WO2018000205A1 - 一种基于多意图的多技能包问答方法、***和机器人 - Google Patents

一种基于多意图的多技能包问答方法、***和机器人 Download PDF

Info

Publication number
WO2018000205A1
WO2018000205A1 PCT/CN2016/087517 CN2016087517W WO2018000205A1 WO 2018000205 A1 WO2018000205 A1 WO 2018000205A1 CN 2016087517 W CN2016087517 W CN 2016087517W WO 2018000205 A1 WO2018000205 A1 WO 2018000205A1
Authority
WO
WIPO (PCT)
Prior art keywords
answer
question
skill
user
answers
Prior art date
Application number
PCT/CN2016/087517
Other languages
English (en)
French (fr)
Inventor
王昊奋
邱楠
杨新宇
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 CN201680001740.4A priority Critical patent/CN106462647A/zh
Priority to PCT/CN2016/087517 priority patent/WO2018000205A1/zh
Publication of WO2018000205A1 publication Critical patent/WO2018000205A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00Handling natural language data
    • G06F40/20Natural language analysis
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90Details of database functions independent of the retrieved data types
    • G06F16/903Querying
    • G06F16/9032Query formulation
    • G06F16/90332Natural language query formulation or dialogue systems

Definitions

  • the invention relates to the field of robot interaction technology, in particular to a multi-intent multi-skilled question and answer method, system and robot.
  • a multi-intent based multi-skill package question and answer method including:
  • the step of answering the question information by using at least two skill packages specifically includes:
  • the step of obtaining the final answer from the answer set according to the user's intention specifically includes:
  • the step of answering the question information through the skill package specifically includes:
  • the step of answering the question information through the skill package specifically includes:
  • the step of obtaining the final answer from the answer set according to the user intention specifically includes:
  • the skill package includes at least a music skill package, a story skill package, a star technology package, and a horoscope skill package.
  • the invention discloses a multi-intent based multi-skill package question answering system, comprising:
  • An intention identification module configured to identify a user's intention according to the question information
  • the parsing module is configured to solve the question information by using at least two skill packages to obtain an answer set including at least two answers;
  • An output module for obtaining a final answer from the set of answers based on the user's intent.
  • the parsing module is further configured to: input the question information into at least two link data corresponding to the at least two skill packages for answering;
  • the output module is further configured to: associate the acquired at least two user intents by a relationship between the at least two link data, and select at least two of the acquired answer sets according to the association between the at least two user intents The answers are combined to get the final answer.
  • the parsing module is specifically configured to: segment the input question information;
  • the parsing module is specifically configured to:
  • the output module is specifically configured to:
  • the skill package includes at least a music skill package, a story skill package, a star technology package, and a horoscope skill package.
  • the present invention discloses a robot comprising a multi-intent based multi-skill package question answering system as described above.
  • the multi-skill package-based question and answer method disclosed by the present invention includes: obtaining question information; identifying user intention according to question information; and answering question information through at least two skill packs Get a set of answers that includes at least two answers; get the final answer from the set of answers based on the user's intent.
  • the user's question information can be answered by at least two skill packages, and then all the answers are obtained and then selected or associated according to the user's intention, thereby obtaining the final answer, wherein the association combination may be according to the user's intention.
  • the invention firstly proposes to manage the function modules of the robot and the robot in the management mode of the skill package. Under the framework of the parallel management, the processing speed and efficiency of the robot can be further improved, and the robot can realize the function more quickly and conveniently.
  • the startup improves the efficiency of the robot interaction with the user and the user's goodwill.
  • FIG. 1 is a flowchart of a multi-intent multi-skill packet question and answer method according to Embodiment 1 of the present invention
  • FIG. 2 is a schematic diagram of a multi-intent multi-skilled question answering system according to a second embodiment of the present invention.
  • Computer devices include user devices and network devices.
  • the user equipment or client includes However, it is not limited to computers, smart phones, PDAs, etc.; network devices include, but are not limited to, a single network server, a server group composed of a plurality of network servers, or a cloud-based cloud composed of a large number of computers or network servers.
  • the computer device can operate alone to carry out the invention, and can also access the network and implement the invention through interoperation with other computer devices in the network.
  • the network in which the computer device is located includes, but is not limited to, the Internet, a wide area network, a metropolitan area network, a local area network, a VPN network, and the like.
