WO2021120779A1 - 一种基于人机对话的用户画像构建方法、系统、终端及存储介质 - Google Patents

一种基于人机对话的用户画像构建方法、系统、终端及存储介质 Download PDF

Info

Publication number
WO2021120779A1
WO2021120779A1 PCT/CN2020/118445 CN2020118445W WO2021120779A1 WO 2021120779 A1 WO2021120779 A1 WO 2021120779A1 CN 2020118445 W CN2020118445 W CN 2020118445W WO 2021120779 A1 WO2021120779 A1 WO 2021120779A1
Authority
WO
WIPO (PCT)
Prior art keywords
attribute
data
dialogue
sentence
subject
Prior art date
Application number
PCT/CN2020/118445
Other languages
English (en)
French (fr)
Inventor
王硕
吴振宇
王建明
Original Assignee
平安科技(深圳)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 平安科技(深圳)有限公司 filed Critical 平安科技(深圳)有限公司
Publication of WO2021120779A1 publication Critical patent/WO2021120779A1/zh

Links

Images

Classifications

    • 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
    • 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
    • G06F16/35Clustering; Classification
    • 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
    • G06F16/36Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367Ontology
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/048Activation functions
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02DCLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00Energy efficient computing, e.g. low power processors, power management or thermal management

Definitions

  • This application relates to the field of human-computer interaction technology and is applied to the field of artificial intelligence, and in particular to a method, system, terminal and storage medium for constructing a user portrait based on human-machine dialogue.
  • dialogue robots have been widely used in many scenarios, such as task robots in vertical scenarios, question answering robots, and small chat robots in open scenarios.
  • the dialog robot often receives some descriptions of the user’s personal information, automatically extracts the user attribute information contained in the human-computer interaction and constructs the user portrait, and then provides the user with different scenarios in a targeted manner.
  • the marketing recommendation under this article has great commercial value.
  • This application provides a method, system, terminal, and storage medium for constructing a user portrait based on human-machine dialogue, which can solve the deficiencies in the prior art to a certain extent.
  • a user portrait construction method based on man-machine dialogue including:
  • the attribute classification model includes a dialog encoder, an attribute classifier, and an entity generator.
  • the dialog data is encoded by the dialog encoder, and all the data are encoded by the attribute classifier.
  • the encoded dialogue data is classified by attribute type, the subject and object corresponding to each attribute type are extracted through the entity generator, and the attribute classification model splices the subject and object with the corresponding attribute type, and outputs the subject and object corresponding to the corresponding attribute type. , Attribute type and triple information of the object;
  • a user portrait construction system based on human-machine dialogue including:
  • Data acquisition module used to acquire dialogue data in the process of human-computer interaction
  • Attribute information extraction module used to input the dialog data into an attribute classification model, the attribute classification model includes a dialog encoder, an attribute classifier, and an entity generator.
  • the dialog data is encoded by the dialog encoder.
  • the attribute classifier classifies the encoded dialogue data by attribute type, extracts the subject and object corresponding to each attribute type through the entity generator, and the attribute classification model classifies the subject and object with the corresponding attribute type Splicing, and output the triple information composed of subject, attribute type and object;
  • User portrait construction module used to construct a user portrait according to the triplet information.
  • a terminal includes a processor and a memory coupled to the processor, wherein:
  • the memory stores program instructions that can be executed by the processor
  • the attribute classification model includes a dialog encoder, an attribute classifier, and an entity generator.
  • the dialog data is encoded by the dialog encoder, and all the data are encoded by the attribute classifier.
  • the encoded dialogue data is classified by attribute type, the subject and object corresponding to each attribute type are extracted through the entity generator, and the attribute classification model splices the subject and object with the corresponding attribute type, and outputs the subject and object corresponding to the corresponding attribute type. , Attribute type and triple information of the object;
  • a storage medium storing program instructions that can be run by a processor, and when the program instructions are executed by the processor, the following steps are implemented:
  • the attribute classification model includes a dialog encoder, an attribute classifier, and an entity generator.
  • the dialog data is encoded by the dialog encoder, and all the data are encoded by the attribute classifier.
  • the encoded dialogue data is classified by attribute type, the subject and object corresponding to each attribute type are extracted through the entity generator, and the attribute classification model splices the subject and object with the corresponding attribute type, and outputs the subject and object corresponding to the corresponding attribute type. , Attribute type and triple information of the object;
  • the method, system, terminal, and storage medium for constructing user portraits based on human-machine dialogue in the embodiments of this application can build an attribute classification model that includes a dialogue encoder, an attribute classifier, and an entity generator.
  • the explicit or implicit user attribute information of different attribute types in the human-machine dialogue is automatically extracted, user portraits are constructed according to the extracted attribute information, and marketing tasks in different scenarios are recommended according to the user portraits.
  • the embodiments of the present application can be deployed in different scenarios and tasks, which improves the flexibility and accuracy of extracting user attribute information.
  • FIG. 1 is a schematic flowchart of a method for constructing a user portrait based on a human-machine dialogue according to a first embodiment of the present application
  • FIG. 2 is a schematic diagram of the structure of an attribute classification model according to an embodiment of the present application.
  • FIG. 3 is a schematic flowchart of a method for constructing a user portrait based on a human-machine dialogue according to a second embodiment of the present application;
  • FIG. 4 is a schematic structural diagram of a user portrait construction system based on human-machine dialogue according to an embodiment of the present application
  • FIG. 5 is a schematic diagram of a terminal structure according to an embodiment of the present application.
  • FIG. 6 is a schematic diagram of the structure of a storage medium according to an embodiment of the present application.
  • first”, “second”, and “third” in this application are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Thus, the features defined with “first”, “second”, and “third” may explicitly or implicitly include at least one of the features.
  • "a plurality of” means at least two, such as two, three, etc., unless otherwise specifically defined. All directional indicators (such as up, down, left, right, front, back%) in the embodiments of this application are only used to explain the relative positional relationship between the components in a specific posture (as shown in the drawings) , Movement status, etc., if the specific posture changes, the directional indication will also change accordingly.
  • FIG. 1 is a schematic flowchart of a method for constructing a user portrait based on a human-machine dialogue in a first embodiment of the present application.
  • the method for constructing a user portrait based on human-machine dialogue in the first embodiment of the present application includes the following steps:
  • S10 Sampling historical human-computer interaction dialogue data to obtain human-machine dialogue sample data
  • the human-machine dialogue sample data is the dialogue between the user and the robot or artificial agent, and the human-machine dialogue data sample includes voice data or/and text data.
  • the human-machine dialogue sample data is voice data
  • the voice data is converted into text format through ASR (Automatic Speech Recognition) technology
  • ASR Automatic Speech Recognition
  • S12 Annotate the user attribute information in the text data according to the triple form of (subject, attribute type, and object) to generate a data set of the training model;
  • the user attribute information is the description of the user's personal information in the process of man-machine dialogue, such as marital status, children's situation, and professional information. For example, if a sentence is "My son is 7 years old this year", the user attribute information label corresponding to the sentence is (son, age, 7 years old).
  • the structure of the attribute classification model is shown in Figure 2, which includes a dialog encoder, an attribute classifier, and an entity generator.
