WO2018006727A1 - 机器人客服转人工客服的方法和装置 - Google Patents

机器人客服转人工客服的方法和装置 Download PDF

Info

Publication number
WO2018006727A1
WO2018006727A1 PCT/CN2017/090322 CN2017090322W WO2018006727A1 WO 2018006727 A1 WO2018006727 A1 WO 2018006727A1 CN 2017090322 W CN2017090322 W CN 2017090322W WO 2018006727 A1 WO2018006727 A1 WO 2018006727A1
Authority
WO
WIPO (PCT)
Prior art keywords
customer service
user
confidence
feature
sample
Prior art date
Application number
PCT/CN2017/090322
Other languages
English (en)
French (fr)
Inventor
陈利霞
杜敏
Original Assignee
阿里巴巴集团控股有限公司
陈利霞
杜敏
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 阿里巴巴集团控股有限公司, 陈利霞, 杜敏 filed Critical 阿里巴巴集团控股有限公司
Publication of WO2018006727A1 publication Critical patent/WO2018006727A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor

Definitions

  • the present application relates to the field of data processing technologies, and in particular, to a method and apparatus for a robot service to a manual customer service.
  • the most commonly used architecture of the customer service center is the co-existence of the robot customer service and the manual customer service.
  • the robot customer service first receives the user, and the robot customer service and the user's conversation round (one user speaking, or one robot customer service speech and one user speaking one as one).
  • the round exceeds the predetermined artificial round threshold
  • the manual customer service is transferred.
  • the robot customer service is actively interacting with the user for multiple rounds, although the predetermined artificial round threshold is reached, but the predetermined artificial round threshold is reached.
  • There is no service barrier that requires manual intervention. This method will cause unnecessary extra work for the manual customer service.
  • the user has an urgent problem and needs to submit the manual processing as soon as possible, it will be due to the predetermined artificial round threshold. Invalid interaction affects the efficiency and user experience of the customer service system.
  • the present application provides a method for a robot service to switch to a manual customer service, including:
  • the evaluation value is used to train the robot customer service and the user's conversation sample marked with effective features and suitable for the artificial point, and the suitable artificial point is an appropriate time point for replacing the robot customer service with the manual customer service;
  • the application also provides a device for changing the robot customer service to the manual customer service, including:
  • An effective feature obtaining unit configured to obtain a valid feature from at least one round of conversation between the robot customer service and the user
  • the current confidence evaluation unit is configured to input the valid feature into the confidence assessment model to obtain a current confidence assessment value of the robot customer service and the user session; the confidence assessment model adopts a session sample of the robot customer service and the user marked with the effective feature and the suitable artificial point Performing training, the appropriate artificial point is an appropriate time point for replacing the robot customer service with manual customer service;
  • the manual customer service switching unit is configured to transfer the user to the manual customer service when the current confidence evaluation value meets the predetermined manual condition.
  • the confidence analysis model of the session sample training of the robot customer service and the user marked with the effective feature and the suitable artificial point is adopted, and the ongoing conversation between the robot client and the user is input to the confidence assessment.
  • the model according to the current confidence evaluation value of the model output, it is judged whether it is suitable for transferring the manual customer service, so that the user's demand for the manual customer service can be identified according to the actual situation of the current session, and the user's satisfaction with the service is improved. It reduces the unnecessary work of the manual customer service and improves the service efficiency of the customer service system.
  • FIG. 1 is a flowchart of a method for a robot service to a manual customer service in the embodiment of the present application
  • FIG. 2 is a hardware structural diagram of a device running an embodiment of the present application
  • FIG. 3 is a logic structural diagram of an apparatus for a robot service to a manual customer service in the embodiment of the present application.
  • the embodiment of the present application proposes a new method of robot customer service to manual customer service, in the robot
  • the customer service and the user's conversation sample manually mark the effective features and fit the artificial point, train the confidence assessment model, and use the confidence assessment model after training and the effective features of the current session to evaluate the current robot service quality to discover
  • the appropriate time point for transferring the manual customer service to determine the artificial point according to the actual situation of the current session can not only improve the user experience, but also avoid unnecessary workload of the manual customer service, thereby solving the problems in the prior art. .
