WO2020038251A1 - 流程定制系统、方法和计算机可读存储介质 - Google Patents

流程定制系统、方法和计算机可读存储介质 Download PDF

Info

Publication number
WO2020038251A1
WO2020038251A1 PCT/CN2019/100310 CN2019100310W WO2020038251A1 WO 2020038251 A1 WO2020038251 A1 WO 2020038251A1 CN 2019100310 W CN2019100310 W CN 2019100310W WO 2020038251 A1 WO2020038251 A1 WO 2020038251A1
Authority
WO
WIPO (PCT)
Prior art keywords
slot information
user
word slot
task
sentence
Prior art date
Application number
PCT/CN2019/100310
Other languages
English (en)
French (fr)
Inventor
刘云峰
徐易楠
吴悦
胡晓
汶林丁
Original Assignee
深圳追一科技有限公司
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 深圳追一科技有限公司 filed Critical 深圳追一科技有限公司
Publication of WO2020038251A1 publication Critical patent/WO2020038251A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/30—Semantic analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • G06Q10/06316—Sequencing of tasks or work

Definitions

  • the present application relates to the field of network technology, and in particular, to a process customization system and method, and a non-volatile computer-readable storage medium.
  • FAQ query and answer
  • task completion Inquiry and answer robot customer service has been relatively mature and can solve most user problems.
  • the application scenario of mission-type robot customer service is more extensive and complicated, involving multiple rounds of dialogue management, word slot collection and inheritance, and logical jumps.
  • An example of a mission robot customer service is as follows:
  • the mission-type robot customer service enables the robot customer service not only to answer questions, but also to solve user problems.
  • the task robot customer service is a robot with a specific task in a specific field.
  • the robot that handles the return of e-commerce cannot solve the task of booking a ticket.
  • To complete the task it needs to be re-developed.
  • open-field task-based robots it will face the problem of collecting too much word slot information, without process assistance and decreasing accuracy, affecting user experience.
  • the purpose of the mission-type robot customer service is to collect the required word slot information, so as to provide customers with corresponding information or solutions.
  • the limitations of the existing technology solutions lead to the need for new tasks when they occur Developers re-developed, making the speed and stability of new tasks on the line unsatisfactory.
  • the maintenance of robots has also risen linearly.
  • a process customization system, method, and non-volatile computer-readable storage medium are provided.
  • a process customization system including:
  • a rule setting module configured to obtain preset types of word slot information to be extracted and their matching rules
  • the task process building module is used to construct a task process, wherein each task process constructed includes the type of word slot information to be collected and the collection order, and online service information feedback actions;
  • An information extraction module configured to extract word slot information for a sentence input by a user according to the type of the word slot information and a matching rule thereof;
  • An intent recognition model for predicting a user's intention based on the sentence entered by the user and the extracted word slot information
  • a task matching module is configured to obtain, when the user intends to complete a task, an approximation of a sentence input by the user and each of the task processes, and match a prescribed number of task processes according to the approximation.
  • a process customization method including:
  • the approximation of the sentence input by the user and each of the task processes is obtained, and a predetermined number of task processes are matched according to the approximation.
  • One or more non-volatile computer-readable storage media containing computer-executable instructions when the computer-executable instructions are executed by one or more processors, cause the processors to perform the following operations:
  • the approximation of the sentence input by the user and each of the task processes is obtained, and a predetermined number of task processes are matched according to the approximation.
  • FIG. 1 is a structural block diagram of a process customization system provided by one or more embodiments.
  • FIG. 2 is a schematic diagram of a courier order task flow provided by one or more embodiments.
  • FIG. 3 is a schematic diagram of an aircraft order task flow provided by one or more embodiments.
  • FIG. 4 is an application environment diagram of a process customization method provided by one or more embodiments.
  • FIG. 5 is a flowchart of a process customization method in one or more embodiments.
  • FIG. 6 is a schematic diagram of a courier inquiry process provided by one or more embodiments.
  • FIG. 7 is a schematic diagram of an internal structure of an electronic device in an embodiment.
  • FIG. 1 is a structural block diagram of a process customization system provided in an embodiment. As shown in Figure 1, the process customization system provided includes:
  • the rule setting module 102 is configured to obtain preset types of word slot information to be extracted and their matching rules.
  • word slot information types mobile phone number, order number, time information, city name, mobile phone verification code. Further matching rules can be formulated for each type of information, such as an order number of eleven or nine consecutive digits, and a collection verification code of five or six consecutive digits. It should be noted that the user can select the common types of word slot information preset by the system, or customize the types and rules. Optionally, the system can obtain the extracted word slot information types preset by the user and their matching rules.
  • the task flow constructing module 104 is configured to construct a task flow chart, wherein each task flow constructed includes a type of word slot information to be collected and a collection order, and an online service information feedback action.
  • FIG. 2 is a schematic diagram of a courier order task flow in an embodiment.
  • the word slot information collection order is: collection order number-> collection mobile phone number-> collection mobile phone verification code-> feedback online query results.
  • FIG. 3 is a schematic diagram of an aircraft order task flow in one embodiment. The task flow is to collect the take-off position and the arrival position at the same time-> feedback the alternative flight information-> feedback the predetermined result.
  • the provided process customization system may further include a pre-processing module for pre-processing a sentence input by a user.
  • the pre-processing operation may include removing verbal words, such as “ ⁇ ”, “ ⁇ ”, punctuation marks, etc., to reduce the number of recognition words of the information extraction module. After preprocessing, it is preferable to perform manual or machine correction again to ensure that important information is not missed.