  • the dialog encoder encodes the text data, and then uses the output of the dialog encoder as the attribute classifier. Input, get a loss related to multi-label classification, and then each activated attribute type is input into the entity generator together with the output of the dialogue encoder, and the subject and object corresponding to each attribute type are generated.
  • the entity generator part will also correspond to a sequence For related losses, the training of the final model is driven by these two parts of losses, forming a framework for multi-task learning.
  • the embodiment of the application adopts a supervised method to perform model training, and the specific training process includes:
  • the dialogue encoder uses GRU (Gated Recurrent Units) to encode text data.
  • the dialog encoder includes a two-way GRU and a one-way GRU.
  • the input received by the two-way GRU is a sequence composed of word embedding vectors (Word Embedding), which is used to separately encode each sentence in the text data; one-way GRU
  • the received input is a sequence composed of encoded sentence embedding vectors (Utterance Embedding), which is used to separately encode the context information of each sentence in the text data.
  • e(wi ,j ) represents the embedding vector of the word w i,j.
  • the new sentence embedding expression contains the context information of each sentence.
  • the embedding expression H i incorporates the information from the previous text u 1 to u i-1 and u i .
  • attribute classifier uses the attribute classifier Designed as a multi-label classifier.
  • attribute classifier uses a simple fully connected layer, that is, a linear mapping layer plus a sigmoid activation function layer, as shown below:
  • P i represents the probability that the i-th sentence in all properties predefined types of distribution
  • represents a sigmoid activation function
  • W and b are the weight parameters of the linear mapping
  • H i is the i-way GRU encoded by The embedding expression vector of a sentence.
  • the entity generator uses GRU to extract the subject and object.
  • the entity generation is defined as a sequence generation problem.
  • the target sequence of the entity generator is the sequence formed by the subject and the object.
  • the input at the beginning of the GRU is the embedding vector of a certain attribute type, and the hidden state at the beginning (Hidden state) is the hidden state obtained at the last moment when the corresponding sentence is encoded by the one-way GRU, so that each attribute type corresponds to an entity generator.
  • the entity generator outputs the subject and object corresponding to each attribute type, and then the entity The output result of the generator is spliced with the corresponding attribute type, and finally the triple information composed of subject, attribute type and object is obtained.
  • the method for constructing user portraits based on human-machine dialogue in the embodiment of the present application can automatically extract different human-machine dialogues under a unified framework by constructing an attribute classification model that includes a dialogue encoder, an attribute classifier, and an entity generator. Explicit or implicit user attribute information of the attribute type, construct user portraits based on the extracted attribute information, and recommend marketing tasks in different scenarios according to the user portraits.
  • the embodiments of the present application can be deployed in different scenarios and tasks, which improves the flexibility and accuracy of extracting user attribute information.
  • FIG. 3 is a schematic flowchart of a method for constructing a user portrait based on a human-machine dialogue according to a second embodiment of the present application.
  • the method for constructing a user portrait based on human-machine dialogue in the second embodiment of the present application includes the following steps:
  • S21 Input the text data into the attribute classification model obtained by training, extract the user attribute information contained in the text data through the attribute classification model, and output the triple information consisting of subject, attribute type, and object;
  • the corresponding summary information is obtained based on the result of the user portrait construction method based on the human-machine dialogue.
  • the summary information is obtained by hashing the result of the user portrait construction method based on the human-machine dialogue.
  • Uploading summary information to the blockchain can ensure its security and fairness and transparency to users.
  • the user can download the summary information from the blockchain to verify whether the result of the user portrait construction method based on man-machine dialogue has been tampered with.
  • the blockchain referred to in this example is a new application mode of computer technology such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm.
  • Blockchain essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information for verification. The validity of the information (anti-counterfeiting) and the generation of the next block.
  • the blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.
  • FIG. 4 is a schematic structural diagram of a user portrait construction system based on human-machine dialogue in an embodiment of the present application.
  • the user portrait construction system 40 based on man-machine dialogue in the embodiment of the present application includes:
  • Data acquisition module 41 used to acquire dialogue data in the process of human-computer interaction
  • the attribute information extraction module 42 is used to input the dialog data into an attribute classification model, the attribute classification model includes a dialog encoder, an attribute classifier, and an entity generator, and the dialog data is encoded by the dialog encoder, The encoded dialogue data is classified by the attribute type by the attribute classifier, the subject and the object corresponding to each attribute type are extracted by the entity generator, and the attribute classification model classifies the subject and the object with the corresponding attribute. Type splicing, and output the triple information composed of subject, attribute type and object;
  • User portrait construction module 43 used to construct a user portrait according to the triplet information.
  • FIG. 5 is a schematic diagram of a terminal structure according to an embodiment of the application.
  • the terminal 50 includes a processor 51 and a memory 52 coupled to the processor 51.
  • the memory 52 stores program instructions for realizing the above-mentioned method for constructing a user portrait based on a human-machine dialogue.
  • the processor 51 is configured to execute program instructions stored in the memory 52 to perform a user portrait construction operation based on a human-machine dialogue.
  • the processor 51 may also be referred to as a CPU (Central Processing Unit, central processing unit).
  • the processor 51 may be an integrated circuit chip with signal processing capability.
  • the processor 51 may also be a general-purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component.
  • DSP digital signal processor
  • ASIC application specific integrated circuit
  • FPGA off-the-shelf programmable gate array
  • the general-purpose processor may be a microprocessor or the processor may also be any conventional processor or the like.
  • FIG. 6 is a schematic structural diagram of a storage medium according to an embodiment of the application.
  • the storage medium of the embodiment of the present application stores a program file 61 that can implement all the above methods, and the storage medium may be non-volatile or volatile.
  • the program file 61 may be stored in the above-mentioned storage medium in the form of a software product, and includes a number of instructions to make a computer device (may be a personal computer, a server, or a network device, etc.) or a processor to execute the program. Apply for all or part of the steps of each implementation method.
  • the aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disks or optical disks and other media that can store program codes.
  • terminal devices such as computers, servers, mobile phones, and tablets.
  • the disclosed system, device, and method can be implemented in other ways.
  • the system embodiment described above is only illustrative.
  • the division of units is only a logical function division, and there may be other divisions in actual implementation, for example, multiple units or components can be combined or integrated. To another system, or some features can be ignored, or not implemented.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms.
  • the functional units in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
  • the above-mentioned integrated unit can be implemented in the form of hardware or software functional unit. The above are only implementations of this application, and do not limit the scope of this application. Any equivalent structure or equivalent process transformation made using the content of the description and drawings of this application, or directly or indirectly applied to other related technical fields, The same reasoning is included in the scope of patent protection of this application.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Biophysics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Databases & Information Systems (AREA)
  • Human Computer Interaction (AREA)
  • Animal Behavior & Ethology (AREA)
  • Machine Translation (AREA)