  • first means “first,” “second,” and the like may be used herein to describe the various elements, but the elements should not be limited by these terms, and the terms are used only to distinguish one element from another.
  • the term “and/or” used herein includes any and all combinations of one or more of the associated listed items. When a unit is referred to as being “connected” or “coupled” to another unit, it can be directly connected or coupled to the other unit, or an intermediate unit can be present.
  • a method for answering and answering based on a multi-skill package including:
  • the method for answering questions based on the multi-skill package includes: obtaining question information; identifying user intent according to the question information; answering the question information through at least two skill packages, and obtaining an answer set including at least two answers; The user intends to get the final answer from the answer set.
  • the user's question information can be answered by at least two skill packages, and then all the answers are obtained and then selected according to the user's intention, thereby obtaining the final answer, so that the obtained answer is more accurate, and by setting multiple skill packages. Can make the question and answer method more widely applicable.
  • the skill package includes at least a music skill package, a story skill package, a star technology package, and a horoscope skill package.
  • the user's questions can be more comprehensively queried and analyzed through different skill packages, and different answers are obtained, and then the answers are combined to obtain the final answer.
  • the first skill package, the second skill package, and the third skill package are included, and the answer set obtained from the three skill packages is compared with the user intention to obtain the final answer.
  • the first skill package may be a music skill package
  • the second skill package may be a story skill package
  • the third skill package may be a horoscope skill package.
  • the step of answering the question information by using at least two skill packages specifically includes:
  • the step of obtaining the final answer from the answer set according to the user's intention specifically includes:
  • the user asks for the message: "What time is it? Will it rain?” In this sentence, you need to use the time skill package and the weather skill package.
  • the answer needs to integrate these two skills packages. For composite intent, then multiple skill packs are required to be executed at the same time, and then combined according to the association between the intents.
  • the answer returned by each skill package is a partial answer. They know the intent to associate data between the links. Contact, and then through the partial answers obtained by each skill package to form a complete answer, according to the relationship between the intentions, the answers are combined and combined to get the complete answer. For example, in the above case, the robot's time skill package is now 9:00 am on June 20, 2016. The weather skill package is sunny today, then the robot will reply "now 9 am, today's weather is fine", thus The answers to each skill pack are summarized to get the final answer.
  • the step of obtaining the final answer from the answer set according to the user's intention specifically includes:
  • Each answer in the answer set is compared, and when each answer is complementary, all answers are merged into a final answer based on the user's intent.
  • the step of obtaining the final answer from the answer set according to the user's intention specifically includes:
  • the robot recognizes the user's intention according to the user's question, and then the system sends the user's question to the three skill packs according to the user's intention, and each skill pack sends the question to the question. Answer and then come up with a set of answers with three answers.
  • the percentage of similarity between the user's intention and the answer obtained by the first skill package is 60%
  • the similarity percentage of the answer obtained by the second skill package is 30%
  • the third skill package is obtained.
  • the percentage of similarity of the answer is 10%, then for these three answers, if the three answers are complementary, then the three answers will be merged to get the final answer; and if the three answers are Mutually exclusive, then choose the one with the highest similarity to the user's intention as the final answer. In this example, the answer from the first skill pack is selected as the final answer.
  • the step of answering the question information through the skill package specifically includes:
  • the step of answering the question information through the skill package specifically includes:
  • the user's question message is: Does Zhao Wei's movie look good?
  • the system analyzes the user's intention.
  • the system analyzes the user's intention to know whether Zhao Wei's movie is popular.
  • the system also divides the question information. For example, it is divided into Zhao Wei. , movie, then the system will query the link data, such as the knowledge base, after the query, for example, get the movie "to youth", judge the public's emotional tendency of the movie, whether to support the movie, or to deny the movie. If there is more support, then the movie is good-looking, otherwise it will not look good.
  • the user's question message is, is the bread delicious?
  • the system analyzes the user's intention.
