WO2018000279A1 - 基于导流的意图识别方法和系统 - Google Patents

基于导流的意图识别方法和系统 Download PDF

Info

Publication number
WO2018000279A1
WO2018000279A1 PCT/CN2016/087770 CN2016087770W WO2018000279A1 WO 2018000279 A1 WO2018000279 A1 WO 2018000279A1 CN 2016087770 W CN2016087770 W CN 2016087770W WO 2018000279 A1 WO2018000279 A1 WO 2018000279A1
Authority
WO
WIPO (PCT)
Prior art keywords
intent
identified
user
matching
precise
Prior art date
Application number
PCT/CN2016/087770
Other languages
English (en)
French (fr)
Inventor
邱楠
杨新宇
王昊奋
Original Assignee
深圳狗尾草智能科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳狗尾草智能科技有限公司 filed Critical 深圳狗尾草智能科技有限公司
Priority to CN201680001738.7A priority Critical patent/CN106471502A/zh
Priority to PCT/CN2016/087770 priority patent/WO2018000279A1/zh
Publication of WO2018000279A1 publication Critical patent/WO2018000279A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33Querying
    • G06F16/3331Query processing
    • G06F16/334Query execution
    • G06F16/3344Query execution using natural language analysis
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33Querying
    • G06F16/332Query formulation
    • G06F16/3329Natural language query formulation or dialogue systems