Abstract

一种基于人机对话的用户画像构建方法、系统、终端及存储介质。所述方法包括:获取人机交互过程中的对话数据;将对话数据转换成文本数据(S20);将文本数据输入属性分类模型,通过属性分类模型提取文本数据中包含的用户属性信息,并输出由主语、属性类型和宾语组成的三元组信息(S21);根据三元组信息构建用户画像(S22)。能够在统一框架下自动提取人机对话中不同属性类型的显式或隐式用户属性信息,提升了用户属性信息提取的灵活性和准确性。

Description

一种基于人机对话的用户画像构建方法、系统、终端及存储介质
本申请要求于2020年08月06日提交中国专利局、申请号为202010784250.4,发明名称为“一种基于人机对话的用户画像构建方法、系统、终端及存储介质”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
技术领域
本申请涉及人机交互技术领域,应用于人工智能领域中,特别是涉及一种基于人机对话的用户画像构建方法、系统、终端及存储介质。
背景技术
随着智能对话技术的发展,对话机器人在众多的场景中得到了广泛的应用,例如垂直场景下的任务型机器人、问答机器人和开放场景下的闲聊机器人等。在用户与对话机器人的交互过程中,对话机器人常常会接收到一些关于用户个人信息的描述,自动化地提取人机交互中包含的用户属性信息并构建用户画像,进而针对性地向用户提供不同场景下的营销推荐,具有很大的商业价值。
传统的人机交互用户属性信息提取方法通常采用基于规则或者一些简单的机器学习方法,但发明人意识到这些方法仅能从句式相对固定的描述中提取到显式用户属性信息,而很难捕捉到用户的隐式属性表达。另外,发明人还发现对于不同类型的用户信息,往往需要设计不同的提取方法,然而目前并没有统一的用户信息提取框架,因此传统的人机交互用户属性信息提取方法灵活性和准确性都相对较差。
发明内容
本申请提供了一种基于人机对话的用户画像构建方法、系统、终端及存储介质,能够在一定程度上解决现有技术中存在的不足。
为解决上述技术问题,本申请采用的技术方案为:
一种基于人机对话的用户画像构建方法,包括:
获取人机交互过程中的对话数据;
将所述对话数据输入属性分类模型,所述属性分类模型包括对话编码器、属性分类器和实体生成器,通过所述对话编码器对所述对话数据进行编码,通过所述属性分类器对所述编码后的对话数据进行属性类型分类,通过所述实体生成器提取各个属性类型对应的主语和宾语,所述属性分类模型将所述主语和宾语与对应的属性类型进行拼接,并输出由主语、属性类型和宾语组成的三元组信息;
根据所述三元组信息构建用户画像。
本申请实施例采取的另一技术方案为:一种基于人机对话的用户画像构建系统,包括:
数据获取模块:用于获取人机交互过程中的对话数据;
属性信息提取模块:用于将所述对话数据输入属性分类模型,所述属性分类模型包括对话编码器、属性分类器和实体生成器,通过所述对话编码器对所述对话数据进行编码,通过所述属性分类器对所述编码后的对话数据进行属性类型分类,通过所述实体生成器提取各个属性类型对应的主语和宾语,所述属性分类模型将所述主语和宾语与对应的属性类型进行拼接,并输出由主语、属性类型和宾语组成的三元组信息;
用户画像构建模块:用于根据所述三元组信息构建用户画像。
本申请实施例采取的又一技术方案为:一种终端,所述终端包括处理器、与所述处理器耦接的存储器,其中,
所述存储器存储有可被所述处理器执行的程序指令;
所述处理器执行所述存储器存储的所述程序指令时实现以下步骤:
获取人机交互过程中的对话数据;
将所述对话数据输入属性分类模型,所述属性分类模型包括对话编码器、属性分类器和实体生成器,通过所述对话编码器对所述对话数据进行编码,通过所述属性分类器对所述编码后的对话数据进行属性类型分类,通过所述实体生成器提取各个属性类型对应的主语和宾语,所述属性分类模型将所述主语和宾语与对应的属性类型进行拼接,并输出由主语、属性类型和宾语组成的三元组信息;
根据所述三元组信息构建用户画像。
本申请实施例采取的又一技术方案为:一种存储介质,存储有处理器可运行的程序指令,所述程序指令被处理器执行时实现以下步骤:
获取人机交互过程中的对话数据;
将所述对话数据输入属性分类模型,所述属性分类模型包括对话编码器、属性分类器和实体生成器,通过所述对话编码器对所述对话数据进行编码,通过所述属性分类器对所述编码后的对话数据进行属性类型分类,通过所述实体生成器提取各个属性类型对应的主语和宾语,所述属性分类模型将所述主语和宾语与对应的属性类型进行拼接,并输出由主语、属性类型和宾语组成的三元组信息;
根据所述三元组信息构建用户画像。
本申请的有益效果是:本申请实施例的基于人机对话的用户画像构建方法、系统、终端及存储介质通过构建包括对话编码器、属性分类器和实体生成器为一体的属性分类模型,能够在统一框架下自动提取人机对话中不同属性类型的显式或隐式用户属性信息,根据提取的属性信息构建用户画像,根据用户画像进行不同场景的营销任务推荐。本申请实施例可以部署到不同的场景任务中,提升了用户属性信息提取的灵活性和准确性。
附图说明
图1是本申请第一实施例的基于人机对话的用户画像构建方法的流程示意图;
图2是本申请实施例的属性分类模型结构示意图;
图3是本申请第二实施例的基于人机对话的用户画像构建方法的流程示意图;
图4是本申请实施例基于人机对话的用户画像构建系统的结构示意图;
图5是本申请实施例的终端结构示意图;
图6是本申请实施例的存储介质结构示意图。
具体实施方式