  • the system analyzes the user's intention to know the taste of the bread, then the system will be divided into Bread, good or bad, then the system will check to judge the public's bias, whether you like bread, if you like more, you will say that the bread is delicious, otherwise it will not be good.
  • this embodiment discloses a multi-intent multi-skill package question answering system, including:
  • the obtaining module 201 is configured to obtain question information.
  • the intent identification module 202 is configured to identify the user's intention according to the question information
  • the parsing module 203 is configured to solve the question information by using at least two skill packages to obtain an answer set including at least two answers;
  • the output module 204 is configured to obtain a final answer from the set of answers according to the user's intention.
  • the user's question information can be answered by at least two skill packages, and then all the answers are obtained and then selected according to the user's intention, thereby obtaining the final answer, so that the obtained answer is more accurate, and by setting multiple skill packages. Can make the question and answer method more widely applicable.
  • the skill package includes at least a music skill package, a story skill package, a star technology package, and a horoscope skill package.
  • the user's questions can be more comprehensively queried and analyzed through different skill packages, and different answers are obtained, and then the answers are combined to obtain the final answer.
  • the parsing module is further configured to: input the question information into at least two link data corresponding to the at least two skill packages for answering;
  • the output module is further configured to: associate the acquired at least two user intents by a relationship between the at least two link data, and select at least two of the acquired answer sets according to the association between the at least two user intents The answers are combined to get the final answer.
  • the user asks for the message: "What time is it? Will it rain?” In this sentence, you need to use the time skill package and the weather skill package.
  • the answer needs to integrate these two skills packages. For composite intent, then multiple skill packs are required to be executed at the same time, and then combined according to the association between the intents.
  • the answer returned by each skill package is a partial answer. They know the intent to associate data between the links. Contact, and then through the partial answers obtained by each skill package to form a complete answer, according to the relationship between the intentions, the answers are combined and combined to get the complete answer. For example, in the above case, the robot's time skill package is now 9:00 am on June 20, 2016. The weather skill package is sunny today, then the robot will reply "now 9 am, today's weather is fine", thus The answers to each skill pack are summarized to get the final answer.
  • the output module is specifically configured to:
  • Each answer in the answer set is compared, and when each answer is complementary, all answers are merged into a final answer based on the user's intent.
  • the output module is specifically configured to:
  • the robot recognizes the user's intention according to the user's question, and then the system sends the user's question to the three skill packs according to the user's intention, and each skill pack sends the question to the question. Answer and then come up with a set of answers with three answers.
  • the percentage of similarity between the user's intention and the answer obtained by the first skill package is 60%
  • the similarity percentage of the answer obtained by the second skill package is 30%
  • the third skill package is obtained.
  • the percentage of similarity of the answer is 10%, then for these three answers, if the three answers are complementary, then the three answers will be merged to get the final answer; and if the three answers are Mutually exclusive, then choose the one with the highest similarity to the user's intention as the final answer. In this example, the answer from the first skill pack is selected as the final answer.
  • the parsing module is specifically configured to: segment the input question information
  • the parsing module is specifically configured to:
  • the user's question message is: Does Zhao Wei's movie look good?
  • the system analyzes the user's intention.
  • the system analyzes the user's intention to know whether Zhao Wei's movie is popular.
  • the system also divides the question information. For example, it is divided into Zhao Wei. , movie, then the system will query the link data, such as the knowledge base, after the query, for example, get the movie "to youth", judge the public's emotional tendency of the movie, whether to support the movie, or to deny the movie. If there is more support, then the movie is good-looking, otherwise it will not look good.
  • the user's question message is, is the bread delicious?
  • the system analyzes the user's intention.
  • the system analyzes the user's intention to know the taste of the bread, then the system will be divided into Bread, good or bad, then the system will check to judge the public's bias, whether you like bread, if you like more, you will say that the bread is delicious, otherwise it will not be good.
  • the embodiment further discloses a robot comprising a multi-intent based multi-skilled question answering system as described above.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Mathematical Physics (AREA)
  • Databases & Information Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Machine Translation (AREA)
  • Manipulator (AREA)