  • the embodiment of the present application can be applied to any device having computing and storage capabilities, such as a physical device or a logical device such as a mobile phone, a tablet computer, a PC (Personal Computer), a notebook, a server, a virtual machine, or the like;
  • a physical device or a logical device such as a mobile phone, a tablet computer, a PC (Personal Computer), a notebook, a server, a virtual machine, or the like;
  • the functions in the embodiments of the present application are implemented by two or more physical or logical devices that share different responsibilities.
  • the flow of the method of the robot customer service to the manual customer service is as shown in FIG. 1 .
  • the machine learning technology is used to establish a confidence assessment model for evaluating whether the robot customer service and the user session should be transferred to a manual customer service at a certain point in time. Specifically, a certain number of robot customer service and user's session records are taken as session samples, and the effective features and the appropriate artificial points in each session are manually marked on the session samples, and the confidence evaluation model is trained by using the session samples.
  • the effective feature is an abstract expression of various factors related to the transfer artificial demand appearing in the session sample.
  • Effective features can be generated and evaluated using a combination of manual or manual and data mining techniques.
  • the technician can summarize and refine according to his work experience, historical data of the robot customer service and the user session, and the solution of the user problem.
  • the artificially summarized and refined features capable of describing the service quality of the robot customer service and the user's intention to transfer the human will be taken as the undetermined feature, and the undetermined feature, the suitable artificial point and the actual artificial point are marked in the On the session sample, combined with the service effect of each session in the session sample, the predetermined data analysis algorithm is used to evaluate and integrate the features, and several effective features are obtained that have obvious influence on the transfer labor demand and comprehensive coverage factors.
  • the historical data of the robot client and the user session can be manually analyzed, and the factors affecting the user's demand for the human service are mined; the factors are refined, classified, and abstracted into pending features, and the pending features are from the session.
  • each session in the session sample is obtained.
  • the service effect is characterized by feature selection and feature extraction algorithms for the analysis of the session samples and service effects that are marked with pending features, suitable for artificial points and actual artificial points, and the effective features are obtained.
  • a variety of data analysis algorithms that can be used to perform feature classification and evaluation can be used to generate effective features, which are not limited in the embodiments of the present application.
  • the data analysis algorithm employed can not only derive effective features based on session samples and service effects, but also give weights to the impact of each valid feature on the transfer of artificial requirements; in these application scenarios, each The weights of the valid features will be used in the training of the confidence assessment model, and/or in the use of the post-training confidence assessment model.
  • the manual customer service In the process of the robot customer service and the user's conversation, the manual customer service usually replaces the robot customer service at the time after the user speaks. In each session of the session sample, if the time point after a user speaks is the appropriate time point for the manual customer service to replace the robot service, it can be marked as suitable for the artificial point. There can be one or more suitable artificial points in a session.
  • the confidence assessment model is used to obtain the confidence assessment value based on the robot customer service and the user's conversation.
  • the confidence assessment value is an estimate of the confidence assessment model's degree of transfer labor demand, or it should be replaced by manual customer service at a certain point in the conversation. A score for the degree of robot customer service.
  • the specific algorithm used in the confidence assessment model in the embodiment of the present application is not limited. For example, it may be an SVR (Support Vector Regression) algorithm, an LR (Logistic Regression) algorithm, or a GBDT (Gradient Boosting Decision). Tree, iterative decision tree) algorithm, etc.
  • the confidence assessment model can be used in the real-time session between the robot customer service and the user to determine whether the user needs to switch to the manual customer service after the user speaks in the session.
  • Step 110 Obtain a valid feature from at least one round of conversation between the robot service and the user.
  • one user speaks, or one robot customer service speaks and one user speaks one round.
  • the first round of the conversation between the robot customer service and the user is the user speaking.
  • the second round and subsequent rounds are a robot customer service statement and a user speech.
  • Each round ends with a user's speech, which is the point at which the robotic customer service can be replaced by manual customer service.
  • the valid features obtained in this step will be used as input to the confidence assessment model to obtain confidence estimates for the current time point.
  • the destination session number of the predetermined number of times before the current time point may be used as the basis for obtaining the effective feature. If the session round that has been performed at the current time point is less than the predetermined number of rounds, the entire session is used as the basis for obtaining the effective feature; The entire session is the basis for obtaining valid features; no limitation.
  • the effective feature can be extracted according to a certain rule directly from one round to multiple rounds of the robot customer service and the user.
  • the identification rules of each effective feature can be set by referring to existing techniques such as semantic analysis, keyword matching, and business content matching, thereby automatically obtaining effective features from the above session; for voice conversation, voice recognition technology can be used first. Convert it to a text session and use the above method to get the valid features.