Definitions

  • the present invention relates to the field of computer application technologies, and in particular, to a method and system for intent recognition based on a flow guiding.
  • the QA Automatic Question Answering
  • the automatic question answering system will search the corresponding answer from the knowledge base or the Internet, and then return the answer directly to the user in a concise form instead of returning it to the user like a search engine. It's a bunch of related web pages. This allows users to easily get the information they want through an automated question and answer system.
  • the automatic question-and-answer technology uses techniques such as knowledge representation, information retrieval, and natural language processing.
  • the automated question answering system enables users to enter questions in natural language rather than a combination of keywords. Returning to the user is a simple and accurate answer, not some related web pages.
  • the question and answer system can better meet the user's search needs, and can quickly find out the answers the user needs.
  • the user does not need to break down his or her own problem into keywords, and the user can directly pass the entire question to the question and answer system.
  • the question and answer system combined with natural language processing technology, can directly submit the answers that the user wants by understanding the problem.
  • the Q&A system is like a knowledgeable expert who can answer any question quickly and accurately.
  • the automatic question answering system is to receive the problem of the user's natural language input and give the corresponding answer, which can solve the problem that the search engine can't solve (semantic analysis, interactive questioning), which is faster, more direct and more accurate than the traditional search engine. .
  • Intent recognition is a research topic in the field of artificial intelligence and natural language processing, mainly used to identify the user's behavioral intentions. For example, in a question-and-answer session, the questioner has a certain intent for each sentence, and the responder answers according to the intention of the other party.
  • the existing automatic question answering system is based on accurate intent recognition and gives a clear answer, which is achieved in the following two ways.
  • the automatic question answering system In the first way, after the user asks a question, the automatic question answering system is in the knowledge base. Find the most similar question and return the corresponding answer.
  • the automatic question answering system In the second mode, after the user asks a question, the automatic question answering system first parses the question keyword, generates a query statement, then queries the database, and finally returns the answer generated by the query. It can be seen that in the existing automatic question answering system, the intent of the user problem must be accurately recognized to give an answer, and the user question with a relatively intentional intention cannot be given an answer.
  • the technical solution adopted by the present invention is: a flow-based intent recognition method, which is characterized in that:
  • the exact intent matching result is returned. Otherwise, the fuzzy entry matching is performed on the user problem to be identified, the fuzzy entry is determined, and the fuzzy entry is returned according to the fuzzy entry. answer.
  • the step of obtaining a user problem to be identified comprises:
  • the step of performing intent matching on the user problem to be identified, and searching for whether there is a precise intention comprises:
  • the fuzzy entry matching is performed on the user problem to be identified, and the fuzzy entry is determined, and
  • the steps of returning a corresponding answer according to the fuzzy entry include:
  • the preset answer is returned through the preset recommended flow map.
  • the invention also provides a flow-based intent recognition system, comprising:
  • An intent identification module configured to perform an intent matching on the user problem to be identified, to find out whether there is an accurate intention
  • the exact intent matching result is returned. Otherwise, the fuzzy entry matching is performed on the user problem to be identified, the fuzzy entry is determined, and the fuzzy entry is returned according to the fuzzy entry. answer.
  • the obtaining module comprises:
  • the collecting unit is configured to collect text information or multi-modal input information input by the user and extract user parameters;
  • a converting unit configured to convert the multimodal input information and the user parameter into text format information
  • a pre-processing unit configured to pre-process the text information or the text format information to obtain the user problem to be identified.
  • the intent identification module comprises:
  • a parsing unit configured to parse keywords of the user problem to be identified
  • a query statement generating unit configured to generate a query statement according to the keyword
  • a query unit configured to perform a query in the exact intent matching database by using the query statement
  • a result output unit configured to have an answer matching the keyword in the exact intent matching database, determine that the user problem to be identified has a precise intent match, and return a precise intent matching result.
  • the intent identification module further includes:
  • a fuzzy entry determining unit configured to perform fuzzy entry matching according to the keyword, and determine a fuzzy entry corresponding to the keyword
  • the determining unit is configured to determine whether there is a precise intention under the fuzzy entry, and if yes, the result output unit determines the precise intent by asking a question and returns a corresponding answer.
  • the result output unit returns a preset answer through the preset recommended flow map.
  • the method and system for intent-based identification based on the flow guiding method provided by the present invention firstly perform accurate intent matching on a user problem to find out whether there is a precise intention corresponding thereto; if not, perform blurring The entry is matched to determine the fuzzy entry corresponding to it; under the corresponding fuzzy entry, for the user problem with precise intention, the precise intention is determined by means of counter-question, etc., for the problem that the precise intention cannot be located, the preset recommended flow is guided.
  • Figure recommend the default answer for the user. Thereby, the false triggering and rejection identification of the intent recognition can be effectively reduced, and the user experience is improved.
  • FIG. 1 is a flowchart of a method for identifying a flow based intent according to an embodiment of the present invention.
  • FIG. 1 is a flowchart of a method for identifying a flow based intent according to an embodiment of the present invention. As shown in FIG. 1, the flow-based intent recognition method includes the following steps:
  • Step S1 Acquire a user problem to be identified.
  • the multi-modal input information includes, but is not limited to, video, face, expression, scene, voiceprint, fingerprint, iris pupil, light perception and the like.
  • Step S2 Perform an intent matching on the user problem to be identified to find out whether there is a precise intent.
  • the fuzzy entry matching is performed on the user problem to be identified, and the fuzzy entry is determined. And returning the corresponding answer according to the fuzzy entry.
  • the keyword of the user problem to be identified is first parsed; then, the query statement is generated according to the keyword, and the query is used to match the exact intent.
  • a query is made in the database, if there is an answer matching the keyword in the exact intent matching database, it is determined that the user problem to be identified has a precise intent match, and returns a precise intent matching result.
  • fuzzy entry matching is performed according to the keyword, and a fuzzy entry corresponding to the keyword is determined; and the fuzzy entry is determined Whether there is a precise intention below, if yes, the exact intent is determined by the counter-question, and the corresponding answer is returned. If there is no precise intention under the fuzzy entry, the preset answer is returned through the preset recommended flow map.
  • An embodiment of the present invention further provides a flow-based intent recognition system, including:
  • An intent identification module configured to perform an intent matching on the user problem to be identified, to find out whether there is an accurate intention
  • the exact intent matching result is returned. Otherwise, the fuzzy entry matching is performed on the user problem to be identified, the fuzzy entry is determined, and the fuzzy entry is returned according to the fuzzy entry. answer.
  • the obtaining module includes:
  • the collecting unit is configured to collect text information or multi-modal input information input by the user and extract user parameters;
  • a converting unit configured to convert the multimodal input information and the user parameter into text format information
  • a pre-processing unit configured to pre-process the text information or the text format information to obtain the user problem to be identified.
  • the intent identification module includes:
  • a parsing unit configured to parse keywords of the user problem to be identified
  • a query statement generating unit configured to generate a query statement according to the keyword
  • a query unit configured to perform a query in the exact intent matching database by using the query statement
  • a result output unit configured to have an answer matching the keyword in the exact intent matching database, determine that the user problem to be identified has a precise intent match, and return a precise intent matching result
  • a fuzzy entry determining unit configured to perform fuzzy entry matching according to the keyword, and determine a fuzzy entry corresponding to the keyword
  • a determining unit configured to determine whether there is a precise intent under the fuzzy entry, and if so, the result output unit determines a precise intent by asking a question and returns a corresponding answer; if there is no precise intention under the fuzzy entry, The result output unit returns the preset answer through the preset recommended flow map.
  • the flow-based intent recognition method and system provided by the present invention first performs precise intent matching on a user problem to find out whether there is a precise intention corresponding thereto; if not, performs fuzzy entry matching to determine a corresponding Fuzzy entry; under the corresponding fuzzy entry, for the user problem with precise intention, the precise intention is determined by means of counter-question, etc.
  • the preset recommended flow map is recommended for the user. answer.