  • the information extraction module 106 is configured to extract word slot information from a sentence input by a user according to the type of the word slot information and its rules.
  • the word slot information collection can be used for regular matching using predefined rules.
  • the specific format is order number
  • the first 11 digits are the mobile phone number
  • the city name is the departure and arrival position, etc.
  • An intent recognition module 108 is configured to predict a user's intention based on the extracted word slot information and a sentence input by the user.
  • the extracted word slot information and the sentence input by the user may be input to the intent recognition model to obtain the user intention input by the intent recognition model.
  • Its model input is the features of user sentence and information extraction model.
  • the user intention here refers to whether the user wants to chat or ask a certain question or complete a task.
  • the features extracted from the information extraction model can effectively filter the intent. For example, if the order number and mobile phone number are extracted from the user sentence, the probability is that the user wants to query the express delivery. For user questions that are judged as gossip, the gossip robot will make a judgment response. If it is a task intent, a task-based robot will be used to perform the workflow described later in this article.
  • the provided process customization system further includes a status tracking module, which is used to transfer the status information of the current round to the next round after each round of dialogue is completed.
  • Status information can include word slot information, task scores for this round, and system-generated answers.
  • Each round of dialogue includes at least one of an information extraction process, an intent recognition process, and a sentence flow for obtaining user input.
  • the task matching module 110 is configured to obtain a degree of approximation between a sentence entered by the user and each task process after determining that the user intends to complete a task, and match a prescribed number of task processes according to the degree of approximation.
  • N can be set according to actual application requirements, for example, it can be 2, 3, 4, etc., which is not limited herein.
  • the workflow of the task matching module is configured as:
  • S2 Calculate the similarity between the sentence entered by the user represented by the word vector and the task similar question corresponding to each task flow.
  • the similarity is the distance between the sentence entered by the user and the task similar question. For example, the cos distance, the larger the cos distance, the higher the similarity.
  • the number of candidate word slots can be reduced by the screening task, which is convenient for screening the word slots corresponding to the task as soon as possible.
  • the process customization system provided by the embodiment of the present application can be developed to adapt to a variety of different tasks at one time.
  • the system can match different task processes according to the sentences entered by the user to complete the tasks, and make the entire task process transparent and easy to maintain. Reduced manpower investment, on the other hand, makes the system more stable and reliable.
  • Each module in the above-mentioned process customization system can be realized in whole or in part through software, hardware, and a combination thereof.
  • the above-mentioned modules may be embedded in the hardware form or independent of the processor in the computer device, or may be stored in the memory of the computer device in the form of software, so that the processor calls and performs the operations corresponding to the above modules.
  • module and “system” are intended to mean computer-related entities, which may be hardware, a combination of hardware and software, software, or software in execution.
  • a module may be, but is not limited to, a process running on a processor, a processor, an object, executable code, a thread of execution, a program, and / or a computer.
  • the application running on the server and the server can be modules.
  • One or more modules can reside within a process and / or thread of execution and a module can be localized on one computer and / or distributed between two or more computers.
  • the provided process customization method can be applied to an application environment as shown in FIG. 4.
  • the application environment includes an electronic device 402.
  • the electronic device 402 can obtain preset types of word slot information to be collected and extraction rules; construct a task flow; extract word slot information from a user-entered sentence according to the type of word slot information and its matching rules; The word slot information predicts the user's intention; when the user intends to complete the task, the approximation of the sentence entered by the user and each task process is obtained, and a specified number of task processes are matched according to the approximation.
  • the electronic device 402 is various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices.
  • the electronic device 402 may also be a server, and the server may be implemented by an independent server or a server cluster composed of multiple servers.
  • FIG. 5 is a flowchart of a process customization method in one or more embodiments. As shown in Figure 5, the method includes:
  • Step 502 Acquire preset types of word slot information to be collected and extraction rules.
  • Step 504 Construct a task flow.
  • Step 506 Extract the word slot information for the sentence input by the user according to the type of the word slot information and its matching rule.
  • Step 508 Predict the user's intention according to the sentence input by the user and the extracted word slot information.
  • Step 510 When the user intends to complete a task, obtain an approximation between the sentence entered by the user and the task process, and match a specified number of task processes according to the approximation.
  • the provided method for customizing a process after matching a prescribed number of task processes according to the approximation, further includes:
  • the operation of returning to extract the word slot information from the user-entered sentence according to the type of word slot information and its matching rules is the operation of transferring to the next round.
  • the provided method can also change the state of the round In the next round, the status of the round includes the collected word slot information, matching task processes, and generated answers, etc., which are not limited here.
  • a method for customizing a process is provided.
  • the task process shown in FIG. 5 is used as an example for description.
  • the specific operations for implementing the method include:
  • Operation 1 Set the type of word slot information to be collected and extraction rules in the system background in advance.
  • Operation two build the task process and store it.
  • the task process is set in advance according to business needs.