  • the original feature may be extracted from one round to multiple rounds of the robot customer service and the user, and then the original feature is split, combined, classified, and/or deleted according to the feature preprocessing rule.
  • Effective feature for a text session, the discovery rules of each original feature can be set by referring to existing techniques such as semantic analysis, keyword matching, and business content matching, thereby automatically extracting original features from one round to multiple rounds of conversation; for voice conversations, Speech recognition technology can be first used to convert a speech conversation into a text conversation, and the original feature is extracted using the discovery rules of the original feature.
  • the feature pre-processing rules describe the mapping from the original feature to the valid feature, including one or more of the following: which or which original feature can be deleted, which original feature can be split into which valid features, which or which are original Features can be attributed to valid features representing a category, which original features can be combined into which effective features, and the like.
  • the feature preprocessing rule may include such a case: if a reply can extract whether the original feature of the answer and the original feature of the answer match, Then combine the two original features into the original feature of the answer; in order to avoid the simultaneous use of these two effective features, the double impact of the same fact on the confidence evaluation value.
  • the confidence assessment model is based on effective features, the use of the post-training confidence assessment model is also input with valid features. Since the training confidence assessment model requires a certain number of manually labeled session samples, the actual client service session situation may change due to the increase of business, changes in business processes, and changes in popular languages. In this implementation, generalized and abstracted features can be used as effective features, and original features, original feature discovery rules, and feature pre-processing rules are set according to the specific conditions of business development and change, so that no actual business is needed. Changes in the situation continue to regenerate session samples and training confidence assessment models, while still maintaining confidence in the accuracy of the model.
  • the original feature discovery rules and/or feature pre-processing rules may be coded in the program that completes this step, or may be written in a configuration file.
  • the original feature may be selected in the undetermined feature, and other features that do not exist in the pending feature may be used as the original feature, and are not limited.
  • step 120 the valid feature is input into the confidence assessment model to obtain a current confidence assessment value of the robot customer service and the user session.
  • step 130 when the current confidence assessment value meets the predetermined manual condition, the user is transferred to the manual customer service.
  • the valid features are input into the confidence evaluation model after the training, and the confidence evaluation value of the current time point is obtained. If the current confidence assessment value meets the predetermined manual condition, it is considered that the current time point requires manual customer service intervention to transfer the user to the manual customer service.
  • the predetermined manual condition may be that the current confidence assessment value is greater than or less than a predetermined confidence threshold, depending on whether the current confidence assessment value in the actual application scenario is higher or not, or is the weaker transition labor demand degree.
  • the predetermined confidence threshold can be determined by the technician comprehensively considering factors such as the degree of fitting of the confidence assessment model and the session sample after training, the proportion of user sessions and the number of manual customer service in the actual application scenario. It can also be set by a certain standard, according to the set standard by the program. The predetermined confidence threshold is automatically determined.
  • the session sample may be input into the trained confidence evaluation model to obtain a sample confidence evaluation value corresponding to the appropriate artificial point in the session sample; a specific value of a series of different predetermined confidence thresholds is set, and the calculation is performed when different values are selected.
  • the coverage ratio is the ratio of the sample confidence evaluation values satisfying the predetermined artificial conditions (specific values greater than or less than the predetermined confidence threshold) among all the sample confidence evaluation values corresponding to the artificial points in the session sample;
  • the accuracy rate is Among all the sample confidence evaluation values that satisfy the predetermined artificial conditions, the proportion of the sample confidence evaluation values corresponding to the suitable artificial points.
  • the predetermined manual condition may also be written into the configuration file, and the predetermined artificial condition read from the configuration file is applied to step 130.
  • the effective feature and the artificial point are manually marked in the session sample of the robot customer service and the user, and the confidence sample is trained by using the labeled session sample, and the confidence evaluation model after the training is used.
  • the effective features in the session are evaluated. According to the output of the model, it is judged whether the manual customer service should be transferred. Therefore, the manual point can be determined according to the actual progress of the current session, which can improve the user experience and avoid manual customer service. Unnecessary workload and improve the service efficiency of the customer service system.
  • a history of a plurality of robot customer service and a user's session is taken as a session sample, and a professional customer service staff analyzes the session sample based on different dimensions (including a service dimension, a user experience dimension, a service track dimension, etc.). Analyze the session sample, summarize and refine the factors that affect the transfer of labor demand, and construct it as a pending feature.
