*?行政區(qū)|.*?市)" +
"?(?<city>[^市]+自治州|.*?地區(qū)|.*?行政單位|.+盟|市轄區(qū)|.*?市|.*?縣)" +
"?(?<county>[^(區(qū)|市|縣|旗|島)]+區(qū)|.*?市|.*?縣|.*?旗|.*?島)" +
"?(?<village>.*)";
Matcher matcher = Pattern.compile(regex).matcher(targetCompanyName);
while(matcher.find()){
String province = matcher.group("province");
log.info("province:{}",province);
if (StringUtils.isNotBlank(province) && targetCompanyName.contains(province)){
targetCompanyName = targetCompanyName.replace(province,"");
}
log.info("處理完省份的公司名稱:{}",targetCompanyName);
String city = matcher.group("city");
log.info("city:{}",city);
if (StringUtils.isNotBlank(city) && targetCompanyName.contains(city)){
targetCompanyName = targetCompanyName.replace(city,"");
}
log.info("處理完城市的公司名稱:{}",targetCompanyName);
String county = matcher.group("county");
log.info("county:{}",county);
if (StringUtils.isNotBlank(county) && targetCompanyName.contains(county)){
targetCompanyName = targetCompanyName.replace(county,"");
}
log.info("處理完區(qū)縣級(jí)的公司名稱:{}",targetCompanyName);
}
String[][] address = AddressUtil.ADDRESS;
for (String [] city: address) {
for (String b : city ) {
if (targetCompanyName.contains(b)){
targetCompanyName = targetCompanyName.replace(b, "");
}
}
}
log.info("處理后的公司名稱:{}",targetCompanyName);
return targetCompanyName;
}
-
地名工具類
public class AddressUtil {
public static final String[][] ADDRESS = {
{"北京"},
{"天津"},
{"安徽","安慶","蚌埠","亳州","巢湖","池州","滁州","阜陽","合肥","淮北","淮南","黃山","六安","馬鞍山","宿州","銅陵","蕪湖","宣城"},
{"澳門"},
{"香港"},
{"福建","福州","龍巖","南平","寧德","莆田","泉州","廈門","漳州"},
{"甘肅","白銀","定西","甘南藏族自治州","嘉峪關(guān)","金昌","酒泉","蘭州","臨夏回族自治州","隴南","平?jīng)?,"慶陽","天水","武威","張掖"},
{"廣東","潮州","東莞","佛山","廣州","河源","惠州","江門","揭陽","茂名","梅州","清遠(yuǎn)","汕頭","汕尾","韶關(guān)","深圳","陽江","云浮","湛江","肇慶","中山","珠海"},
{"廣西","百色","北海","崇左","防城港","貴港","桂林","河池","賀州","來賓","柳州","南寧","欽州","梧州","玉林"},
{"貴州","安順","畢節(jié)地區(qū)","貴陽","六盤水","黔東南苗族侗族自治州","黔南布依族苗族自治州","黔西南布依族苗族自治州","銅仁地區(qū)","遵義"},
{"海南","海口","三亞","直轄縣級(jí)行政區(qū)劃"},
{"河北","保定","滄州","承德","邯鄲","衡水","廊坊","秦皇島","石家莊","唐山","邢臺(tái)","張家口"},
{"河南","安陽","鶴壁","焦作","開封","洛陽","漯河","南陽","平頂山","濮陽","三門峽","商丘","新鄉(xiāng)","信陽","許昌","鄭州","周口","駐馬店"},
{"黑龍江","大慶","大興安嶺地區(qū)","哈爾濱","鶴崗","黑河","雞西","佳木斯","牡丹江","七臺(tái)河","齊齊哈爾","雙鴨山","綏化","伊春"},
{"湖北","鄂州","恩施土家族苗族自治州","黃岡","黃石","荊門","荊州","十堰","隨州","武漢","咸寧","襄樊","孝感","宜昌"},
{"湖南","長沙","常德","郴州","衡陽","懷化","婁底","邵陽","湘潭","湘西土家族苗族自治州","益陽","永州","岳陽","張家界","株洲"},
{"吉林","白城","白山","長春","吉林","遼源","四平","松原","通化","延邊朝鮮族自治州"},