Abstract

一种基于导流的意图识别方法,包括:获取待识别的用户问题(S1);对待识别的用户问题进行精确意图匹配,如果有,则返回精确意图匹配结果,如果没有,则进行模糊入口匹配(S2)。本方法可以减少意图识别的误触发和拒识别,提升用户体验。

Description

基于导流的意图识别方法和系统 技术领域
本发明涉及计算机应用技术领域,尤其涉及一种基于导流的意图识别方法和系统。
背景技术
自动问答(QA,Automatic Question Answering)技术是伴随着自然语言的语义处理技术而发展起来的。人们可以用普通的问句对自动问答系统提问,自动问答系统将从知识库或者互联网中搜索相应的答案,然后把答案以简洁的形式直接返回给用户,而不是像搜索引擎那样返回给用户的是一堆相关的网页。这样用户就可以通过自动问答系统方便地获得自己想要的信息。自动问答技术综合运用了知识表示、信息检索、自然语言处理等技术。自动问答系统能够使用户以自然语言输入问题,而不是关键词的组合。而返回给用户的是简洁、准确的答案,而不是一些相关的网页。所以,问答系统能更好的满足用户的检索需求,能更快地找出用户所需的答案。对于问答系统,用户不需要把自己的问题分解成关键字,用户可以把整个问题直接交给问答系统。问答系统结合自然语言处理技术,通过对问题理解,能够直接提交给用户想要的答案。问答系统就像一个知识渊博的专家,可以快速准确地回答任何问题。比如,用户提交一个问题“上海的简称是什么?”,问答系统将会直接给出答案“上海的简称是沪”。自动问答系统是,通过接收用户自然语言形式输入的问题,给出相应答案,可以解决搜索引擎解决不了的问题(语义分析,交互反问),相较于传统的搜索引擎,更快速、直接、准确。
意图识别是人工智能和自然语言处理领域中的一个倍受关注的研究方向,主要用于识别用户的行为意图。例如,在问答对话中,提问者每句话都带有一定的意图,应答方则根据对方的意图进行回答。
现有的自动问答系统是基于精确意图识别,并给出明确答案,其通过以下两种方式来实现。在第一种方式中,用户提出问题后,自动问答系统在知识库 找到最相似的问题,然后返回对应的答案。在第二种方式中,用户提出问题后,自动问答系统首先解析问题关键词,再生成查询语句,之后查询数据库,最后返回查询生成的答案。由此可见,在现有的自动问答系统中,必须精确识别用户问题的意图才能给出答案,而对于意图比较模糊的用户问题,则无法给出答案。
发明内容
本发明的目的在于提供一种基于导流的意图识别方法和系统以减少意图识别的误触发和拒识别,提升用户体验。
本发明为了解决上述技术问题,采用的技术方案是:一种基于导流的意图识别方法,其特征在于,包括:
获取待识别的用户问题;
对所述待识别的用户问题进行意图匹配,查找是否有精确意图,
其中,如果所述待识别的用户问题有精确意图匹配,则返回精确意图匹配结果,反之,则对所述待识别的用户问题进行模糊入口匹配,确定模糊入口,并根据所述模糊入口返回对应答案。
优选地,获取待识别的用户问题的所述步骤包括:
采集用户输入的文字信息或多模态输入信息并提取用户参数;
如果输入的是多模态输入信息,将所述多模态输入信息和所述用户参数转换为文本格式信息;
对所述文字信息或所述文本格式信息进行预处理得到所述待识别的用户问题。
优选地,对所述待识别的用户问题进行意图匹配,查找是否有精确意图的所述步骤包括:
解析所述待识别的用户问题的关键词;
根据所述关键词生成查询语句;
利用所述查询语句在精确意图匹配数据库中进行查询;
如果在所述精确意图匹配数据库中存在与所述关键词匹配的答案,则判定所述待识别的用户问题有精确意图匹配,并返回精确意图匹配结果。
优选地,对所述待识别的用户问题进行模糊入口匹配,确定模糊入口,并 根据所述模糊入口返回对应答案的所述步骤包括:
根据所述关键词进行模糊入口匹配,确定与所述关键词对应的模糊入口;
判断所述模糊入口下面是否存在精确意图,如果有,则通过反问确定精确意图,并返回对应的答案。
优选地,如果所述模糊入口下面不存在精确意图,则通过预设的推荐流量导图,返回预设的答案。
本发明还提供一种基于导流的意图识别系统,其特征在于,包括:
获取模块,用于获取待识别的用户问题;
意图识别模块,用于对所述待识别的用户问题进行意图匹配,查找是否有精确意图,
其中,如果所述待识别的用户问题有精确意图匹配,则返回精确意图匹配结果,反之,则对所述待识别的用户问题进行模糊入口匹配,确定模糊入口,并根据所述模糊入口返回对应答案。
优选地,所述获取模块包括:
采集单元,用于采集用户输入的文字信息或多模态输入信息并提取用户参数;
转换单元,用于将所述多模态输入信息和所述用户参数转换为文本格式信息;
预处理单元,用于对所述文字信息或所述文本格式信息进行预处理得到所述待识别的用户问题。
优选地,所述意图识别模块包括:
解析单元,用于解析所述待识别的用户问题的关键词;
查询语句生成单元,用于根据所述关键词生成查询语句;
查询单元,用于利用所述查询语句在精确意图匹配数据库中进行查询;
结果输出单元,用于在所述精确意图匹配数据库中存在与所述关键词匹配的答案使,判定所述待识别的用户问题有精确意图匹配,并返回精确意图匹配结果。
优选地,所述意图识别模块还包括:
模糊入口确定单元,用于根据所述关键词进行模糊入口匹配,确定与所述关键词对应的模糊入口;
判断单元,用于判断所述模糊入口下面是否存在精确意图,如果有,则所述结果输出单元通过反问确定精确意图,并返回对应的答案。
优选地,如果所述模糊入口下面不存在精确意图,则所述结果输出单元通过预设的推荐流量导图,返回预设的答案。