Abstract

一种基于多意图的多技能包问答方法,包括:获取提问信息(S101);根据提问信息识别用户意图(S102);通过至少两个技能包对提问信息进行解答,得到包括至少两个答案的答案集合(S103);根据用户意图从答案集合中获取最终答案(S104)。可以通过至少两个技能包对用户的提问信息进行解答,然后对所有得到答案再根据用户的意图进行挑选或关联组合,从而获得最终答案,其中关联组合可以是将根据用户的意图,将多个技能包得出的答案进行组合。所述方法可以使答案更精确,适用范围更广。

Description

一种基于多意图的多技能包问答方法、***和机器人 技术领域
本发明涉及机器人交互技术领域,尤其涉及一种多意图的基于多技能包问答方法、***和机器人。
背景技术
随着社会和科技的发展,机器人越来越多的应用到与人类的工作、生活沟通中,人类可以向机器人提出问题,机器人经过一系列的过程后向人类回复结果,类似于人与人沟通的过程。为了让机器人在与人沟通的过程中更加拟人化,更加模拟真实的与人沟通的场景,人们探索了很多种方法,然而在现有的问答***中,存在回答的答案准确率不高的情况。
因此,如何提供一种回答的答案正确率更高的问答方法,成为亟需解决的技术问题。
发明内容
本发明的目的是提供一种基于多意图的多技能包问答方法、***和机器人,以实现适用范围更广的问答和机器人交互。
本发明的目的是通过以下技术方案来实现的:
一种基于多意图的多技能包问答方法,包括:
获取提问信息;
根据提问信息识别用户意图;
通过至少两个技能包对提问信息进行解答,得到包括至少两个答案的答案集合;
根据用户意图从答案集合中获取最终答案。
优选的,所述通过至少两个技能包对提问信息进行解答的步骤具体包括:
将提问信息输入到与至少两个技能包对应的至少两个链接数据中进行解答;
所述根据用户意图从答案集合中获取最终答案的步骤具体包括:
通过至少两个链接数据之间的关系,对获取的至少两个用户意图进行 关联,根据至少两个用户意图之间的关联,将获取的答案集合中的至少两个答案进行关联组合,得到最终答案。
优选的,所述通过技能包对提问信息进行解答的步骤具体包括:
将输入的提问信息进行切分;
根据切分的结果在链接数据库中查找中间结果;
将中间结果输入到数据库中获取情感偏向,根据情感偏向获取答案。
优选的,所述通过技能包对提问信息进行解答的步骤具体包括:
将输入的提问信息进行切分;
根据切分的结果在数据库中查询获取答案。
优选的,所述根据用户意图从答案集合中获取最终答案的步骤具体包括:
将答案集合中的每个答案进行对比,当答案之间有排斥时,将用户意图与每个答案进行对比,选择与用户意图相似度最高的答案作为最终答案。
优选的,所述技能包至少包括音乐技能包、故事技能包、明星百科技能包和星座运势技能包。
本发明公开一种基于多意图的多技能包问答***,包括:
获取模块,用于获取提问信息;
意图识别模块,用于根据提问信息识别用户意图;
解析模块,用于通过至少两个技能包对提问信息进行解答,得到包括至少两个答案的答案集合;
输出模块,用于根据用户意图从答案集合中获取最终答案。
优选的,所述解析模块还用于:将提问信息输入到与至少两个技能包对应的至少两个链接数据中进行解答;
所述输出模块还用于:通过至少两个链接数据之间的关系,对获取的至少两个用户意图进行关联,根据至少两个用户意图之间的关联,将获取的答案集合中的至少两个答案进行关联组合,得到最终答案。
优选的,所述解析模块具体用于:将输入的提问信息进行切分;
根据切分的结果在链接数据库中查找中间结果;
将中间结果输入到数据库中获取情感偏向,根据情感偏向获取答案。。
优选的,所述解析模块具体用于:
将输入的提问信息进行切分;
根据切分的结果在数据库中查询获取答案。
优选的,所述输出模块具体用于:
将答案集合中的每个答案进行对比,当答案之间有排斥时,将用户意图与每个答案进行对比,选择与用户意图相似度最高的答案作为最终答案。
优选的,所述技能包至少包括音乐技能包、故事技能包、明星百科技能包和星座运势技能包。
本发明公开一种机器人,包括如上述任一所述的一种基于多意图的多技能包问答***。
相比现有技术,本发明具有以下优点:本发明公开的一种基于多技能包的问答方法,包括:获取提问信息;根据提问信息识别用户意图;通过至少两个技能包对提问信息进行解答,得到包括至少两个答案的答案集合;根据用户意图从答案集合中获取最终答案。这样就可以通过至少两个技能包对用户的提问信息进行解答,然后对所有得到答案再根据用户的意图进行挑选或关联组合等,从而获得最终答案,其中关联组合可以是将根据用户的意图,将多个技能包得出的答案进行组合,这样获得的答案就更加精确,而且通过设置多个技能包,可以让问答方法的适用范围更加广。本发明首次提出以技能包的管理方式,对机器人及机器人中的功能模块等进行管理,在并行管理的框架下,可以进一步地提高机器人的处理速度和效率,方便机器人更快更便捷的实现功能的启动,提高了机器人与用户交互的效率和用户好感度。
附图说明
图1是本发明实施例一的一种基于多意图的多技能包问答方法的流程图;
图2是本发明实施例二的一种基于多意图的多技能包问答***的示意图。
具体实施方式