  • Operation three Get the statement entered by the user.
  • Operation four Use the information extraction model to extract information based on the type of word slot information and extraction rules to be collected by the user to obtain the extracted word slot information.
  • the sentence may be pre-processed, such as removing unnecessary mood particles and the like.
  • Operation five The extracted word slot information and the corresponding word slot information type and extraction rules are input as input parameters to an intent recognition model to obtain a predicted user intention.
  • Operation 6 The intent recognition module determines whether the user intends to complete the task or chat; if it is a chat, the robot is not transferred to the chat robot and no subsequent operations are performed, otherwise the subsequent operations are performed.
  • Operation 7 Match the corresponding task process according to the judged user's intention, score the tasks according to the degree of matching, and sort the task priorities according to the scores.
  • Operation eight The system judges whether all the word slot information required for the task with the highest priority is collected according to the word slot information that has been collected. If the word slot information is collected, the related services are accessed, otherwise the status tracking module will obtain the word slot information. , Matching tasks, and system-generated answers go to the next round of system interaction with customers.
  • Operation Nine The system generates corresponding answers to guide the user to complete the relevant word slot information input.
  • Operation ten Repeat operations three to ten until the task flow is completed.
  • Operation eleven access related services.
  • Operation 12 feedback the service results.
  • Operation two Establish the task flow of the courier order inquiry and store it.
  • Operation three Get the sentence entered by the user: "Where's my courier?".
  • Operation four The information extraction model extracts the "courier” in “where is the courier” according to the type of word slot information to be collected and the extraction rules.
  • the pre-processing module pre-processes the sentence, removing "I”, "", “Le” and question marks.
  • Operation five The features of the extracted "courier" and information extraction models are input as input parameters to the intent recognition model.
  • Operation 7 The intent recognition model predicts that the user's intent is to prepare to query the courier.
  • Operation eight Match the user's intention to the corresponding express order query task flow.
  • Operation 9 The system judges that the order number information and mobile phone number information have not been collected, so it generates an answer to prompt the user to enter the order number and mobile phone number.
  • Operation 10 Transfer the obtained word slot information, matched tasks, and system-generated answers to the next round of system interaction with customers.
  • Operation eleven The user enters a sentence containing an order number and a mobile phone number.
  • Operation 12 The information extraction model extracts the order number and mobile phone number according to the type of word slot information and extraction rules to be collected.
  • Operation 13 The features of the extracted order number, mobile phone number, and information extraction model are input to the intent recognition model as input parameters.
  • Operation 14 The intent recognition model predicts that the user's intention is to be ready to complete a task.
  • Operation 15 Match the corresponding courier order query task process.
  • Operation sixteen The system judges that the mobile phone verification code needs to be collected according to the collected order number and collection number.
  • Operation seventeen The user enters a collection verification code.
  • Operation 18 After the system verifies and collects the verification code, it accesses the online service of express logistics information query and feedbacks the results.
  • Any process or method description in a flowchart or otherwise described herein can be understood as representing a module, fragment, or portion of code that includes one or more executable instructions for implementing the operation of a particular logical function or process
  • the scope of the preferred embodiments of this application includes additional implementations in which the functions may be performed out of the order shown or discussed, including performing the functions in a substantially simultaneous manner or in the reverse order according to the functions involved, which should It is understood by those skilled in the art to which the embodiments of the present application pertain.
  • an electronic device is further provided.
  • the electronic device includes a memory and a processor.
  • the memory stores computer-readable instructions.
  • the processor causes the processor to perform the operations described in the foregoing embodiment. Provided process customization methods.
  • the electronic device may be a customer service robot.
  • FIG. 7 is a schematic diagram of an internal structure of an electronic device in an embodiment.
  • the electronic device includes a processor, a memory, a network interface, a display screen, and an input device connected through a system bus.
  • the processor of the electronic device is used to provide computing and control capabilities.
  • the memory of the electronic device includes a non-volatile storage medium and an internal memory.
  • the non-volatile storage medium stores an operating system and a computer program.
  • the internal memory provides an environment for running an operating system and computer programs in a non-volatile storage medium.
  • the network interface of the electronic device is used to communicate with an external terminal through a network connection.
  • the computer program is executed by a processor to implement a process customization method.
  • the display screen of the electronic device may be a liquid crystal display screen or an electronic ink display screen.
  • the input device of the electronic device may be a touch layer covered on the display screen, or a button, a trackball, or a touchpad provided on the electronic device casing. , Or an external keyboard, trackpad, or mouse.
  • the process customization system provided in this application may be implemented in the form of a computer program, and the computer program may be run on an electronic device as shown in FIG. 7.
  • the memory of the electronic device can store each program module constituting the process customization system.
  • the computer program constituted by each program module causes the processor to perform operations in the process customization method of each embodiment of the application described in this specification.
  • one or more non-volatile computer-readable storage media containing computer-executable instructions are provided, and when the computer-executable instructions are executed by one or more processors, the processors enable Operation of the provided process customization method.