下面将结合本申请实施例中的附图,对本申请实施例中的技术方案进行清楚、完整地描述,显然,所描述的实施例仅是本申请的一部分实施例,而不是全部的实施例。基于本申请中的实施例,本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例,都属于本申请保护的范围。
本申请中的术语“第一”、“第二”、“第三”仅用于描述目的,而不能理解为指示或暗示相对重要性或者隐含指明所指示的技术特征的数量。由此,限定有“第一”、“第二”、“第三”的特征可以明示或者隐含地包括至少一个该特征。本申请的描述中,“多个”的含义是至少两个,例如两个,三个等,除非另有明确具体的限定。本申请实施例中所有方向性指示(诸如上、下、左、右、前、后……)仅用于解释在某一特定姿态(如附图所示)下各部件之间的相对位置关系、运动情况等,如果该特定姿态发生改变时,则该方向性指示也相应地随之改变。此外,术语“包括”和“具有”以及它们任何变形,意图在于覆盖不排他的包含。例如包含了一系列步骤或单元的过程、方法、系统、产品或设备没有限定于已列出的步骤或单元,而是可选地还包括没有列出的步骤或单元,或可选地还包括对于这些过程、方法、产品或设备固有的其它步骤或单元。
在本文中提及“实施例”意味着,结合实施例描述的特定特征、结构或特性可以包含在本申请的至少一个实施例中。在说明书中的各个位置出现该短语并不一定均是指相同的实施例,也不是与其它实施例互斥的独立的或备选的实施例。本领域技术人员显式地和隐式地理解的是,本文所描述的实施例可以与其它实施例相结合。
请参阅图1,是本申请第一实施例的基于人机对话的用户画像构建方法的流程示意图。本申请第一实施例的基于人机对话的用户画像构建方法包括以下步骤:
S10:对历史人机交互对话数据进行采样,得到人机对话样本数据;
本步骤中,人机对话样本数据即为用户与机器人或人工坐席之间的对话,人机对话数据样本包括语音数据或/和文本数据等。
S11:将人机对话样本数据转换为文本数据;
本步骤中,如果人机对话样本数据是语音数据,则通过ASR(Automatic Speech Recognition, 自动语音识别)技术将语音数据转换为文本格式;如果人机对话样本数据是通过文本交互产生的,则直接采用原始的文本数据。
S12:对文本数据中的用户属性信息按照(主语、属性类型、宾语)的三元组形式进行标注,生成训练模型的数据集;
本步骤中,用户属性信息即为用户在人机对话过程中对个人信息的描述,例如婚姻状况、子女情况、职业信息等。例如某一个句子为“我儿子今年7岁”,则该句子对应的用户属性信息标签为(儿子,年龄,7岁)。
S13:根据数据集训练属性分类模型,属性分类模型输出由主语、属性类型和宾语组成的三元组信息;
本步骤中,属性分类模型结构如图2所示,其包括对话编码器、属性分类器和实体生成器,首先对话编码器对文本数据进行编码,然后以对话编码器的输出作为属性分类器的输入,得到一个多标签分类相关的损失,随后每个激活的属性类型连同对话编码器的输出一起输入实体生成器,生成每个属性类型对应的主语和宾语,实体生成器部分也会对应一个序列相关的损失,最终模型的训练是在这两部分损失的驱动下进行的,构成一个多任务学习的框架。
本申请实施例采用有监督的方式进行模型训练,具体训练过程包括:
1、通过对话编码器对文本数据进行编码;其中,对话编码器采用GRU(Gated Recurrent Units,门控循环单元)对文本数据进行编码。具体的,对话编码器包括一个双向GRU和一个单向GRU,双向GRU接收的输入是单词嵌入向量(Word Embedding)组成的序列,用于对文本数据中的每一个句子分别进行编码;单向GRU接收的输入是编码后的句子嵌入向量(UtteranceEmbedding)组成的序列,用于对文本数据中每一个句子的上下文信息分别进行编码。
例如:假设一段文本数据由n个句子(Utterance)组成,表示成C={u 1,u 2,u 3,…,u n},第i个句子由K i个单词(Word)组成,表示成u i={w i,1,w i,2,w i,3,…,w i,Ki}。则对第i个句子的编码过程表示成下面的形式,得到单句级别的编码序列{h 1,h 2,h 3,…,h n}:
Figure PCTCN2020118445-appb-000001
Figure PCTCN2020118445-appb-000002
Figure PCTCN2020118445-appb-000003
上式中,e(w i,j)表示单词w i,j的嵌入向量。
得到单句级别的编码序列{h 1,h 2,h 3,…,h n}后,将其送入单向GRU进行对话级别的编码, 将GRU每一时刻的隐状态(Hidden state)构成新的句子嵌入表达序列{H 1,H 2,H 3,…,H n},编码过程表示成下面的形式:
Figure PCTCN2020118445-appb-000004
其中,由于GRU本身的序列编码特性,经过上述编码后,新的句子嵌入表达中包含每一个句子的上下文信息。例如,对于句子u i,其嵌入表达H i中融合了前文u 1到u i-1和u i的信息。
2、通过属性分类器对编码后的文本数据进行属性类型分类;其中,由于用户可能在一轮表达中产生多个类型的属性信息,例如性别、年龄或职业等,因此本申请将属性分类器设计为一个多标签分类器。具体地,属性分类器采用一个简单的全连接层,即线性映射层加上sigmoid激活函数层,如下所示:
P i=σ(WH i+b)   (5)
上式中,P i表示第i个句子在所有预先定义的属性类型下的概率分布,σ表示sigmoid激活函数,W和b是线性映射的权重参数,H i是单向GRU编码得到的第i个句子的嵌入表达向量。