虽然流程图将各项操作描述成顺序的处理,但是其中的许多操作可以被并行地、并发地或者同时实施。各项操作的顺序可以被重新安排。当其操作完成时处理可以被终止,但是还可以具有未包括在附图中的附加步骤。处理可以对应于方法、函数、规程、子例程、子程序等等。
计算机设备包括用户设备与网络设备。其中,用户设备或客户端包括 但不限于电脑、智能手机、PDA等;网络设备包括但不限于单个网络服务器、多个网络服务器组成的服务器组或基于云计算的由大量计算机或网络服务器构成的云。计算机设备可单独运行来实现本发明,也可接入网络并通过与网络中的其他计算机设备的交互操作来实现本发明。计算机设备所处的网络包括但不限于互联网、广域网、城域网、局域网、VPN网络等。
在这里可能使用了术语“第一”、“第二”等等来描述各个单元,但是这些单元不应当受这些术语限制,使用这些术语仅仅是为了将一个单元与另一个单元进行区分。这里所使用的术语“和/或”包括其中一个或更多所列出的相关联项目的任意和所有组合。当一个单元被称为“连接”或“耦合”到另一单元时,其可以直接连接或耦合到所述另一单元,或者可以存在中间单元。
这里所使用的术语仅仅是为了描述具体实施例而不意图限制示例性实施例。除非上下文明确地另有所指,否则这里所使用的单数形式“一个”、“一项”还意图包括复数。还应当理解的是,这里所使用的术语“包括”和/或“包含”规定所陈述的特征、整数、步骤、操作、单元和/或组件的存在,而不排除存在或添加一个或更多其他特征、整数、步骤、操作、单元、组件和/或其组合。
下面结合附图和较佳的实施例对本发明作进一步说明。
实施例一
如图1所示,本实施例中公开一种基于多技能包的问答方法,包括:
S101、获取提问信息;
S102、根据提问信息识别用户意图;
S103、通过至少两个技能包对提问信息进行解答,得到包括至少两个答案的答案集合;
S104、根据用户意图从答案集合中获取最终答案。
本发明公开的一种基于多技能包的问答方法,包括:获取提问信息;根据提问信息识别用户意图;通过至少两个技能包对提问信息进行解答,得到包括至少两个答案的答案集合;根据用户意图从答案集合中获取最终答案。这样就可以通过至少两个技能包对用户的提问信息进行解答,然后对所有得到答案再根据用户的意图进行挑选,从而获得最终答案,这样获得的答案就更加精确,而且通过设置多个技能包,可以让问答方法的适用范围更加广。
其中,所述技能包至少包括音乐技能包、故事技能包、明星百科技能包和星座运势技能包。这样就可以通过不同的技能包对用户的提问进行更加全面的查询和分析,得到不同的答案,然后综合这些答案获取最终的答案。
例如,本实施例中,包括第一技能包、第二技能包和第三技能包,从这三个技能包中得出的答案集合中,与用户意图进行对比来获取最终的答案。其中,第一技能包可以是音乐技能包,第二技能包可以是故事技能包,第三技能包可以是星座运势技能包。
根据其中一个示例,,所述通过至少两个技能包对提问信息进行解答的步骤具体包括:
将提问信息输入到与至少两个技能包对应的至少两个链接数据中进行解答;
所述根据用户意图从答案集合中获取最终答案的步骤具体包括:
通过至少两个链接数据之间的关系,对获取的至少两个用户意图进行关联,根据至少两个用户意图之间的关联,将获取的答案集合中的至少两个答案进行关联组合,得到最终答案。
例如,用户提问信息为:“现在几点了?会不会下雨”,这句话,需要用到时间技能包和天气技能包,答案就需要整合这两个技能包得来。对于复合意图的,那么就需要多个技能包同时执行,然后根据意图之间的关联来进行合并组合,每个技能包返回的答案是部分答案,他们通过链接数据知道意图关联的数据之间的联系,然后通过每个技能包得到的部分答案进行关联从而形成完整的答案,从而根据意图的之间的关系,将答案进行关联组合,得到完整的答案。例如上述案例,机器人的时间技能包得到的是现在是2016年6月20日上午9点,天气技能包今天天气晴朗,那么机器人就会回复“现在是上午9点,今天天气晴朗”,从而将每个技能包的答案进行汇总,得到最终答案。
根据其中一个示例,所述根据用户意图从答案集合中获取最终答案的步骤具体包括:
将答案集合中的每个答案进行对比,当每个答案之间可互补时,根据用户意图将所有答案融合为最终答案。
根据其中一个示例,所述根据用户意图从答案集合中获取最终答案的步骤具体包括:
将答案集合中的每个答案进行对比,当答案之间有排斥时,将用户意图与每个答案进行对比,选择与用户意图相似度最高的答案作为最终答案。
例如,用户问机器人问题后,机器人根据用户的问句识别用户的意图,然后***根据这个用户的意图将用户的问句向三个技能包都发送,发送后每个技能包对该问句进行回答,然后得出有三个回答的答案集合。得到的答案中,例如用户的意图与第一个技能包得到的答案的相似度百分比为60%,与第二个技能包得到的答案的相似度百分比为30%,与第三个技能包得到的答案的相似度百分比为10%,那么就对于这三个回答,如果三个答案是可以互补的,那么就会将这个三个回答进行融合,得到最终的答案;而如果这三个答案是相互排斥的,那么就会选择一个与用户意图的相似度最高的一个作为最终答案,本示例中选择第一个技能包得到的答案作为最终答案。
根据其中一个示例,所述通过技能包对提问信息进行解答的步骤具体包括:
将输入的提问信息进行切分;
根据切分的结果在链接数据库中查找中间结果;
将中间结果输入到数据库中获取情感偏向,根据情感偏向获取答案。
根据其中另一个示例,所述通过技能包对提问信息进行解答的步骤具体包括:
将输入的提问信息进行切分;
根据切分的结果在数据库中查询获取答案。
例如,用户的提问信息是:赵薇演的电影好看么?在***获取了这样提问信息后,就会分析用户的意图,***分析用户的意图就是想知道赵薇的某一个电影是否受欢迎,***也会对提问信息作出切分,例如,切分为,赵薇,电影,那么***就会查询链接数据,例如知识库,查询之后例如得到电影“致青春”,判断该电影的大众的情感倾向,是支持该电影,还是否定该电影。如果支持的多,那么说明该电影好看,否则不好看。
又如,用户的提问信息是,面包好吃吗,在***获取这样的提问信息后,就会分析用户的意图,***分析用户的意图就是想知道面包的口味,那么***就会切分为,面包,好不好吃,那么***就会查询判断大众的偏向,是否喜欢面包,如果喜欢的多就会说面包好吃,否则就会回复不好吃。
实施例二
如图2所示,本实施例公开一种基于多意图的多技能包问答***,包括:
获取模块201,用于获取提问信息;
意图识别模块202,用于根据提问信息识别用户意图;
解析模块203,用于通过至少两个技能包对提问信息进行解答,得到包括至少两个答案的答案集合;
输出模块204,用于根据用户意图从答案集合中获取最终答案。
这样就可以通过至少两个技能包对用户的提问信息进行解答,然后对所有得到答案再根据用户的意图进行挑选,从而获得最终答案,这样获得的答案就更加精确,而且通过设置多个技能包,可以让问答方法的适用范围更加广。
其中,所述技能包至少包括音乐技能包、故事技能包、明星百科技能包和星座运势技能包。这样就可以通过不同的技能包对用户的提问进行更加全面的查询和分析,得到不同的答案,然后综合这些答案获取最终的答案。
根据其中另一个示例,所述解析模块还用于:将提问信息输入到与至少两个技能包对应的至少两个链接数据中进行解答;
所述输出模块还用于:通过至少两个链接数据之间的关系,对获取的至少两个用户意图进行关联,根据至少两个用户意图之间的关联,将获取的答案集合中的至少两个答案进行关联组合,得到最终答案。
例如,用户提问信息为:“现在几点了?会不会下雨”,这句话,需要用到时间技能包和天气技能包,答案就需要整合这两个技能包得来。对于复合意图的,那么就需要多个技能包同时执行,然后根据意图之间的关联来进行合并组合,每个技能包返回的答案是部分答案,他们通过链接数据知道意图关联的数据之间的联系,然后通过每个技能包得到的部分答案进行关联从而形成完整的答案,从而根据意图的之间的关系,将答案进行关联组合,得到完整的答案。例如上述案例,机器人的时间技能包得到的是现在是2016年6月20日上午9点,天气技能包今天天气晴朗,那么机器人就会回复“现在是上午9点,今天天气晴朗”,从而将每个技能包的答案进行汇总,得到最终答案。
根据其中另一个示例,所述输出模块具体用于:
将答案集合中的每个答案进行对比,当每个答案之间可互补时,根据用户意图将所有答案融合为最终答案。
根据其中另一个示例,所述输出模块具体用于:
将答案集合中的每个答案进行对比,当答案之间有排斥时,将用户意图与每个答案进行对比,选择与用户意图相似度最高的答案作为最终答案。
例如,用户问机器人问题后,机器人根据用户的问句识别用户的意图,然后***根据这个用户的意图将用户的问句向三个技能包都发送,发送后每个技能包对该问句进行回答,然后得出有三个回答的答案集合。得到的答案中,例如用户的意图与第一个技能包得到的答案的相似度百分比为60%,与第二个技能包得到的答案的相似度百分比为30%,与第三个技能包得到的答案的相似度百分比为10%,那么就对于这三个回答,如果三个答案是可以互补的,那么就会将这个三个回答进行融合,得到最终的答案;而如果这三个答案是相互排斥的,那么就会选择一个与用户意图的相似度最高的一个作为最终答案,本示例中选择第一个技能包得到的答案作为最终答案。
根据其中一个示例,所述解析模块具体用于:将输入的提问信息进行切分;
根据切分的结果在链接数据库中查找中间结果;
将中间结果输入到数据库中获取情感偏向,根据情感偏向获取答案。。
根据其中另一个示例,所述解析模块具体用于:
将输入的提问信息进行切分;
根据切分的结果在数据库中查询获取答案。
例如,用户的提问信息是:赵薇演的电影好看么?在***获取了这样提问信息后,就会分析用户的意图,***分析用户的意图就是想知道赵薇的某一个电影是否受欢迎,***也会对提问信息作出切分,例如,切分为,赵薇,电影,那么***就会查询链接数据,例如知识库,查询之后例如得到电影“致青春”,判断该电影的大众的情感倾向,是支持该电影,还是否定该电影。如果支持的多,那么说明该电影好看,否则不好看。
又如,用户的提问信息是,面包好吃吗,在***获取这样的提问信息后,就会分析用户的意图,***分析用户的意图就是想知道面包的口味,那么***就会切分为,面包,好不好吃,那么***就会查询判断大众的偏向,是否喜欢面包,如果喜欢的多就会说面包好吃,否则就会回复不好吃。
本实施例还公开一种机器人,包括如上述任一所述的一种基于多意图的多技能包问答***。
以上内容是结合具体的优选实施方式对本发明所作的进一步详细说明,不能认定本发明的具体实施只局限于这些说明。对于本发明所属技术领域的普通技术人员来说,在不脱离本发明构思的前提下,还可以做出若干简单推演或替换,都应当视为属于本发明的保护范围。