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Theoretical Computer Science (AREA)
  • Strategic Management (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Economics (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Educational Administration (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Machine Translation (AREA)

Abstract

一种流程定制方法，包括：获取预先设定需要收集的词槽信息种类及提取规则；构建任务流程；根据所述词槽信息种类及其匹配规则对用户输入的语句提取词槽信息；根据所述用户输入的语句、提取的词槽信息预测用户意图；当所述用户意图为完成任务时，获取所述用户输入的语句与每个所述任务流程的近似度，根据所述近似度匹配规定数量的任务流程。

Description

流程定制系统、方法和计算机可读存储介质
相关申请的交叉引用
本申请要求于2018年08月24日提交中国专利局、申请号为2018109714518、发明名称为“一种可定制化流程的任务型客服机器人系统及其工作方法”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
本申请涉及网络技术领域，具体涉及一种流程定制系统、方法和非易失性计算机可读存储介质。
背景技术
客服是企业解答用户疑问，解决用户难题的重要渠道，而传统的电话客服以及在线客服都需要企业花费大量的人力物力去维护。如今，客服机器人的兴起使得人工客服的需求量下降，极大地节约了企业成本。
按机器人客服需要完成的目标划分，可分成两类，分别是查询解答类(FAQ)和完成任务类。查询解答型机器人客服已经相对成熟，能解决大部分用户难题。而任务型机器人客服应用场景则更加广泛和复杂，涉及多轮对话管理，词槽收集继承，逻辑跳转等多个难题。任务型机器人客服示例如下：
任务型机器人客服使机器人客服不仅具有答疑的能力，更具有解决用户问题的能力。在一般的解决方案中，任务机器人客服是特定领域特定任务的机器人，例如电商办理退货的机器人不能解决预定机票的任务，要完成该任 务需要重新开发。而对于开放领域的任务型机器人，则会面临需要收集词槽信息过多，没有流程辅助而准确率下降的问题，影响用户体验。
任务型机器人客服的目的是收集需要的词槽信息，从而为客户提供相应的资讯或者解决方案，而在保证准确率的前提下，现有技术方案的局限性导致当有新的任务出现时需要开发人员进行重新开发，使新任务上线速度和稳定性都不理想，另外，随着新任务越来越多，机器人维护难度也直线上升。
发明内容
根据本申请的各种实施例，提供一种流程定制系统、方法和非易失性计算机可读存储介质。
一种流程定制系统，包括：
规则设定模块，用于获取预先设定需要提取的词槽信息种类及其匹配规则；
任务流程构建模块，用于构建任务流程，其中，构建的每个任务流程包括需要收集的词槽信息种类及收集顺序、线上服务信息反馈动作；
信息提取模块，用于根据所述词槽信息种类及其匹配规则对用户输入的语句提取词槽信息；
意图识别模型，用于根据所述用户输入的语句、提取的词槽信息预测用户意图；及
任务匹配模块，用于当所述用户意图为完成任务时，获取所述用户输入的语句与每个所述任务流程的近似度，根据所述近似度匹配规定数量的任务流程。
一种流程定制方法，包括：
获取预先设定需要收集的词槽信息种类及提取规则；
构建任务流程；
根据所述词槽信息种类及其匹配规则对用户输入的语句提取词槽信息；
根据所述用户输入的语句、提取的词槽信息预测用户意图；及
当所述用户意图为完成任务时，获取所述用户输入的语句与每个所述任务流程的近似度，根据所述近似度匹配规定数量的任务流程。
一个或多个包含计算机可执行指令的非易失性计算机可读存储介质，当所述计算机可执行指令被一个或多个处理器执行时，使得所述处理器执行以下操作：
获取预先设定需要收集的词槽信息种类及提取规则；
构建任务流程；
根据所述词槽信息种类及其匹配规则对用户输入的语句提取词槽信息；
根据所述用户输入的语句、提取的词槽信息预测用户意图；及
当所述用户意图为完成任务时，获取所述用户输入的语句与每个所述任务流程的近似度，根据所述近似度匹配规定数量的任务流程。
本申请的一个或多个实施例的细节在下面的附图和描述中提出。本申请的其它特征、目的和优点将从说明书、附图以及权利要求书变得明显。
附图说明
为了更清楚地说明本申请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申请的一些实施例，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。
为了更好地描述和说明本申请公开的实施例和/或示例，可以参考一副或者多副附图。用于描述附图的附加细节或示例不应当被认为是对所公开的发明、目前描述的实施例和/或示例以及目前理解的这些发明的最佳模式中的任何一者的范围的限制。
图1为一个或多个实施例提供的流程定制系统的结构框图。
图2为一个或多个实施例提供的快递订单任务流程的示意图。
图3为一个或多个实施例提供的飞机订单任务流程的示意图。
图4为一个或多个实施例提供的流程定制方法的应用环境图。
图5为一个或多个实施例中流程定制方法的流程图。
图6为一个或多个实施例提供的快递查询流程的示意图。
图7为一个实施例中电子设备的内部结构示意图。
具体实施方式
为使本申请的目的、技术方案和优点更加清楚，下面将对本申请的技术方案进行详细的描述。显然，所描述的实施例仅仅是本申请一面分实施例，而不是全面的实施例。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的前提下所得到的所有其它实施方式，都属于本申请所保护的范围。