  • Marking pending features, appropriate artificial points, and actual artificial points in the session sample using the session samples to construct data for the pending features of the structure according to the user feedback of each session in the session sample, the user's problem solving situation, and the user's satisfaction situation. Analysis, from which the effective features are evaluated and the weights of the valid features are derived.
  • the effective features and suitable artificial points are marked in the session sample, and the confidence evaluation model of the LR algorithm is trained by using the labeled session samples. Enter the session in the session sample into the trained LR The confidence evaluation model obtains the sample confidence evaluation value corresponding to the suitable artificial point in the session sample, and the technician determines the predetermined information threshold according to the coverage rate and accuracy rate of the session sample according to the LR confidence evaluation model.
  • the technician writes the original feature discovery rules and feature pre-processing rules in the configuration file and saves them at the predetermined location.
  • This application example loads the original feature discovery rules and feature preprocessing rules from the configuration file after starting the run.
  • the valid features belonging to the current time point are input into the trained LR confidence evaluation model to obtain the current confidence evaluation value.
  • the LR confidence assessment model is used to obtain a higher evaluation value, indicating that the transfer manual demand is stronger.
  • the current confidence evaluation value is greater than the predetermined information threshold, the user is transferred to the manual customer service, otherwise the robot customer service continues to talk to the user. .
  • the embodiment of the present application further provides a device for changing the robot customer service to the manual customer service.
  • the device may be implemented by software, or may be implemented by hardware or a combination of hardware and software.
  • the CPU Central Process Unit
  • the device in which the device of the robot service to the manual customer service usually includes other hardware such as a chip for transmitting and receiving wireless signals, and / Other hardware such as boards used to implement network communication functions.
  • FIG. 3 is a schematic diagram of a device for changing a human service of a robot to a manual customer service according to an embodiment of the present application, including an effective feature acquisition unit, a current confidence assessment unit, and a manual customer service transfer unit, wherein: the effective feature acquisition unit is used to Obtaining valid features in at least one round of the user's session; the current confidence assessment unit is configured to input the valid features into the confidence assessment model to obtain a current confidence assessment value of the robot customer service and the user session; the confidence assessment model is marked with effective features and suitable People
  • the robot service of the worksite is trained with the user's session sample, and the suitable manual point is an appropriate time point for replacing the robot customer service with the manual customer service; the manual customer service switching unit is used when the current confidence evaluation value satisfies the predetermined manual condition. Transfer the user to the manual customer service.
  • the effective feature obtaining unit is specifically configured to: extract original features from at least one round of conversation between the robot customer service and the user, and split, combine, classify, and/or delete the original features according to the feature preprocessing rules. Get valid features.
  • the device further includes a configuration file acquiring unit, configured to acquire a configuration file, where the configuration file includes an original feature discovery rule and/or a feature pre-processing rule; the effective feature obtaining unit is specifically configured to: In at least one round of conversation between the customer service and the user, the original feature is extracted according to the original feature discovery rule.
  • a configuration file acquiring unit configured to acquire a configuration file, where the configuration file includes an original feature discovery rule and/or a feature pre-processing rule; the effective feature obtaining unit is specifically configured to: In at least one round of conversation between the customer service and the user, the original feature is extracted according to the original feature discovery rule.
  • the configuration file further includes: scheduling a manual condition.
  • the original features include: business relevance, answer matching degree, answer repetition number, whether the answer is bottom, whether the answer is a question, the user explicitly proposes to change the labor, and the user has a potential for artificial change, the user's emotional tendency, The user is explaining at least one of his or her own problems.
  • the effective feature is evaluated by a predetermined data analysis algorithm based on a sample of the robot and the user that is marked with a pending feature, a suitable artificial point, and an actual artificial point, and a service effect of the session. And the integration is obtained; the pending feature can describe the service quality of the robot customer service and the user's intention to transfer.
  • the effective features have respective weights, which are calculated by the predetermined data analysis algorithm; the confidence assessment model performs training according to the weight of the effective features.
  • the predetermined manual condition includes: the current confidence assessment value is greater than or less than a predetermined confidence threshold; the predetermined confidence threshold is determined according to coverage and accuracy of the plurality of sample confidence assessment values; and the sample confidence evaluation value is The output obtained by inputting the session sample into the confidence evaluation model; the coverage ratio is the ratio of the sample confidence evaluation values satisfying the predetermined artificial condition among all the sample confidence evaluation values corresponding to the artificial points in the session sample; The accuracy rate is the proportion of the sample confidence evaluation value that satisfies the predetermined artificial condition, and the proportion of the sample confidence evaluation value corresponding to the suitable artificial point.