实施本发明实施例,具有如下有益效果:本发明提供的基于导流的意图识别方法和系统,首先对用户问题进行精确意图匹配,查找是否存在与之对应的精确意图;如果没有,则进行模糊入口匹配,确定与之对应的模糊入口;在对应的模糊入口下面,对于存在精确意图的用户问题,通过反问等手段确定精确意图,对于无法定位精确意图的问题,则通过预设的推荐流量导图,为用户推荐预设的答案。由此,可以有效减少意图识别的误触发和拒识别,提升用户体验。
附图说明
为了更清楚地说明本发明实施例或现有技术中的技术方案,下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍,显而易见地,下面描述中的附图仅仅是本发明的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动的前提下,还可以根据这些附图获得其他的附图。
图1为本发明一实施例提供的基于导流的意图识别方法的流程图。
具体实施方式
下面将结合本发明实施例中的附图,对本发明实施例中的技术方案进行清楚、完整地描述,显然,所描述的实施例仅仅是本发明一部分实施例,而不是全部的实施例。基于本发明中的实施例,本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例,都属于本发明保护的范围。
图1为本发明一实施例提供的基于导流的意图识别方法的流程图。如图1所示,基于导流的意图识别方法包括以下步骤:
步骤S1:获取待识别的用户问题。
具体地,在本发明一实施例中,首先需要采集用户输入的文字信息或多模态输入信息并提取用户参数;其次,如果输入的是多模态输入信息,将所述多模态输入信息和所述用户参数转换为文本格式信息;最后,对所述文字信息或 所述文本格式信息进行预处理得到所述待识别的用户问题。其中,多模态输入信息包括但不限于,视频、人脸、表情、场景、声纹、指纹、虹膜瞳孔、光感等信息。
步骤S2:对所述待识别的用户问题进行意图匹配,查找是否有精确意图。
具体地,在本发明一实施例中,如果所述待识别的用户问题有精确意图匹配,则返回精确意图匹配结果,反之,则对所述待识别的用户问题进行模糊入口匹配,确定模糊入口,并根据所述模糊入口返回对应答案。
进一步地,在对待识别的用户问题进行意图匹配时,首先要解析所述待识别的用户问题的关键词;之后,根据所述关键词来生成查询语句,并利用所述查询语句在精确意图匹配数据库中进行查询,如果在所述精确意图匹配数据库中存在与所述关键词匹配的答案,则判定所述待识别的用户问题有精确意图匹配,并返回精确意图匹配结果。
进一步地,如果在所述精确意图匹配数据库中不存在与所述关键词匹配的答案,则根据所述关键词进行模糊入口匹配,确定与所述关键词对应的模糊入口;判断所述模糊入口下面是否存在精确意图,如果有,则通过反问确定精确意图,并返回对应的答案,如果所述模糊入口下面不存在精确意图,则通过预设的推荐流量导图,返回预设的答案。
例如,如果用户输入的是“我想听周杰伦的七里香”,则会命中精确意图“开始播放音乐”,系统会直接播放七里香。如果用户输入的是“我想去旅游”,则首先无法匹配到目前的所有精确意图,接着在模糊意图中查找,匹配到旅游相关入口。接下来,通过反问用户,“你需要找酒店还是定机票”,如果用户回复“定机票”,则返回预定旅游机票的精确意图,如果用户不回复或回复的内容不足以确定精确意图,则返回一个模糊意图的通用回答进行导流。本发明一实施例还提供一种基于导流的意图识别系统,包括:
获取模块,用于获取待识别的用户问题;
意图识别模块,用于对所述待识别的用户问题进行意图匹配,查找是否有精确意图,
其中,如果所述待识别的用户问题有精确意图匹配,则返回精确意图匹配结果,反之,则对所述待识别的用户问题进行模糊入口匹配,确定模糊入口,并根据所述模糊入口返回对应答案。
进一步地,所述获取模块包括:
采集单元,用于采集用户输入的文字信息或多模态输入信息并提取用户参数;
转换单元,用于将所述多模态输入信息和所述用户参数转换为文本格式信息;
预处理单元,用于对所述文字信息或所述文本格式信息进行预处理得到所述待识别的用户问题。
进一步地,所述意图识别模块包括:
解析单元,用于解析所述待识别的用户问题的关键词;
查询语句生成单元,用于根据所述关键词生成查询语句;
查询单元,用于利用所述查询语句在精确意图匹配数据库中进行查询;
结果输出单元,用于在所述精确意图匹配数据库中存在与所述关键词匹配的答案使,判定所述待识别的用户问题有精确意图匹配,并返回精确意图匹配结果;
模糊入口确定单元,用于根据所述关键词进行模糊入口匹配,确定与所述关键词对应的模糊入口;
判断单元,用于判断所述模糊入口下面是否存在精确意图,如果有,则所述结果输出单元通过反问确定精确意图,并返回对应的答案;如果所述模糊入口下面不存在精确意图,则所述结果输出单元通过预设的推荐流量导图,返回预设的答案。
有利地,本发明提供的基于导流的意图识别方法和系统,首先对用户问题进行精确意图匹配,查找是否存在与之对应的精确意图;如果没有,则进行模糊入口匹配,确定与之对应的模糊入口;在对应的模糊入口下面,对于存在精确意图的用户问题,通过反问等手段确定精确意图,对于无法定位精确意图的问题,则通过预设的推荐流量导图,为用户推荐预设的答案。由此,可以有效减少意图识别的误触发和拒识别,提升用户体验。
以上所揭露的仅为本发明一种较佳实施例而已,当然不能以此来限定本发明之权利范围,本领域普通技术人员可以理解实现上述实施例的全部或部分流程,并依本发明权利要求所作的等同变化,仍属于发明所涵盖的范围。