下面对本申请系统进行说明。图1为一个实施例中提供的流程定制系统的结构框图。如图1所示，提供的流程定制系统包括：
规则设定模块102，用于获取预先设定需要提取的词槽信息种类及其匹配规则。
词槽信息种类举例：手机号、订单号、时间信息、城市名称、手机验证 码。还可对各个信息种类进行进一步的匹配规则制定，如订单号为十一位或者九位连续数字、收集验证码为连续五位或者六位数字。需要说明的是，用户可以选择系统预设好的常见的词槽信息种类，也可以自定义种类及规则。可选地，系统可以获取用户预先设定的提取的词槽信息种类及其匹配规则。
任务流程构建模块104，用于,构建任务流程图，其中，构建的每个任务流程包括需要收集的词槽信息种类及收集顺序、线上服务信息反馈动作。
举例说明：如图2所示为一个实施例中快递订单任务流程的示意图，词槽信息收集顺序为：收集订单号->收集手机号->收集手机验证码->反馈线上查询结果。如图3所示为一个实施例中飞机订单任务流程的示意图，任务流程为：同时收集起飞位置和到达位置->反馈回可供选择的航班信息->反馈回预定结果。
在一个实施例中，提供的流程定制系统还可以包括预处理模块，用于对用户输入的语句进行预处理。预处理操作可以包括去掉虚词，如“了”、“的”、标点符号等等，以减少信息提取模块的识别字数。预处理完后，优选再次进行人工或者机器修正，以保证不遗漏重要信息。
信息提取模块106，用于根据词槽信息种类及其规则对用户输入的语句进行提取词槽信息。
词槽信息收集可以是使用预定义规则进行正则匹配，比如匹配特定格式为订单号，1开头的11位数字为手机号，城市名为出发到达位置等，也可以使用深度学习的序列标注模型标注出词槽位置，还可以结合以上两种方法。
意图识别模块108，用于根据提取的词槽信息、用户的输入的语句预测用户意图。
具体地，可以将提取的词槽信息和用户输入的语句输入至意图识别模型， 以得到意图识别模型输入的用户意图。其模型输入为用户语句及信息提取模型的特征。这里的用户意图指的是用户是想闲聊还是询问某个问题还是完成某个任务等。从信息提取模型提取出来的特征可以有效过滤意图，如在用户语句中抽取出订单号和手机号的话，大概率用户是想查询快递。对于判断为闲聊的用户问句，会交由闲聊机器人进行判断回复，如果是任务意图，则会利用任务型机器人进行本文后续描述的工作流程。
在一个实施例中，提供的流程定制系统还包括状态追踪模块，用于在每轮对话完成后将本轮的状态信息转到下一轮。状态信息可以包括词槽信息，本轮任务打分，系统生成的答案等。每轮对话包括一次信息提取流程、一次意图识别流程、一次获取用户输入的语句流程中的至少一种。
任务匹配模块110，用于在确定用户意图为完成任务后，获取用户输入的语句与每个任务流程的近似度，根据近似度匹配规定数量的任务流程。
可选地，根据用户问句与各个任务的匹配度进行任务打分并进行排序，取前N个匹配对应任务流程。其中，N可以根据实际应用需求设定，例如可以是2、3、4等，在此不做限定。
具体地，任务匹配模块的工作流程被配置为：
S1：将用户输入的语句进行分词，转化为对应的词向量表示。
S2：计算词向量表示的用户输入的语句与每个任务流程对应的任务相似问句的相似度，相似度为用户输入的语句与任务相似问句的距离。例如为cos距离，cos距离越大，相似度越高。
S3:对相似度进行排序。
S4:选择相似度排名在前的预设数量的任务流程。
通过筛选任务可以缩小候选词槽的数量，便于尽快筛选出任务对应的词 槽。
本申请实施例提供的流程定制系统，可以一次开发适应多种不同的任务，系统可以根据用户输入的语句匹配不同的任务流程进而完成任务，并且让整个任务流程变得透明和容易维护，一方面减少了人力投入，另一方面也使系统更加稳定可靠。
上述流程定制系统中的各个模块可全部或部分通过软件、硬件及其组合来实现。上述各模块可以硬件形式内嵌于或独立于计算机设备中的处理器中，也可以以软件形式存储于计算机设备中的存储器中，以便于处理器调用执行以上各个模块对应的操作。
如在本申请中所使用的，术语“模块”和“系统”等旨在表示计算机相关的实体，它可以是硬件、硬件和软件的组合、软件、或者执行中的软件。例如，模块可以是但不限于是，在处理器上运行的进程、处理器、对象、可执行码、执行的线程、程序和/或计算机。作为说明，运行在服务器上的应用程序和服务器都可以是模块。一个或多个模块可以驻留在进程和/或执行的线程中，并且模块可以位于一个计算机内和/或分布在两个或更多的计算机之间。
在一个实施例中，提供的流程定制方法，可以应用于如图4所示的应用环境中。该应用环境包括电子设备402。电子设备402可以获取预先设定需要收集的词槽信息种类及提取规则；构建任务流程；根据词槽信息种类及其匹配规则对用户输入的语句提取词槽信息；根据用户输入的语句、提取的词槽信息预测用户意图；当用户意图为完成任务时，获取用户输入的语句与每个任务流程的近似度，根据近似度匹配规定数量的任务流程。其中，电子设备402是各种个人计算机、笔记本电脑、智能手机、平板电脑和便携式可穿 戴设备等。可选地，电子设备402也可以是服务器，服务器可以用独立的服务器或者是多个服务器组成的服务器集群来实现。
图5为一个或多个实施例中流程定制方法的流程图。如图5所示，该方法包括：
步骤502，获取预先设定需要收集的词槽信息种类及提取规则。
步骤504，构建任务流程。
步骤506，根据词槽信息种类及其匹配规则对用户输入的语句提取词槽信息。
步骤508，根据用户输入的语句、提取的词槽信息预测用户意图。
步骤510，当用户意图为完成任务时，获取用户输入的语句与所述任务流程的近似度，根据近似度匹配规定数量的任务流程。
在一个实施例中，提供的流程定制方法中，根据近似度匹配规定数量的任务流程之后，还包括：
S1，按照近似度的高度分数高低进行任务流程的优先级排序。
S2，根据已经收集的词槽信息判断是否将目前优先级最高的任务流程所需的全部词槽信息收集完成。
S3，当收集未完成时，生成对应的答案，并返回根据词槽信息种类及其匹配规则对用户输入的语句提取词槽信息的操作，其中，答案用于引导用户输入对应的词槽信息。