  • the confidence assessment model adopts a support vector regression SVR algorithm, a logistic regression LR algorithm, or an iterative decision tree GBDT algorithm.
  • a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
  • processors CPUs
  • input/output interfaces network interfaces
  • memory volatile and non-volatile memory
  • the memory may include non-persistent memory, random access memory (RAM), and/or non-volatile memory in a computer readable medium, such as read only memory (ROM) or flash memory.
  • RAM random access memory
  • ROM read only memory
  • Memory is an example of a computer readable medium.
  • Computer readable media includes both permanent and non-persistent, removable and non-removable media.
  • Information storage can be implemented by any method or technology.
  • the information can be computer readable instructions, data structures, modules of programs, or other data.
  • Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory. (ROM), electrically erasable programmable read only memory (EEPROM), flash memory or other memory technology, compact disk read only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, Magnetic tape cartridges, magnetic tape storage or other magnetic storage devices or any other non-transportable media can be used to store information that can be accessed by a computing device.
  • computer readable media does not include temporary storage of computer readable media, such as modulated data signals and carrier waves.
  • embodiments of the present application can be provided as a method, system, or computer program. Order product.
  • the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment in combination of software and hardware.
  • the application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) including computer usable program code.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种机器人客服转人工客服的方法，包括：从机器人客服与用户的至少一轮会话中获取有效特征（110）；将有效特征输入信心评估模型，得到机器人客服与用户会话的当前信心评估值（120）；所述信心评估模型采用标记有有效特征和适宜出人工点的机器人客服与用户的会话样本进行训练，所述适宜出人工点为以人工客服代替机器人客服的适当时点；在当前信心评估值满足预定出人工条件时，将用户转接人工客服（130）。该方法能够根据当前会话的实际情况来识别用户转人工客服的需求程度，在提高用户对服务的满意程度的同时，减少了人工客服的不必要工作，提高了客服系统的服务效率。

Description

机器人客服转人工客服的方法和装置 技术领域
本申请涉及数据处理技术领域，尤其涉及一种机器人客服转人工客服的方法和装置。
背景技术
随着互联网的发展，基于人工智能技术的虚拟机器人在企业用户服务领域的应用越来越广泛。机器人客服不需要休息，可以更加快速和标准化的响应用户的问题，以语音对话或文字聊天的形式与用户进行沟通，将人工客服从大量重复性问答中解放出来。