3、将编码后的文本数据和属性分类器输出的属性类型一起输入实体生成器,通过实体生成器提取各个属性类型对应的主语和宾语;其中,实体生成器采用GRU进行主语和宾语的提取。将实体生成定义成一个序列生成问题,实体生成器的目标序列为由主语和宾语拼接成的序列,GRU起始时刻的输入为某一种属性类型的嵌入向量,起始时刻的隐状态(Hidden state)为对应句子由单向GRU进行编码的最后一个时刻得到的隐状态,以使得各个属性类型分别对应一种实体生成器,通过实体生成器输出各个属性类型对应的主语和宾语,然后将实体生成器的输出结果和对应的属性类型进行拼接,最终得到由主语、属性类型和宾语组成的三元组信息。
S14:根据三元组信息构建用户画像。
基于上述,本申请实施例的基于人机对话的用户画像构建方法通过构建包括对话编码器、属性分类器和实体生成器为一体的属性分类模型,能够在统一框架下自动提取人机对话中不同属性类型的显式或隐式用户属性信息,根据提取的属性信息构建用户画像,根据用户画像进行不同场景的营销任务推荐。本申请实施例可以部署到不同的场景任务中,提升了用户属性信息提取的灵活性和准确性。
请参阅图3,是本申请第二实施例的基于人机对话的用户画像构建方法的流程示意图。本申请第二实施例的基于人机对话的用户画像构建方法包括以下步骤:
S20:获取人机交互过程中的对话数据,将对话数据转换成文本数据;
S21:将文本数据输入训练得到的属性分类模型,通过属性分类模型提取文本数据中包含的用户属性信息,并输出由主语、属性类型、宾语组成的三元组信息;
S22:根据三元组信息构建用户画像。
在一个可选的实施方式中,还可以:将所述的基于人机对话的用户画像构建方法的结果上传至区块链中。
具体地,基于所述的基于人机对话的用户画像构建方法的结果得到对应的摘要信息,具体来说,摘要信息由所述的基于人机对话的用户画像构建方法的结果进行散列处理得到,比如利用sha256s算法处理得到。将摘要信息上传至区块链可保证其安全性和对用户的公正透明性。用户可以从区块链中下载得该摘要信息,以便查证所述的基于人机对话的用户画像构建方法的结果是否被篡改。本示例所指区块链是分布式数据存储、点对点传输、共识机制、加密算法等计算机技术的新型应用模式。区块链(Blockchain),本质上是一个去中心化的数据库,是一串使用密码学方法相关联产生的数据块,每一个数据块中包含了一批次网络交易的信息,用于验证其信息的有效性(防伪)和生成下一个区块。区块链可以包括区块链底层平台、平台产品服务层以及应用服务层等。
请参阅图4,是本申请实施例基于人机对话的用户画像构建系统的结构示意图。本申请实施例基于人机对话的用户画像构建系统40包括:
数据获取模块41:用于获取人机交互过程中的对话数据;
属性信息提取模块42:用于将所述对话数据输入属性分类模型,所述属性分类模型包括对话编码器、属性分类器和实体生成器,通过所述对话编码器对所述对话数据进行编码,通过所述属性分类器对所述编码后的对话数据进行属性类型分类,通过所述实体生成器提取各个属性类型对应的主语和宾语,所述属性分类模型将所述主语和宾语与对应的属性类型进行拼接,并输出由主语、属性类型和宾语组成的三元组信息;
用户画像构建模块43:用于根据所述三元组信息构建用户画像。
请参阅图5,为本申请实施例的终端结构示意图。该终端50包括处理器51、与处理器51耦接的存储器52。
存储器52存储有用于实现上述基于人机对话的用户画像构建方法的程序指令。
处理器51用于执行存储器52存储的程序指令以执行基于人机对话的用户画像构建操作。
其中,处理器51还可以称为CPU(Central Processing Unit,中央处理单元)。处理器51可能是一种集成电路芯片,具有信号的处理能力。处理器51还可以是通用处理器、数字信号处理器(DSP)、专用集成电路(ASIC)、现成可编程门阵列(FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
请参阅图6,图6为本申请实施例的存储介质的结构示意图。本申请实施例的存储介质存储有能够实现上述所有方法的程序文件61,所述存储介质可以是非易失性,也可以是易失 性。其中,该程序文件61可以以软件产品的形式存储在上述存储介质中,包括若干指令用以使得一台计算机设备(可以是个人计算机,服务器,或者网络设备等)或处理器(processor)执行本申请各个实施方式方法的全部或部分步骤。而前述的存储介质包括:U盘、移动硬盘、只读存储器(ROM,Read-Only Memory)、随机存取存储器(RAM,Random Access Memory)、磁碟或者光盘等各种可以存储程序代码的介质,或者是计算机、服务器、手机、平板等终端设备。
在本申请所提供的几个实施例中,应该理解到,所揭露的系统,装置和方法,可以通过其它的方式实现。例如,以上所描述的系统实施例仅仅是示意性的,例如,单元的划分,仅仅为一种逻辑功能划分,实际实现时可以有另外的划分方式,例如多个单元或组件可以结合或者可以集成到另一个系统,或一些特征可以忽略,或不执行。另一点,所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口,装置或单元的间接耦合或通信连接,可以是电性,机械或其它的形式。
另外,在本申请各个实施例中的各功能单元可以集成在一个处理单元中,也可以是各个单元单独物理存在,也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现,也可以采用软件功能单元的形式实现。以上仅为本申请的实施方式,并非因此限制本申请的专利范围,凡是利用本申请说明书及附图内容所作的等效结构或等效流程变换,或直接或间接运用在其他相关的技术领域,均同理包括在本申请的专利保护范围内。