其中，返回根据词槽信息种类及其匹配规则对用户输入的语句提取词槽信息的操作即为转入下一轮的操作，在一个实施例中，提供的方法还可以将该轮的状态转到下一轮，其中，该轮的状态包括收集的词槽信息、匹配的任务流程、生成的答案等，在此不做限定。
S4，当收集完成时，则接入与目前优先级最高的任务流程对应的服务，并进入下一优先级的人物流程。
在一个实施例中，提供了一种流程定制方法，以图5所示的任务流程为例进行说明，实现该方法的具体操作包括：
操作一：预先在系统后台设定好所需要收集的词槽信息种类及提取规则。
操作二：构建任务流程，并进行存储。其中，任务流程是预先根据业务需要设定的。
操作三：获取用户输入的语句。
操作四：通过信息提取模型根据所需要收集的词槽信息种类及提取规则对用户输入的语句进行信息提取，得到提取的词槽信息。
可选地，在提取词槽信息之前，可以对语句进行预处理，如去除不需的语气助词等。
操作五：将提取的词槽信息及对应的词槽信息种类及提取规则作为输入参数输入到意图识别模型，获得预测的用户意图。
操作六：意图识别模块判断出来用户意图是完成任务还是闲聊；如果是闲聊，则不转到闲聊机器人，不进行后续操作，否则进行后续操作。
操作七：根据判断的用户意图匹配对应的任务流程,根据匹配程度对任务进行打分,按照分数高低进行任务优先级排序。
当匹配到多个任务流程时，根据匹配到的多个任务流程的优先级进行收集。举例说明：用户输入“帮我订个飞机票，再开个发票”，这时会先进行“订机票”流程，再进行“开发票|”流程，因为订机票优先级要高于开发票，没有机票也没法开发票。
操作八：系统根据已经收集的词槽信息判断是否收集完了目前优先级最 高的任务所需的全部词槽信息，如果收集完了，则接入相关服务，否则状态追踪模块将所获得的词槽信息、所匹配任务、系统生成的答案转到下一轮的系统与客户互动状态。
操作九：系统生成对应的答案引导用户完成相关词槽信息输入。
操作十：重复操作三到操作十，直到完成任务流程。
需要说明的是，当用户继续输入时，首先会根据之前进入的任务流程进行收集，当用户提及其他任务意图时，也会收集新的任务意图对应的词槽。
操作十一：接入相关服务。
操作十二：反馈服务结果。
下面以图6所示为快递查询流程例进行说明。
操作一：预先在系统后台设定词槽信息，包括“快递”和/或“包裹”文字、订单号、手机号。设定订单号是9位连续数字；手机号是11为连续数字。
操作二：建立快递订单查询的任务流程，并进行存储。
操作三：获取用户输入的语句：“我的快递到哪了？”。
操作四：信息提取模型根据所需要收集的词槽信息种类及提取规则将“快递到哪”中的“快递”提取出来。
预处理模块对语句进行预处理，去掉“我”、“的”、“了”及问号。
操作五：被提取的“快递”及信息提取模型的特征作为输入参数输入到意图识别模型。
操作七：意图识别模型预测用户意图是准备查询快递。
操作八：将用户意图匹配对应的快递订单查询任务流程。
操作九：系统判断还未收集到订单号信息、手机号信息，因此生成答案提示用户输入订单号及手机号。
操作十：将所获得的词槽信息、所匹配任务、系统生成的答案转到下一轮的系统与客户互动状态。
操作十一：用户输入含有订单号、手机号的语句。
操作十二：信息提取模型根据所需要收集的词槽信息种类及提取规则将订单号、手机号提取出来。
对语句进行预处理，去掉“我”、“的”、“了”及问号。
操作十三：被提取的订单号、手机号及信息提取模型的特征作为输入参数输入到意图识别模型。
操作十四：意图识别模型预测用户意图是准备完成任务。
操作十五：匹配对应的快递订单查询任务流程。
操作十六：系统根据已经收集的订单号、收集号判断还需要收集手机验证码。
操作十七：用户输入收集验证码。
操作十八：系统验证收集验证码后，接入快递物流信息查询线上服务，并反馈结果。
可以理解的是，上述各实施例中相同或相似部分可以相互参考，在一些实施例中未详细说明的内容可以参见其他实施例中相同或相似的内容。
流程图中或在此以其他方式描述的任何过程或方法描述可以被理解为，表示包括一个或更多个用于实现特定逻辑功能或过程的操作的可执行指令的代码的模块、片段或部分，并且本申请的优选实施方式的范围包括另外的实现，其中可以不按所示出或讨论的顺序，包括根据所涉及的功能按基本同时的方式或按相反的顺序，来执行功能，这应被本申请的实施例所属技术领域的技术人员所理解。
在一个实施例中，还提供一种电子设备，该电子设备包括存储器和处理器，存储器中存储有计算机可读指令，该计算机可读指令被处理器执行时，使得处理器执行上述实施例所提供的流程定制方法。
可选地，电子设备可以是客服机器人。
图7为一个实施例中电子设备的内部结构示意图。如图7所示，该电子设备包括通过系统总线连接的处理器、存储器、网络接口、显示屏和输入装置。其中，该电子设备的处理器用于提供计算和控制能力。该电子设备的存储器包括非易失性存储介质、内存储器。该非易失性存储介质存储有操作系统和计算机程序。该内存储器为非易失性存储介质中的操作系统和计算机程序的运行提供环境。该电子设备的网络接口用于与外部的终端通过网络连接通信。该计算机程序被处理器执行时以实现一种流程定制方法。该电子设备的显示屏可以是液晶显示屏或者电子墨水显示屏，该电子设备的输入装置可以是显示屏上覆盖的触摸层，也可以是电子设备外壳上设置的按键、轨迹球或触控板，还可以是外接的键盘、触控板或鼠标等。
在一个实施例中，本申请提供的流程定制系统可以实现为一种计算机程序的形式，计算机程序可在如图7所示的电子设备上运行。电子设备的存储器中可存储组成该流程定制系统的各个程序模块。各个程序模块构成的计算机程序使得处理器执行本说明书中描述的本申请各个实施例的流程定制方法中的操作。
在一个实施例中，提供了一个或多个包含计算机可执行指令的非易失性计算机可读存储介质，当计算机可执行指令被一个或多个处理器执行时，使得处理器上述实施例所提供的流程定制方法的操作。
以上的，仅为本申请的具体实施方式，但本申请的保护范围并不局限于 此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应以的权利要求的保护范围为准。