Claims (10)

  1. 一种基于导流的意图识别方法,其特征在于,包括:
    获取待识别的用户问题;
    对所述待识别的用户问题进行意图匹配,查找是否有精确意图,
    其中,如果所述待识别的用户问题有精确意图匹配,则返回精确意图匹配结果,反之,则对所述待识别的用户问题进行模糊入口匹配,确定模糊入口,并根据所述模糊入口返回对应答案。
  2. 根据权利要求1所述的基于导流的意图识别方法,其特征在于,获取待识别的用户问题的所述步骤包括:
    采集用户输入文字信息或多模态输入信息并提取用户参数;
    如果输入的是多模态输入信息,将所述多模态输入信息和所述用户参数转换为文本格式信息;
    对所述文字信息或所述文本格式信息进行预处理得到所述待识别的用户问题。
  3. 根据权利要求1所述的基于导流的意图识别方法,其特征在于,对所述待识别的用户问题进行意图匹配,查找是否有精确意图的所述步骤包括:
    解析所述待识别的用户问题的关键词;
    根据所述关键词生成查询语句;
    利用所述查询语句在精确意图匹配数据库中进行查询;
    如果在所述精确意图匹配数据库中存在与所述关键词匹配的答案,则判定所述待识别的用户问题有精确意图匹配,并返回精确意图匹配结果。
  4. 根据权利要求3所述的基于导流的意图识别方法,其特征在于,对所述待识别的用户问题进行模糊入口匹配,确定模糊入口,并根据所述模糊入口返回对应答案的所述步骤包括:
    根据所述关键词进行模糊入口匹配,确定与所述关键词对应的模糊入口;
    判断所述模糊入口下面是否存在精确意图,如果有,则通过反问确定精确意图,并返回对应的答案。
  5. 根据权利要求4所述的基于多媒体的智能机器人交互方法,其特征在于,如果所述模糊入口下面不存在精确意图,则通过预设的推荐流量导图,返回预 设的答案。
  6. 一种基于导流的意图识别系统,其特征在于,包括:
    获取模块,用于获取待识别的用户问题;
    意图识别模块,用于对所述待识别的用户问题进行意图匹配,查找是否有精确意图,
    其中,如果所述待识别的用户问题有精确意图匹配,则返回精确意图匹配结果,反之,则对所述待识别的用户问题进行模糊入口匹配,确定模糊入口,并根据所述模糊入口返回对应答案。
  7. 根据权利要求6所述的基于导流的意图识别系统,其特征在于,所述获取模块包括:
    采集单元,用于采集用户输入的文字信息或多模态输入信息并提取用户参数;
    转换单元,用于将所述多模态输入信息和所述用户参数转换为文本格式信息;
    预处理单元,用于对所述文字信息或所述文本格式信息进行预处理得到所述待识别的用户问题。
  8. 根据权利要求6所述的基于导流的意图识别系统,其特征在于,所述意图识别模块包括:
    解析单元,用于解析所述待识别的用户问题的关键词;
    查询语句生成单元,用于根据所述关键词生成查询语句;
    查询单元,用于利用所述查询语句在精确意图匹配数据库中进行查询;
    结果输出单元,用于在所述精确意图匹配数据库中存在与所述关键词匹配的答案使,判定所述待识别的用户问题有精确意图匹配,并返回精确意图匹配结果。
  9. 根据权利要求8所述的基于导流的意图识别系统,其特征在于,所述意图识别模块还包括:
    模糊入口确定单元,用于根据所述关键词进行模糊入口匹配,确定与所述关键词对应的模糊入口;
    判断单元,用于判断所述模糊入口下面是否存在精确意图,如果有,则所述结果输出单元通过反问确定精确意图,并返回对应的答案。
  10. 根据权利要求9所述的基于导流的意图识别系统,其特征在于,如果所述模糊入口下面不存在精确意图,则所述结果输出单元通过预设的推荐流量导图,返回预设的答案。
PCT/CN2016/087770 2016-06-29 2016-06-29 基于导流的意图识别方法和系统 WO2018000279A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN201680001738.7A CN106471502A (zh) 2016-06-29 2016-06-29 基于导流的意图识别方法和系统
PCT/CN2016/087770 WO2018000279A1 (zh) 2016-06-29 2016-06-29 基于导流的意图识别方法和系统

