自然语言处理,是以哈工大社会计算与信息检索研究中心研发的“语言技术平台(LTP)”为基础,为用户提供高效精准的中文(简体)自然语言处理服务。该自然语言基础处理服务包括:词法分析、依存句法分析、语义角色标注、语义依存(依存树)分析、语义依存(依存图)分析五类,其中词法分析又可以分为:中文分词、词性标注、命名实体识别。
模块简介:
以上能力是通过HTTPAPI的方式给开发者提供一个通用的接口,适用于一次性交互数据传输的AI服务场景。相较于SDK,API具有轻量、跨语言的特点,不过请注意该接口使用的HTTPAPI协议不支持跨域。
集成自然语言处理API时,需按照以下要求。
接口地址示例:
在调用该业务接口时
IP白名单规则
注:
*X-CheckSum*生成示例:
X-Param生成示例:
注:一般基础类库会默认进行urlencode处理,请注意不要重复处理。
返回值为json串,各字段如下:
)datajson对象对应具体的文本分析结果,例如分词结果descstring描述sidstring会话ID其中sid字段主要用于追查问题,如果出现问题,可以提供sid给讯飞技术人员帮助确认问题。
data各字段说明如下:
结果示例如下:
失败结果:
{"code":"10105","desc":"illegalaccess|invalidX-Appid","data":{},"sid":"ltp0000a744@ch78290eb1e128000100"}成功结果:
中文分词(cws)
{"code":"0","data":{"word":["他","叫","汤姆","去","拿","外衣","。"]},"desc":"success","sid":"ltp00000006@ch7ea90e9a28b8000100"}词性标注(pos)
{"code":"0","data":{"pos":["r","v","nh","v","v","n","wp"]},"desc":"success","sid":"ltp00000007@ch7ea90e9a28ef000100"}"r","v","nh","v","v","n","wp"分别对应cws返回结果中的"他","叫","汤姆","去","拿","外衣","。"
依存句法分析(dp)
{"code":"0","data":{"dp":[{"parent":1,"relate":"SBV"},{"parent":-1,"relate":"HED"},{"parent":1,"relate":"DBL"},{"parent":4,"relate":"ADV"},{"parent":1,"relate":"VOB"},{"parent":4,"relate":"VOB"},{"parent":1,"relate":"WP"}]},"desc":"success","sid":"ltp00000005@ch7ea90e9a2858000100"}以dp[0]为例,"他"的父节点是"叫",他们中间的关系为SBV,即主谓关系。
命名实体识别(ner)
{"code":"0","data":{"ner":["O","O","S-Nh","O","O","O","O"]},"desc":"success","sid":"ltp00000008@ch7ea90e9a2928000100"}"O","O","S-Nh","O","O","O","O"分别对应cws返回结果中的"他","叫","汤姆","去","拿","外衣","。"
语义角色标注(srl)
{"code":"0","data":{"srl":[{"beg":0,"end":0,"id":1,"type":"A0"},{"beg":2,"end":2,"id":1,"type":"A1"},{"beg":3,"end":5,"id":1,"type":"A2"},{"beg":5,"end":5,"id":4,"type":"A1"}]},"desc":"success","sid":"ltp0000a741@ch78290eb1df9e000100"}语义依存(依存树)分析(sdp)
{"code":"0","data":{"sdp":[{"parent":2,"relate":"Agt"},{"parent":0,"relate":"Root"},{"parent":2,"relate":"Datv"},{"parent":2,"relate":"eSucc"},{"parent":4,"relate":"ePurp"},{"parent":5,"relate":"Pat"},{"parent":2,"relate":"mPunc"}]},"desc":"success","sid":"ltp00000002@ch409d0e9a29ec000100"}以sdp[0]为例,"他"的父节点是"叫",他们中间的关系为Agt,即施事关系。
语义依存(依存图)分析(sdgp)
{"code":"0","data":{"sdgp":[{"id":0,"parent":1,"relate":"Agt"},{"id":1,"parent":-1,"relate":"Root"},{"id":2,"parent":1,"relate":"Datv"},{"id":2,"parent":3,"relate":"Agt"},{"id":2,"parent":4,"relate":"Agt"},{"id":3,"parent":1,"relate":"eSucc"},{"id":4,"parent":3,"relate":"ePurp"},{"id":5,"parent":4,"relate":"Pat"},{"id":6,"parent":4,"relate":"mPunc"}]},"desc":"success","sid":"ltp0000000a@dx4f2f0f1f5931000100"}以sdgp[0]为例,"他"的父节点是"叫",他们中间的关系为Agt,即施事关系。
python脚本示例(python3)
说明:将脚本中TEXT,API_KEY,APPID,{func},换成相应的待分析文本,讯飞开放平台提供的apiKey,讯飞开放平台应用的appid以及功能模块名称即可,运行脚本可打印相应结果。
备注:核心的语义角色为A0-5六种,A0通常表示动作的施事,A1通常表示动作的影响等,A2-5根据谓语动词不同会有不同的语义含义。
注:demo只是一个简单的调用示例,不适合直接放在复杂多变的生产环境使用
答:语义依存分析(依存图),分析句子各个语言单位之间的语义关联,并将语义关联以依存结构呈现。