Claims (13)

  1. 一种基于多意图的多技能包问答方法,其特征在于,包括:
    获取提问信息;
    根据提问信息识别用户意图;
    通过至少两个技能包对提问信息进行解答,得到包括至少两个答案的答案集合;
    根据用户意图从答案集合中获取最终答案。
  2. 根据权利要求1所述的问答方法,其特征在于,所述通过至少两个技能包对提问信息进行解答的步骤具体包括:
    将提问信息输入到与至少两个技能包对应的至少两个链接数据中进行解答;
    所述根据用户意图从答案集合中获取最终答案的步骤具体包括:
    通过至少两个链接数据之间的关系,对获取的至少两个用户意图进行关联,根据至少两个用户意图之间的关联,将获取的答案集合中的至少两个答案进行关联组合,得到最终答案。
  3. 根据权利要求1所述的问答方法,其特征在于,所述通过技能包对提问信息进行解答的步骤具体包括:
    将输入的提问信息进行切分;
    根据切分的结果在链接数据库中查找中间结果;
    将中间结果输入到数据库中获取情感偏向,根据情感偏向获取答案。
  4. 根据权利要求1所述的问答方法,其特征在于,所述通过技能包对提问信息进行解答的步骤具体包括:
    将输入的提问信息进行切分;
    根据切分的结果在数据库中查询获取答案。
  5. 根据权利要求1所述的问答方法,其特征在于,所述根据用户意图从答案集合中获取最终答案的步骤具体包括:
    将答案集合中的每个答案进行对比,当答案之间有排斥时,将用户意图与每个答案进行对比,选择与用户意图相似度最高的答案作为最终答案。
  6. 根据权利要求1所述的问答方法,其特征在于,所述技能包至少包括音乐技能包、故事技能包、明星百科技能包和星座运势技能包。
  7. 一种基于多意图的多技能包问答***,其特征在于,包括:
    获取模块,用于获取提问信息;
    意图识别模块,用于根据提问信息识别用户意图;
    解析模块,用于通过至少两个技能包对提问信息进行解答,得到包括至少两个答案的答案集合;
    输出模块,用于根据用户意图从答案集合中获取最终答案。
  8. 根据权利要求7所述的问答***,其特征在于,所述解析模块还用于:将提问信息输入到与至少两个技能包对应的至少两个链接数据中进行解答;
    所述输出模块还用于:通过至少两个链接数据之间的关系,对获取的至少两个用户意图进行关联,根据至少两个用户意图之间的关联,将获取的答案集合中的至少两个答案进行关联组合,得到最终答案。
  9. 根据权利要求7所述的问答***,其特征在于,所述解析模块具体用于:将输入的提问信息进行切分;
    根据切分的结果在链接数据库中查找中间结果;
    将中间结果输入到数据库中获取情感偏向,根据情感偏向获取答案。。
  10. 根据权利要求7所述的问答***,其特征在于,所述解析模块具体用于:
    将输入的提问信息进行切分;
    根据切分的结果在数据库中查询获取答案。
  11. 根据权利要求7所述的问答***,其特征在于,所述输出模块具体用于:
    将答案集合中的每个答案进行对比,当答案之间有排斥时,将用户意图与每个答案进行对比,选择与用户意图相似度最高的答案作为最终答案。
  12. 根据权利要求7所述的问答***,其特征在于,所述技能包至少包括音乐技能包、故事技能包、明星百科技能包和星座运势技能包。
  13. 一种机器人,其特征在于,包括如权利要求7至12任一所述的一种基于多意图的多技能包问答***。
PCT/CN2016/087517 2016-06-28 2016-06-28 一种基于多意图的多技能包问答方法、***和机器人 WO2018000205A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN201680001740.4A CN106462647A (zh) 2016-06-28 2016-06-28 一种基于多意图的多技能包问答方法、***和机器人
PCT/CN2016/087517 WO2018000205A1 (zh) 2016-06-28 2016-06-28 一种基于多意图的多技能包问答方法、***和机器人

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2016/087517 WO2018000205A1 (zh) 2016-06-28 2016-06-28 一种基于多意图的多技能包问答方法、***和机器人

Publications (1)

Publication Number Publication Date
WO2018000205A1 true WO2018000205A1 (zh) 2018-01-04

Family

ID=58215737

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2016/087517 WO2018000205A1 (zh) 2016-06-28 2016-06-28 一种基于多意图的多技能包问答方法、***和机器人

Country Status (2)

Country Link
CN (1) CN106462647A (zh)
WO (1) WO2018000205A1 (zh)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109408800A (zh) * 2018-08-23 2019-03-01 优视科技(中国)有限公司 对话机器人***及相关技能配置方法
WO2020109344A1 (en) 2018-11-29 2020-06-04 F. Hoffmann-La Roche Ag Occular administration device for antisense oligonucleotides
WO2020109343A1 (en) 2018-11-29 2020-06-04 F. Hoffmann-La Roche Ag Combination therapy for treatment of macular degeneration
CN111710338A (zh) * 2020-06-28 2020-09-25 上海优扬新媒信息技术有限公司 一种话术播放方法及装置
US20220198292A1 (en) * 2020-12-21 2022-06-23 International Business Machines Corporation Automation of virtual assistant training
WO2023178192A1 (en) 2022-03-15 2023-09-21 Compugen Ltd. Il-18bp antagonist antibodies and their use in monotherapy and combination therapy in the treatment of cancer

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111125320A (zh) * 2018-10-31 2020-05-08 重庆小雨点小额贷款有限公司 数据处理方法、装置、服务器及计算机可读存储介质
CN109783733B (zh) * 2019-01-15 2020-11-06 腾讯科技(深圳)有限公司 用户画像生成装置及方法、信息处理装置及存储介质
CN110428263A (zh) * 2019-06-19 2019-11-08 平安科技(深圳)有限公司 客户预判方法、装置、终端及存储介质
CN112784146A (zh) * 2019-11-04 2021-05-11 北京搜狗科技发展有限公司 一种数据处理方法、装置和用于数据处理的装置
CN112148848A (zh) * 2020-08-28 2020-12-29 出门问问(苏州)信息科技有限公司 一种问答处理方法及装置
CN113282727B (zh) * 2021-06-03 2024-04-16 北京捷通华声科技股份有限公司 问答处理方法、装置、计算机可读存储介质及处理器
CN114661885B (zh) * 2022-05-26 2022-10-11 深圳追一科技有限公司 问答处理方法、装置、计算机设备和存储介质