Claims (20)

  1. 一种基于人机对话的用户画像构建方法,其中,包括:
    获取人机交互过程中的对话数据;
    将所述对话数据输入属性分类模型,所述属性分类模型包括对话编码器、属性分类器和实体生成器,通过所述对话编码器对所述对话数据进行编码,通过所述属性分类器对所述编码后的对话数据进行属性类型分类,通过所述实体生成器提取各个属性类型对应的主语和宾语,所述属性分类模型将所述主语和宾语与对应的属性类型进行拼接,并输出由主语、属性类型和宾语组成的三元组信息;
    根据所述三元组信息构建用户画像。
  2. 根据权利要求1所述的基于人机对话的用户画像构建方法,其中,所述对话数据包括语音数据或/和文本数据。
  3. 根据权利要求2所述的基于人机对话的用户画像构建方法,其中,所述获取人机交互过程中的对话数据还包括:
    通过ASR技术将所述语音数据转换为文本数据。
  4. 根据权利要求1所述的基于人机对话的用户画像构建方法,其中,所述通过对话编码器对对话数据进行编码包括:
    所述对话编码器包括一个双向GRU,所述双向GRU的输入为单词嵌入向量组成的序列,用于对所述文本数据中的每一个句子分别进行编码,得到单句级别的编码序列{h 1,h 2,h 3,…,h n}:
    Figure PCTCN2020118445-appb-100001
    Figure PCTCN2020118445-appb-100002
    Figure PCTCN2020118445-appb-100003
    上式中,e(w i,j)表示单词w i,j的嵌入向量。
  5. 根据权利要求4所述的基于人机对话的用户画像构建方法,其中,所述通过对话编码器对对话数据进行编码还包括:
    所述对话编码器还包括一个单向GRU,所述单向GRU的输入为所述单句级别的编码序列{h 1,h 2,h 3,…,h n},根据GRU本身的序列编码特性将GRU每一时刻的隐状态构成新的句子嵌入表达序列{H 1,H 2,H 3,…,H n},使得所述新的句子嵌入表达中包含每一个句子的上下文信息:
    Figure PCTCN2020118445-appb-100004
  6. 根据权利要求5所述的基于人机对话的用户画像构建方法,其中,所述通过属性分类器对编码后的对话数据进行属性类型分类包括:
    所述属性分类器采用线性映射层加sigmoid激活函数层进行属性类型分类:
    P i=σ(WH i+b)
    上式中,P i表示第i个句子在所有预先定义的属性类型下的概率分布,σ表示sigmoid激活函数,W和b是线性映射的权重参数,H i是所述单向GRU编码得到的第i个句子的嵌入表达向量。
  7. 根据权利要求6所述的基于人机对话的用户画像构建方法,其中,所述通过实体生成器提取各个属性类型对应的主语和宾语包括:
    所述实体生成器采用GRU进行主语和宾语的提取,将实体生成定义成一个序列生成问题,所述实体生成器的目标序列为由主语和宾语拼接成的序列,所述GRU起始时刻的输入为某一种属性类型的嵌入向量,起始时刻的隐状态为对应句子由所述单向GRU进行编码的最后一个时刻得到的隐状态,使得各个属性类型分别对应一种实体生成器,通过所述实体生成器输出各个属性类型对应的主语和宾语。
  8. 一种基于人机对话的用户画像构建系统,其中,包括:
    数据获取模块:用于获取人机交互过程中的对话数据;
    属性信息提取模块:用于将所述对话数据输入属性分类模型,所述属性分类模型包括对话编码器、属性分类器和实体生成器,通过所述对话编码器对所述对话数据进行编码,通过所述属性分类器对所述编码后的对话数据进行属性类型分类,通过所述实体生成器提取各个属性类型对应的主语和宾语,所述属性分类模型将所述主语和宾语与对应的属性类型进行拼接,并输出由主语、属性类型和宾语组成的三元组信息;
    用户画像构建模块:用于根据所述三元组信息构建用户画像。
  9. 一种终端,其中,所述终端包括处理器、与所述处理器耦接的存储器,其中,
    所述存储器存储有可被所述处理器执行的程序指令;
    所述处理器执行所述存储器存储的所述程序指令时实现以下步骤:
    获取人机交互过程中的对话数据;
    将所述对话数据输入属性分类模型,所述属性分类模型包括对话编码器、属性分类器和实体生成器,通过所述对话编码器对所述对话数据进行编码,通过所述属性分类器对所述编码后的对话数据进行属性类型分类,通过所述实体生成器提取各个属性类型对应的主语和宾语,所述属性分类模型将所述主语和宾语与对应的属性类型进行拼接,并输出由主语、属性类型和宾语组成的三元组信息;
    根据所述三元组信息构建用户画像。
  10. 根据权利要求9所述的终端,其中,所述对话数据包括语音数据或/和文本数据。
  11. 根据权利要求10所述的终端,其中,所述获取人机交互过程中的对话数据还包括:
    通过ASR技术将所述语音数据转换为文本数据。
  12. 根据权利要求9所述的终端,其中,所述通过对话编码器对对话数据进行编码包括:
    所述对话编码器包括一个双向GRU,所述双向GRU的输入为单词嵌入向量组成的序列,用于对所述文本数据中的每一个句子分别进行编码,得到单句级别的编码序列{h 1,h 2,h 3,…,h n}:
    Figure PCTCN2020118445-appb-100005