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2016/087770 WO2018000279A1 (zh) 2016-06-29 2016-06-29 基于导流的意图识别方法和系统

Publications (1)

Publication Number Publication Date
WO2018000279A1 true WO2018000279A1 (zh) 2018-01-04

Family

ID=58230945

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2016/087770 WO2018000279A1 (zh) 2016-06-29 2016-06-29 基于导流的意图识别方法和系统

Country Status (2)

Country Link
CN (1) CN106471502A (zh)
WO (1) WO2018000279A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111400340A (zh) * 2020-03-12 2020-07-10 杭州城市大数据运营有限公司 一种自然语言处理方法、装置、计算机设备和存储介质
CN111400453A (zh) * 2020-03-16 2020-07-10 上海百事通信息技术股份有限公司 一种基于法律咨询的智能交互系统及方法

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106933807A (zh) * 2017-03-20 2017-07-07 北京光年无限科技有限公司 备忘录事件提醒方法及系统
CN109508376A (zh) * 2018-11-23 2019-03-22 四川长虹电器股份有限公司 可在线纠错更新的意图识别方法及装置
CN109657444B (zh) * 2018-12-29 2021-06-25 上海智臻智能网络科技股份有限公司 安全验证方法及装置、存储介质、终端
CN110119509A (zh) * 2019-05-10 2019-08-13 福建工程学院 一种基于语义分析的备忘录事件智能提醒方法
CN112256846A (zh) * 2020-09-28 2021-01-22 南方电网深圳数字电网研究院有限公司 一种人机对话交互方法及系统

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101286161A (zh) * 2008-05-28 2008-10-15 华中科技大学 一种基于概念的智能中文问答系统
CN104199825A (zh) * 2014-07-23 2014-12-10 清华大学 一种信息查询方法和系统
CN104598445A (zh) * 2013-11-01 2015-05-06 腾讯科技(深圳)有限公司 自动问答系统和方法
CN105677822A (zh) * 2016-01-05 2016-06-15 首都师范大学 一种基于对话机器人的招生自动问答方法及系统