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103229162A (zh) * 2010-09-28 2013-07-31 国际商业机器公司 使用候选答案逻辑综合提供问题答案
US20150161230A1 (en) * 2013-12-11 2015-06-11 International Business Machines Corporation Generating an Answer from Multiple Pipelines Using Clustering
CN104933084A (zh) * 2015-05-04 2015-09-23 上海智臻网络科技有限公司 一种用于获得答案信息的方法、装置和设备
CN105068661A (zh) * 2015-09-07 2015-11-18 百度在线网络技术(北京)有限公司 基于人工智能的人机交互方法和***
CN105471712A (zh) * 2015-11-25 2016-04-06 深圳狗尾草智能科技有限公司 一种机器人答复***及其答复方法
CN105487663A (zh) * 2015-11-30 2016-04-13 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***
CN105512228A (zh) * 2015-11-30 2016-04-20 北京光年无限科技有限公司 一种基于智能机器人的双向问答数据处理方法和***
CN105574133A (zh) * 2015-12-15 2016-05-11 苏州贝多环保技术有限公司 一种多模态的智能问答***及方法
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***

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104598445B (zh) * 2013-11-01 2019-05-10 腾讯科技(深圳)有限公司 自动问答***和方法

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103229162A (zh) * 2010-09-28 2013-07-31 国际商业机器公司 使用候选答案逻辑综合提供问题答案
US20150161230A1 (en) * 2013-12-11 2015-06-11 International Business Machines Corporation Generating an Answer from Multiple Pipelines Using Clustering
CN104933084A (zh) * 2015-05-04 2015-09-23 上海智臻网络科技有限公司 一种用于获得答案信息的方法、装置和设备
CN105068661A (zh) * 2015-09-07 2015-11-18 百度在线网络技术(北京)有限公司 基于人工智能的人机交互方法和***
CN105471712A (zh) * 2015-11-25 2016-04-06 深圳狗尾草智能科技有限公司 一种机器人答复***及其答复方法
CN105487663A (zh) * 2015-11-30 2016-04-13 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***
CN105512228A (zh) * 2015-11-30 2016-04-20 北京光年无限科技有限公司 一种基于智能机器人的双向问答数据处理方法和***
CN105574133A (zh) * 2015-12-15 2016-05-11 苏州贝多环保技术有限公司 一种多模态的智能问答***及方法
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***

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109408800A (zh) * 2018-08-23 2019-03-01 优视科技(中国)有限公司 对话机器人***及相关技能配置方法
CN109408800B (zh) * 2018-08-23 2024-03-01 阿里巴巴(中国)有限公司 对话机器人***及相关技能配置方法
WO2020109344A1 (en) 2018-11-29 2020-06-04 F. Hoffmann-La Roche Ag Occular administration device for antisense oligonucleotides
WO2020109343A1 (en) 2018-11-29 2020-06-04 F. Hoffmann-La Roche Ag Combination therapy for treatment of macular degeneration
CN111710338A (zh) * 2020-06-28 2020-09-25 上海优扬新媒信息技术有限公司 一种话术播放方法及装置
US20220198292A1 (en) * 2020-12-21 2022-06-23 International Business Machines Corporation Automation of virtual assistant training
WO2023178192A1 (en) 2022-03-15 2023-09-21 Compugen Ltd. Il-18bp antagonist antibodies and their use in monotherapy and combination therapy in the treatment of cancer

Also Published As

Publication number Publication date
CN106462647A (zh) 2017-02-22

Similar Documents

Publication Publication Date Title
WO2018000205A1 (zh) 一种基于多意图的多技能包问答方法、***和机器人
CN110347863B (zh) 话术推荐方法和装置及存储介质
CN105592237B (zh) 一种会话切换的方法、装置及智能客服机器人
US10834145B2 (en) Providing of recommendations determined from a collaboration session system and method
US20180190295A1 (en) Voice recognition
CN109214674B (zh) 客服***管理方法、客服***及电子设备
CN110019744A (zh) 辅助生成会议纪要的方法、装置、设备及计算机存储介质
CN107170446A (zh) 语义处理服务器及用于语义处理的方法
CN103561086A (zh) 呈现日程提醒信息的方法、终端设备及云服务器
KR20210157848A (ko) 컴퓨터로 구현하는 회의 예약 방법, 장치, 기기 및 매체
WO2016179939A1 (zh) 一种智能客服处理方法及其装置
US11410568B2 (en) Dynamic evaluation of event participants using a smart context-based quiz system
CN109544417A (zh) 一种学习效果确定方法、装置、存储介质及终端设备
CN109389870A (zh) 一种应用于电子教学中的数据自适应调整方法及其装置
CN110609970B (zh) 用户身份识别方法、装置、存储介质及电子设备
US20200250608A1 (en) Providing feedback by evaluating multi-modal data using machine learning techniques
CN111223015B (zh) 课程推荐方法、装置及终端设备
CN104882039A (zh) 一种基于移动终端的教学互动方法及服务器
CN109002434A (zh) 客服问答匹配方法、服务器及存储介质
US10841368B2 (en) Method for presenting schedule reminder information, terminal device, and cloud server
CN110659368A (zh) 知识图谱构建方法、装置、电子设备及可读存储介质
US20230403328A1 (en) Arrangements for detecting bi-directional artificial intelligence (ai) voice communications and negotiating direct digital communications
WO2016184341A1 (zh) 一种业务处理方法和设备
CN110532565B (zh) 语句处理方法及装置、以及电子设备
CN109840446A (zh) 多人脸识别引擎的管理控制方法、装置及***

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: 16906607

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: 16906607

Country of ref document: EP

Kind code of ref document: A1