    Figure PCTCN2020118445-appb-100006
    Figure PCTCN2020118445-appb-100007
    上式中,e(w i,j)表示单词w i,j的嵌入向量。
  13. 根据权利要求12所述的终端,其中,所述通过对话编码器对对话数据进行编码还包括:
    所述对话编码器还包括一个单向GRU,所述单向GRU的输入为所述单句级别的编码序列{h 1,h 2,h 3,…,h n},根据GRU本身的序列编码特性将GRU每一时刻的隐状态构成新的句子嵌入表达序列{H 1,H 2,H 3,…,H n},使得所述新的句子嵌入表达中包含每一个句子的上下文信息:
    Figure PCTCN2020118445-appb-100008
  14. 根据权利要求13所述的终端,其中,所述通过属性分类器对编码后的对话数据进行属性类型分类包括:
    所述属性分类器采用线性映射层加sigmoid激活函数层进行属性类型分类:
    P i=σ(WH i+b)
    上式中,P i表示第i个句子在所有预先定义的属性类型下的概率分布,σ表示sigmoid激活函数,W和b是线性映射的权重参数,H i是所述单向GRU编码得到的第i个句子的嵌入表达向量。
  15. 根据权利要求14所述的终端,其中,所述通过实体生成器提取各个属性类型对应的主语和宾语包括:
    所述实体生成器采用GRU进行主语和宾语的提取,将实体生成定义成一个序列生成问题,所述实体生成器的目标序列为由主语和宾语拼接成的序列,所述GRU起始时刻的输入为某一种属性类型的嵌入向量,起始时刻的隐状态为对应句子由所述单向GRU进行编码的最后一个时刻得到的隐状态,使得各个属性类型分别对应一种实体生成器,通过所述实体生成器输出各个属性类型对应的主语和宾语。
  16. 一种存储介质,其中,存储有处理器可运行的程序指令,所述程序指令被处理器执行时实现以下步骤:
    获取人机交互过程中的对话数据;
    将所述对话数据输入属性分类模型,所述属性分类模型包括对话编码器、属性分类器和实体生成器,通过所述对话编码器对所述对话数据进行编码,通过所述属性分类器对所述编码后的对话数据进行属性类型分类,通过所述实体生成器提取各个属性类型对应的主语和宾语,所述属性分类模型将所述主语和宾语与对应的属性类型进行拼接,并输出由主语、属性类型和宾语组成的三元组信息;
    根据所述三元组信息构建用户画像。
  17. 根据权利要求16所述的存储介质,其中,所述通过对话编码器对对话数据进行编码包括:
    所述对话编码器包括一个双向GRU,所述双向GRU的输入为单词嵌入向量组成的序列,用于对所述文本数据中的每一个句子分别进行编码,得到单句级别的编码序列{h 1,h 2,h 3,…,h n}:
    Figure PCTCN2020118445-appb-100009
    Figure PCTCN2020118445-appb-100010
    Figure PCTCN2020118445-appb-100011
    上式中,e(w i,j)表示单词w i,j的嵌入向量。
  18. 根据权利要求17所述的存储介质,其中,所述通过对话编码器对对话数据进行编码还包括:
    所述对话编码器还包括一个单向GRU,所述单向GRU的输入为所述单句级别的编码序列{h 1,h 2,h 3,…,h n},根据GRU本身的序列编码特性将GRU每一时刻的隐状态构成新的句子嵌入表达序列{H 1,H 2,H 3,…,H n},使得所述新的句子嵌入表达中包含每一个句子的上下文信息:
    Figure PCTCN2020118445-appb-100012
  19. 根据权利要求18所述的存储介质,其中,所述通过属性分类器对编码后的对话数据进行属性类型分类包括:
    所述属性分类器采用线性映射层加sigmoid激活函数层进行属性类型分类:
    P i=σ(WH i+b)
    上式中,P i表示第i个句子在所有预先定义的属性类型下的概率分布,σ表示sigmoid激活函数,W和b是线性映射的权重参数,H i是所述单向GRU编码得到的第i个句子的嵌入表达向量。
  20. 根据权利要求19所述的存储介质,其中,所述通过实体生成器提取各个属性类型对应的主语和宾语包括:
    所述实体生成器采用GRU进行主语和宾语的提取,将实体生成定义成一个序列生成问题, 所述实体生成器的目标序列为由主语和宾语拼接成的序列,所述GRU起始时刻的输入为某一种属性类型的嵌入向量,起始时刻的隐状态为对应句子由所述单向GRU进行编码的最后一个时刻得到的隐状态,使得各个属性类型分别对应一种实体生成器,通过所述实体生成器输出各个属性类型对应的主语和宾语。
PCT/CN2020/118445 2020-08-06 2020-09-28 一种基于人机对话的用户画像构建方法、系统、终端及存储介质 WO2021120779A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010784250.4A CN111914076B (zh) 2020-08-06 2020-08-06 一种基于人机对话的用户画像构建方法、系统、终端及存储介质
CN202010784250.4 2020-08-06