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105335400B (zh) * 2014-07-22 2018-11-23 阿里巴巴集团控股有限公司 针对用户的提问意图获取答案信息的方法及装置
CN105487663B (zh) * 2015-11-30 2018-09-11 北京光年无限科技有限公司 一种面向智能机器人的意图识别方法和系统

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101286161A (zh) * 2008-05-28 2008-10-15 华中科技大学 一种基于概念的智能中文问答系统
CN104598445A (zh) * 2013-11-01 2015-05-06 腾讯科技(深圳)有限公司 自动问答系统和方法
CN104199825A (zh) * 2014-07-23 2014-12-10 清华大学 一种信息查询方法和系统
CN105677822A (zh) * 2016-01-05 2016-06-15 首都师范大学 一种基于对话机器人的招生自动问答方法及系统

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111400340A (zh) * 2020-03-12 2020-07-10 杭州城市大数据运营有限公司 一种自然语言处理方法、装置、计算机设备和存储介质
CN111400340B (zh) * 2020-03-12 2024-01-09 杭州城市大数据运营有限公司 一种自然语言处理方法、装置、计算机设备和存储介质
CN111400453A (zh) * 2020-03-16 2020-07-10 上海百事通信息技术股份有限公司 一种基于法律咨询的智能交互系统及方法
CN111400453B (zh) * 2020-03-16 2024-03-08 上海百事通公共法律服务有限公司 一种基于法律咨询的智能交互系统及方法

Also Published As

Publication number Publication date
CN106471502A (zh) 2017-03-01

Similar Documents

Publication Publication Date Title
WO2018000279A1 (zh) 基于导流的意图识别方法和系统
RU2747425C2 (ru) Система ответа на вопросы из разных областей знаний в режиме реального времени
CN107943998B (zh) 一种基于知识图谱的人机对话控制系统及方法
CN115238101B (zh) 一种面向多类型知识库的多引擎智能问答系统
CN107665708B (zh) 智能语音交互方法及系统
CN102262634B (zh) 一种自动问答方法及系统
JP2017511922A (ja) スマート質問回答の実現方法、システム、および記憶媒体
CN110335595A (zh) 基于语音识别的插问对话方法、装置及存储介质
KR102033388B1 (ko) 질의 응답 장치 및 방법
CN109410927A (zh) 离线命令词与云端解析结合的语音识别方法、装置和系统
US11756532B2 (en) Intelligence-driven virtual assistant for automated idea documentation
JP5780617B2 (ja) 関連知識情報を獲得・検索する方法及び装置
CN111400453B (zh) 一种基于法律咨询的智能交互系统及方法
CN111858876A (zh) 一种知识库的生成方法、文本查找方法和装置
CN114676259B (zh) 一种基于因果感知交互网络的对话情绪识别方法
US11188533B1 (en) Generating query answers from a user's history
CN108763355B (zh) 一种基于用户的智能机器人交互数据处理系统及方法
CN114742032A (zh) 交互式数据分析方法、装置、设备、介质及程序产品
CN112035655B (zh) 银行人力资源系统及工作方法
Bruhn Target First: On “Bidirectionality and Metaphor”
Clifton et al. Bangor at TREC 2004: Question Answering Track.
Stoyanchev et al. Combining structured and unstructured knowledge in an interactive search dialogue system
Aavik 11 Studying Privileged Men’s Career Narratives from an Intersectional Perspective: The Methodological Challenge of the Invisibility of Privilege
CN115810422B (zh) 基于ai技术的互联网智能自动诊疗应答系统
García-Robledo Gabriela et al. Ontology-Based Question Answering System for an Academic Domain

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 16906680

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 16906680

Country of ref document: EP

Kind code of ref document: A1