Claims (15)

  1. 一种流程定制系统，包括：
    规则设定模块，用于获取预先设定需要提取的词槽信息种类及其匹配规则；
    任务流程构建模块，用于构建任务流程，其中，构建的每个任务流程包括需要收集的词槽信息种类及收集顺序、线上服务信息反馈动作；
    信息提取模块，用于根据所述词槽信息种类及其匹配规则对用户输入的语句提取词槽信息；
    意图识别模块，用于根据所述用户输入的语句、提取的词槽信息预测用户意图；及
    任务匹配模块，用于当所述用户意图为完成任务时，获取所述用户输入的语句与每个所述任务流程的近似度，根据所述近似度匹配规定数量的任务流程。
  2. 根据权利要求1所述的系统，其特征在于，还包括预处理模块，用于对用户输入的语句进行预处理。
  3. 根据权利要求1所述的系统，其特征在于，所述系统还包括：
    状态追踪模块，用于在每轮对话完成后将该轮的状态转到下一轮，其中，所述该轮的状态包括该轮的词槽信息、该轮任务打分，该轮系统生成的答案，所述每轮对话为一次词槽信息提取流程、一次意图识别流程、一次获取用户输入的语句流程中的至少一种。
  4. 根据权利要求1所述的系统，其特征在于，任务匹配模块被配置为：
    将所述用户输入的语句进行分词，转化为对应的词向量表示；
    计算所述词向量表示的用户输入的语句与每个所述任务流程对应的任务 相似问句的相似度，所述相似度为用户输入的语句与任务相似问句的距离；
    对所述相似度进行排序；及
    选择所述相似度排名在前的预设数量的任务流程。
  5. 根据权利要求4所述的系统，其特征在于，所述用户输入的语句与任务相似问句的距离为COS距离。
  6. 一种流程定制方法，包括：
    获取预先设定需要收集的词槽信息种类及提取规则；
    构建任务流程；
    根据所述词槽信息种类及其匹配规则对用户输入的语句提取词槽信息；
    根据所述用户输入的语句、提取的词槽信息预测用户意图；及
    当所述用户意图为完成任务时，获取所述用户输入的语句与每个所述任务流程的近似度，根据所述近似度匹配规定数量的任务流程。
  7. 根据权利要求6所述的方法，其特征在于，所述根据所述近似度匹配规定数量的任务流程之后，还包括：
    按照所述近似度的高度分数高低进行任务流程的优先级排序；
    根据已经收集的词槽信息判断是否将所述目前优先级最高的任务流程所需的全部词槽信息收集完成；
    当收集未完成时，生成对应的答案，并返回所述根据所述词槽信息种类及其匹配规则对用户输入的语句提取词槽信息的操作，其中，所述答案用于引导用户输入对应的词槽信息；及
    当收集完成时，则接入与所述目前优先级最高的任务流程对应的服务，并进入下一优先级的任务流程。
  8. 根据权利要求7所述的方法，其特征在于，还包括：
    当收集未完成时，则将该轮的状态转到下一轮，其中，该轮的状态包括收集的词槽信息、匹配的任务流程、生成的答案。
  9. 根据权利要求6所述的方法，其特征在于，所述根据所述词槽信息种类及其匹配规则对用户输入的语句提取词槽信息之前，还包括：
    对所述用户输入的语句进行预处理。
  10. 根据权利要求6所述的方法，其特征在于，所述根据所述词槽信息种类及其匹配规则对用户输入的语句提取词槽信息，包括：
    利用正则表达式根据所述词槽信息及其匹配规则对所述用户输入的语句进行匹配，得到所述词槽信息。
  11. 一个或多个包含计算机可执行指令的非易失性计算机可读存储介质，当所述计算机可执行指令被一个或多个处理器执行时，使得所述处理器执行以下操作：
    获取预先设定需要收集的词槽信息种类及提取规则；
    构建任务流程；
    根据所述词槽信息种类及其匹配规则对用户输入的语句提取词槽信息；
    根据所述用户输入的语句、提取的词槽信息预测用户意图；及
    当所述用户意图为完成任务时，获取所述用户输入的语句与每个所述任务流程的近似度，根据所述近似度匹配规定数量的任务流程。
  12. 根据权利要求11所述的非易失性计算机可读存储介质，其特征在于，所述处理器执行所述根据所述近似度匹配规定数量的任务流程之后，还执行如下操作：
    按照所述近似度的高度分数高低进行任务流程的优先级排序；
    根据已经收集的词槽信息判断是否将所述目前优先级最高的任务流程所 需的全部词槽信息收集完成；
    当收集未完成时，生成对应的答案，并返回所述根据所述词槽信息种类及其匹配规则对用户输入的语句提取词槽信息的操作，其中，所述答案用于引导用户输入对应的词槽信息；及
    当收集完成时，则接入与所述目前优先级最高的任务流程对应的服务，并进入下一优先级的人物流程。
  13. 根据权利要求12所述的非易失性计算机可读存储介质，其特征在于，所述处理器还执行：
    当收集未完成时，则将该轮的状态转到下一轮，其中，该轮的状态包括收集的词槽信息、匹配的任务流程、生成的答案。
  14. 根据权利要求10所述的非易失性计算机可读存储介质，其特征在于，所述处理器执行所述根据所述词槽信息种类及其匹配规则对用户输入的语句提取词槽信息之前，还执行如下操作：
    对所述用户输入的语句进行预处理。
  15. 根据权利要求10所述的非易失性计算机可读存储介质，其特征在于，所述处理器执行所述根据所述词槽信息种类及其匹配规则对用户输入的语句提取词槽信息时，还执行如下操作：
    利用正则表达式根据所述词槽信息及其匹配规则对所述用户输入的语句进行匹配，得到所述词槽信息。
PCT/CN2019/100310 2018-08-24 2019-08-13 流程定制系统、方法和计算机可读存储介质 WO2020038251A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810971451.8A CN109145100A (zh) 2018-08-24 2018-08-24 一种可定制化流程的任务型客服机器人系统及其工作方法
CN201810971451.8 2018-08-24

Publications (1)

Publication Number Publication Date
WO2020038251A1 true WO2020038251A1 (zh) 2020-02-27

Family

ID=64827783

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/100310 WO2020038251A1 (zh) 2018-08-24 2019-08-13 流程定制系统、方法和计算机可读存储介质

Country Status (2)