Publications (1)

Publication Number Publication Date
WO2021120779A1 true WO2021120779A1 (zh) 2021-06-24

Family

ID=73287298

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/118445 WO2021120779A1 (zh) 2020-08-06 2020-09-28 一种基于人机对话的用户画像构建方法、系统、终端及存储介质

Country Status (2)

Country Link
CN (1) CN111914076B (zh)
WO (1) WO2021120779A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113761167A (zh) * 2021-09-09 2021-12-07 上海明略人工智能(集团)有限公司 一种会话信息抽取方法、系统、电子设备及存储介质
CN113901213A (zh) * 2021-10-08 2022-01-07 西安电子科技大学 一种面向任务型对话的实景信息分类方法及系统

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113268641B (zh) * 2020-12-14 2022-05-06 广东省广告集团股份有限公司 基于大数据的用户数据处理方法及大数据服务器
CN112287076B (zh) * 2020-12-25 2021-04-02 南京云问网络技术有限公司 一种基于用户聊天记录的标签挖掘方法及设备
CN113051384B (zh) * 2021-04-26 2023-09-19 中国平安人寿保险股份有限公司 基于对话的用户画像抽取方法及相关装置

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106599060A (zh) * 2016-11-16 2017-04-26 竹间智能科技(上海)有限公司 获取用户画像的方法及系统
CN107025278A (zh) * 2017-03-27 2017-08-08 竹间智能科技(上海)有限公司 基于人机对话的用户画像自动提取方法及装置
US20190236471A1 (en) * 2018-01-26 2019-08-01 International Business Machines Corporation Identifying Intent in Dialog Data Through Variant Assessment
CN110287290A (zh) * 2019-06-26 2019-09-27 平安科技(深圳)有限公司 基于阅读理解的营销线索提取方法、装置及计算机可读存储介质

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130246182A1 (en) * 2012-03-16 2013-09-19 Augme Technologies, Inc. Apparatuses, methods and systems for acquiring information from objects in a virtual space

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106599060A (zh) * 2016-11-16 2017-04-26 竹间智能科技(上海)有限公司 获取用户画像的方法及系统
CN107025278A (zh) * 2017-03-27 2017-08-08 竹间智能科技(上海)有限公司 基于人机对话的用户画像自动提取方法及装置
US20190236471A1 (en) * 2018-01-26 2019-08-01 International Business Machines Corporation Identifying Intent in Dialog Data Through Variant Assessment
CN110287290A (zh) * 2019-06-26 2019-09-27 平安科技(深圳)有限公司 基于阅读理解的营销线索提取方法、装置及计算机可读存储介质

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113761167A (zh) * 2021-09-09 2021-12-07 上海明略人工智能(集团)有限公司 一种会话信息抽取方法、系统、电子设备及存储介质
CN113761167B (zh) * 2021-09-09 2023-10-20 上海明略人工智能(集团)有限公司 一种会话信息抽取方法、系统、电子设备及存储介质
CN113901213A (zh) * 2021-10-08 2022-01-07 西安电子科技大学 一种面向任务型对话的实景信息分类方法及系统
CN113901213B (zh) * 2021-10-08 2024-04-02 西安电子科技大学 一种面向任务型对话的实景信息分类方法及系统

Also Published As

Publication number Publication date
CN111914076B (zh) 2023-07-21
CN111914076A (zh) 2020-11-10

Similar Documents

Publication Publication Date Title
WO2021120779A1 (zh) 一种基于人机对话的用户画像构建方法、系统、终端及存储介质
WO2020253060A1 (zh) 语音识别方法、模型的训练方法、装置、设备及存储介质
WO2022095378A1 (zh) 基于人工智能的培训方法、装置、计算机设备及存储介质
WO2021114840A1 (zh) 基于语义分析的评分方法、装置、终端设备及存储介质
WO2023241410A1 (zh) 数据处理方法、装置、设备及计算机介质
CN112732911A (zh) 基于语义识别的话术推荐方法、装置、设备及存储介质
WO2021174922A1 (zh) 语句情感分类方法及相关设备
WO2022252636A1 (zh) 基于人工智能的回答生成方法、装置、设备及存储介质
WO2021204017A1 (zh) 文本意图识别方法、装置以及相关设备
WO2023273628A1 (zh) 一种视频循环识别方法、装置、计算机设备及存储介质
WO2021135457A1 (zh) 基于循环神经网络的情绪识别方法、装置及存储介质
CN113887229A (zh) 地址信息的识别方法、装置、计算机设备及存储介质
CN112632244A (zh) 一种人机通话的优化方法、装置、计算机设备及存储介质
CN112418059A (zh) 一种情绪识别的方法、装置、计算机设备及存储介质
CN115438215A (zh) 图文双向搜索及匹配模型训练方法、装置、设备及介质
WO2023226239A1 (zh) 对象情绪的分析方法、装置和电子设备
CN111767694A (zh) 文本生成方法、装置和计算机可读存储介质
CN115687934A (zh) 意图识别方法、装置、计算机设备及存储介质
US11893990B2 (en) Audio file annotation
CN113903338A (zh) 面签方法、装置、电子设备和存储介质
CN111723649B (zh) 一种基于语义分解的短视频事件检测方法
US10706086B1 (en) Collaborative-filtering based user simulation for dialog systems
CN116680392A (zh) 一种关系三元组的抽取方法和装置
CN113515617B (zh) 一种对话生成模型的方法、装置以及设备
CN113420869B (zh) 基于全方向注意力的翻译方法及其相关设备

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: 20901190

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: 20901190

Country of ref document: EP

Kind code of ref document: A1