{"江蘇","常州","淮安","連云港","南京","南通","蘇州","宿遷","泰州","無錫","徐州","鹽城","揚(yáng)州","鎮(zhèn)江"},
{"江西","撫州","贛州","吉安","景德鎮(zhèn)","九江","南昌","萍鄉(xiāng)","上饒","新余","宜春","鷹潭"},
{"遼寧","鞍山","本溪","朝陽","大連","丹東","撫順","阜新","葫蘆島","錦州","遼陽","盤錦","沈陽","鐵嶺","營口"},
{"內(nèi)蒙古","阿拉善盟","巴彥淖爾","包頭","赤峰","鄂爾多斯","呼和浩特","呼倫貝爾","通遼","烏海","烏蘭察布","錫林郭勒盟","興安盟"},
{"寧夏回族","固原","石嘴山","吳忠","銀川","中衛(wèi)"},
{"青海","果洛藏族自治州","海北藏族自治州","海東地區(qū)","海南藏族自治州","海西蒙古族藏族自治州","黃南藏族自治州","西寧","玉樹藏族自治州"},
{"山東","濱州","德州","東營","菏澤","濟(jì)南","濟(jì)寧","萊蕪","聊城","臨沂","青島","日照","泰安","威海","濰坊","煙臺(tái)","棗莊","淄博"},
{"山西","長治","大同","晉城","晉中","臨汾","呂梁","朔州","太原","忻州","陽泉","運(yùn)城"},
{"陜西","安康","寶雞","漢中","商洛","銅川","渭南","西安","咸陽","延安","榆林"},
{"上海"},
{"四川","阿壩藏族羌族自治州","巴中","成都","達(dá)州","德陽","甘孜藏族自治州","廣安","廣元","樂山","涼山彝族自治州","瀘州","眉山","綿陽","內(nèi)江","南充","攀枝花","遂寧","雅安","宜賓","資陽","自貢"},
{"西藏","阿里地區(qū)","昌都地區(qū)","拉薩","林芝地區(qū)","那曲地區(qū)","日喀則地區(qū)","山南地區(qū)"},
{"新疆維吾爾","阿克蘇地區(qū)","阿勒泰地區(qū)","巴音郭楞蒙古自治州","博爾塔拉蒙古自治州","昌吉回族自治州","哈密地區(qū)","和田地區(qū)","喀什地區(qū)","克拉瑪依","克孜勒蘇柯爾克孜自治州","塔城地區(qū)","吐魯番地區(qū)","烏魯木齊","伊犁哈薩克自治州","直轄縣級(jí)行政區(qū)劃"},
{"云南","保山","楚雄彝族自治州","大理白族自治州","德宏傣族景頗族自治州","迪慶藏族自治州","紅河哈尼族彝族自治州","昆明","麗江","臨滄","怒江僳僳族自治州","普洱","曲靖","文山壯族苗族自治州","西雙版納傣族自治州","玉溪","昭通"},
{"浙江","杭州","湖州","嘉興","金華","麗水","寧波","衢州","紹興","臺(tái)州","溫州","舟山"},
{"重慶"},
{"臺(tái)灣","臺(tái)北","高雄","基隆","臺(tái)中","臺(tái)南","新竹","嘉義"},
};
}
分詞相關(guān)代碼
-
pom文件：引入IK分詞器相關(guān)依賴
<!-- ikAnalyzer 中文分詞器 -->
<dependency>
<groupId>com.janeluo</groupId>
<artifactId>ikanalyzer</artifactId>
<version>2012_u6</version>
<exclusions>
<exclusion>
<groupId>org.apache.lucene</groupId>
<artifactId>lucene-core</artifactId>
</exclusion>
<exclusion>
<groupId>org.apache.lucene</groupId>
<artifactId>lucene-queryparser</artifactId>
</exclusion>
<exclusion>
<groupId>org.apache.lucene</groupId>
<artifactId>lucene-analyzers-common</artifactId>
</exclusion>
</exclusions>
</dependency>
<!-- lucene-queryParser 查詢分析器模塊 -->
<dependency>
<groupId>org.apache.lucene</groupId>
<artifactId>lucene-queryparser</artifactId>
<version>7.3.0</version>
</dependency>
-
IKAnalyzerSupport類：用于配置分詞器
@Slf4j
public class IKAnalyzerSupport {
/**
* IK分詞
* @param target
* @return
*/
public static List<String> iKSegmenterToList(String target) throws Exception {
if (StringUtils.isEmpty(target)){
return new ArrayList();
}
List<String> result = new ArrayList<>();
StringReader sr = new StringReader(target);