Country Link
CN (1) CN109145100A (zh)
WO (1) WO2020038251A1 (zh)

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109145100A (zh) * 2018-08-24 2019-01-04 深圳追科技有限公司 一种可定制化流程的任务型客服机器人系统及其工作方法
CN109829729A (zh) * 2019-01-18 2019-05-31 青牛智胜(深圳)科技有限公司 一种智能外呼系统及方法
CN109857487A (zh) * 2019-02-02 2019-06-07 上海奔影网络科技有限公司 用于任务执行的数据处理方法及装置
CN110110338A (zh) * 2019-05-13 2019-08-09 哈尔滨理工大学 一种基于lstm与槽填充的对话管理模型使用方法
CN110297850B (zh) * 2019-05-28 2021-10-15 北京奇艺世纪科技有限公司 特征作业集发布方法、装置、电子设备及存储介质
CN110516035A (zh) * 2019-07-05 2019-11-29 同济大学 一种混合模块的人机交互方法和系统
CN110472030A (zh) * 2019-08-08 2019-11-19 网易（杭州）网络有限公司 人机交互方法、装置和电子设备
CN111510566B (zh) * 2020-03-16 2021-05-28 深圳追一科技有限公司 通话标签的确定方法、装置、计算机设备和存储介质
CN114493164B (zh) * 2021-12-30 2024-04-09 重庆特斯联智慧科技股份有限公司 基于边缘计算的机器人任务解析方法和系统
CN114888804B (zh) * 2022-05-18 2024-03-12 深圳鹏行智能研究有限公司 基于工作链的机器人控制装置及方法、介质、机器人

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104731895A (zh) * 2015-03-18 2015-06-24 北京京东尚科信息技术有限公司 自动应答的方法和装置
US20160275444A1 (en) * 2012-03-28 2016-09-22 Global Eprocure Procurement System
CN107590216A (zh) * 2017-08-31 2018-01-16 北京百度网讯科技有限公司 答案获得方法、装置和计算机设备
CN107958059A (zh) * 2017-12-01 2018-04-24 北京百度网讯科技有限公司 智能问答方法、装置、终端和计算机可读存储介质
CN109145100A (zh) * 2018-08-24 2019-01-04 深圳追科技有限公司 一种可定制化流程的任务型客服机器人系统及其工作方法

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9830044B2 (en) * 2013-12-31 2017-11-28 Next It Corporation Virtual assistant team customization
CN107015964B (zh) * 2017-03-22 2021-10-19 北京光年无限科技有限公司 面向智能机器人开发的自定义意图实现方法及装置
CN107133052B (zh) * 2017-05-27 2021-01-05 杭州迪脉信息科技有限公司 流程创建的方法及装置
CN108197167A (zh) * 2017-12-18 2018-06-22 深圳前海微众银行股份有限公司 人机对话处理方法、设备及可读存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160275444A1 (en) * 2012-03-28 2016-09-22 Global Eprocure Procurement System
CN104731895A (zh) * 2015-03-18 2015-06-24 北京京东尚科信息技术有限公司 自动应答的方法和装置
CN107590216A (zh) * 2017-08-31 2018-01-16 北京百度网讯科技有限公司 答案获得方法、装置和计算机设备
CN107958059A (zh) * 2017-12-01 2018-04-24 北京百度网讯科技有限公司 智能问答方法、装置、终端和计算机可读存储介质
CN109145100A (zh) * 2018-08-24 2019-01-04 深圳追科技有限公司 一种可定制化流程的任务型客服机器人系统及其工作方法

Also Published As

Publication number Publication date
CN109145100A (zh) 2019-01-04

Similar Documents

Publication Publication Date Title
WO2020038251A1 (zh) 流程定制系统、方法和计算机可读存储介质
US10725827B2 (en) Artificial intelligence based virtual automated assistance
CA3085315C (en) Systems and methods for chatbot generation
US11093854B2 (en) Emoji recommendation method and device thereof
US11693894B2 (en) Conversation oriented machine-user interaction
US20210224306A1 (en) System, Apparatus and Methods for Providing an Intent Suggestion to a User in a Text-Based Conversational Experience with User Feedback
US11538468B2 (en) Using semantic frames for intent classification
JP2022547631A (ja) 自然言語処理のためのストップワードデータ拡張
US11694032B2 (en) Template-based intent classification for chatbots
WO2020108063A1 (zh) 特征词的确定方法、装置和服务器
CN115398437A (zh) 改进的域外(ood)检测技术
US20190251417A1 (en) Artificial Intelligence System for Inferring Grounded Intent
US11928430B2 (en) Detecting unrelated utterances in a chatbot system
WO2020019878A1 (zh) 任务型机器人词槽填充方法、计算机设备和存储介质
CN116724305A (zh) 上下文标签与命名实体识别模型的集成
CN112417090A (zh) 利用未提交的用户输入数据来提高任务性能
CN115398436A (zh) 用于自然语言处理的噪声数据扩充
US20220237376A1 (en) Method, apparatus, electronic device and storage medium for text classification
CN116802629A (zh) 用于自然语言处理的多因素建模
CN116635862A (zh) 用于自然语言处理的域外数据扩充
CN116615727A (zh) 用于自然语言处理的关键词数据扩充工具
CN114663155A (zh) 广告投放选品方法及其装置、设备、介质、产品
WO2023029354A1 (zh) 文本信息提取方法、装置、存储介质及计算机设备
CN113177154A (zh) 搜索词推荐方法、装置、电子设备及存储介质
CN117314139A (zh) 业务流程的建模方法、装置、终端设备及存储介质

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: 19851002

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 24.06.2021

122 Ep: pct application non-entry in european phase

Ref document number: 19851002

Country of ref document: EP

Kind code of ref document: A1