// false:關(guān)閉智能分詞 (對(duì)分詞的精度影響較大)
IKSegmenter ik = new IKSegmenter(sr, true);
Lexeme lex;
while((lex=ik.next())!=null) {
String lexemeText = lex.getLexemeText();
result.add(lexemeText);
}
return result;
}
}
-
ServiceImpl類：進(jìn)行分詞處理
/**
* 對(duì)目標(biāo)公司名稱進(jìn)行分詞
* @param targetCompanyName
* @return
*/
private String splitWord(String targetCompanyName){
log.info("對(duì)處理后端公司名稱進(jìn)行分詞");
List<String> splitWord = new ArrayList<>();
String result = targetCompanyName;
try {
splitWord = iKSegmenterToList(targetCompanyName);
result = splitWord.stream().map(String::valueOf).distinct().collect(Collectors.joining("|")) ;
log.info("分詞結(jié)果:{}",result);
} catch (Exception e) {
log.error("分詞報(bào)錯(cuò):{}",e.getMessage());
}
return result;
}
匹配
-
ServiceImpl類：匹配核心代碼
public JsonResult matchCompanyName(CompanyDTO companyDTO, String accessToken, String localIp) {
// 對(duì)公司名稱進(jìn)行處理
String sourceCompanyName = companyDTO.getCompanyName();
String targetCompanyName = sourceCompanyName;
log.info("處理前公司名稱:{}",targetCompanyName);
// 處理圓括號(hào)
targetCompanyName = targetCompanyName.replaceAll("[（]|[）]|[(]|[)]","");
// 處理公司相關(guān)關(guān)鍵詞
targetCompanyName = targetCompanyName.replaceAll("[(集團(tuán)|股份|有限|責(zé)任|分公司)]", "");
if (!targetCompanyName.contains("銀行")){
// 去除行政區(qū)域
targetCompanyName = formatCompanyName(targetCompanyName);
}
// 分詞
String splitCompanyName = splitWord(targetCompanyName);
// 匹配
List<Company> matchedCompany = companyRepository.queryMatchCompanyName(splitCompanyName,targetCompanyName);
List<String> result = new ArrayList();
for (Company companyInfo : matchedCompany) {
result.add(companyInfo.getCompanyName());
if (companyDTO.getCompanyId().equals(companyInfo.getCompanyId())){
result.remove(companyInfo.getCompanyName());
}
}
return JsonResult.successResult(result);
}
-
Repository類：編寫SQL語句
/**
* 模糊匹配公司名稱
* @param companyNameRegex 分詞后的公司名稱
* @param companyName 分詞前的公司名稱
* @return
*/
@Query(value =
"SELECT * FROM company WHERE isDeleted = '0' and companyName REGEXP ?1
ORDER BY length(REPLACE(companyName,?2,''))/length(companyName) ",
nativeQuery = true)
List<Company> queryMatchCompanyName(String companyNameRegex,String companyName);
按照匹配度排序這個(gè)功能點(diǎn)，LENGTH(companyName)返回companyName的長度，LENGTH(REPLACE(companyName, ?2, ''))計(jì)算出companyName中關(guān)鍵詞出現(xiàn)的次數(shù)。通過這種方式，我們可以根據(jù)匹配程度進(jìn)行排序，匹配次數(shù)越多的公司名稱排序越靠前。
評(píng)論
圖片
表情
<b id="afajh"><abbr id="afajh"></abbr></b>