对一些非常规的用户问题，机器人客服往往难以给出令用户满意的答复。目前客服中心最为常用的架构是机器人客服与人工客服并存，缺省由机器人客服先接待用户，当机器人客服与用户的会话轮次(以一次用户发言、或者一次机器人客服发言和一次用户发言为一个轮次)超过预定的出人工轮次阈值时，转接人工客服。这种方式在大多数情况下捕捉不到合适的由机器人客服到人工客服的转接点，例如机器人客服正在与用户进行多轮次的有效交互，虽然达到了预定的出人工轮次阈值，但并没有遇到需要人工介入的服务障碍，这种方式会造成人工客服不必要的额外工作；而用户有紧急的问题需要尽快提交人工处理时，又会因达到预定的出人工轮次阈值前的无效交互影响客服系统的效率和用户体验。
发明内容
有鉴于此，本申请提供一种机器人客服转人工客服的方法，包括：
从机器人客服与用户的至少一轮会话中获取有效特征；
将有效特征输入信心评估模型，得到机器人客服与用户会话的当前信心 评估值；所述信心评估模型采用标记有有效特征和适宜出人工点的机器人客服与用户的会话样本进行训练，所述适宜出人工点为以人工客服代替机器人客服的适当时点；
在当前信心评估值满足预定出人工条件时，将用户转接人工客服。
本申请还提供了一种机器人客服转人工客服的装置，包括：
有效特征获取单元，用于从机器人客服与用户的至少一轮会话中获取有效特征；
当前信心评估单元，用于将有效特征输入信心评估模型，得到机器人客服与用户会话的当前信心评估值；所述信心评估模型采用标记有有效特征和适宜出人工点的机器人客服与用户的会话样本进行训练，所述适宜出人工点为以人工客服代替机器人客服的适当时点；
人工客服转接单元，用于在当前信心评估值满足预定出人工条件时，将用户转接人工客服。
由以上技术方案可见，本申请的实施例中，采用标记有有效特征和适宜出人工点的机器人客服与用户的会话样本训练信心评估模型，将正在进行的机器人客户与用户的会话输入到信心评估模型后，根据模型输出的当前信心评估值来判断当前是否适合转接人工客服，从而能够根据当前会话的实际情况来识别用户转人工客服的需求程度，在提高用户对服务的满意程度的同时，减少了人工客服的不必要工作，提高了客服系统的服务效率。
附图说明
图1是本申请实施例中一种机器人客服转人工客服的方法的流程图；
图2是运行本申请实施例的设备的一种硬件结构图；
图3是本申请实施例中一种机器人客服转人工客服的装置的逻辑结构图。
具体实施方式
本申请的实施例提出一种新的机器人客服转人工客服的方法，在机器人 客服与用户的会话样本中人工标注有效特征和适宜出人工点，对信心评估模型进行训练，并利用训练后的信心评估模型和当前会话的有效特征，对当前的机器人服务质量进行评估，来发现转接人工客服的合适时点，以便根据当前会话的实际进行情况来确定出人工点，既能提升用户的体验，又能避免人工客服不必要的工作负荷，从而解决现有技术中存在的问题。
本申请的实施例可以应用在任何具有计算和存储能力的设备上，例如可以是手机、平板电脑、PC(Personal Computer，个人电脑)、笔记本、服务器、虚拟机等物理设备或逻辑设备；也可以由两个或两个以上分担不同职责的物理或逻辑设备、相互协同来实现本申请实施例中的各项功能。
本申请的实施例中，机器人客服转人工客服的方法的流程如图1所示。
本申请的实施例中，采用机器学习技术来建立信心评估模型，用来对机器人客服与用户会话过程中的某个时点是否应转接人工客服进行评估。具体而言，将一定数量的机器人客服与用户的会话记录作为会话样本，在会话样本上人工标注有效特征和各个会话过程中的适宜出人工点，利用会话样本对信心评估模型进行训练。
其中，有效特征是对会话样本中出现的与转接人工需求相关的各种因素的抽象表达。有效特征可以采用人工、或人工和数据挖掘技术相结合的方式来生成、评估。一种实现方式中，可以由技术人员根据自身的工作经验、机器人客服与用户会话的历史数据、用户问题解决情况等方面来总结、提炼得出。
在另一种实现方式中，可以先将人工总结、提炼的能够描述机器人客服的服务质量和用户的转人工意愿的特征作为待定特征，将待定特征、适宜出人工点和实际出人工点标注在会话样本上，结合会话样本中每个会话的服务效果，采用预定的数据分析算法对待定特征进行评估和整合，得到对转接人工需求影响明显、覆盖因素全面的若干个有效特征。例如，可以人工分析机器人客户与用户会话的历史数据，从中挖掘影响用户对人工服务的需求的因素；对这些因素进行提炼、归类，抽象化为待定特征，待定特征从会话上下 文、业务、用户体验、服务轨迹等各个不同方面完备的描述了影响转接人工需求的上述因素；按照用户反馈、用户的问题解决情况以及用户的满意度情况等来得出会话样本中各个会话的服务效果，采用特征选择、特征抽取算法对标记有待定特征、适宜出人工点和实际出人工点的会话样本和服务效果进行分析，得出有效特征。
各种能够用来进行特征分类和评估的数据分析算法都可以用于生成有效特征，本申请的实施例中不做限定。在一些应用场景中，所采用的数据分析算法不仅能够基于会话样本和服务效果得出有效特征，还能够给出每个有效特征对转接人工需求的影响的权重；在这些应用场景中，每个有效特征的权重将用于信心评估模型的训练，和/或训练后信心评估模型的使用中。
在机器人客服和用户的会话过程中，人工客服通常在用户发言之后的时点代替机器人客服。在会话样本的各个会话中，如果某个用户发言之后的时点是以人工客服代替机器人客服的适当时点，则可以将其标记为适宜出人工点。一个会话中可以有一个到多个适宜出人工点。
信心评估模型用来以机器人客服和用户的会话为基础获得信心评估值，信心评估值是信心评估模型对转接人工需求程度的估计值，或者说，对会话中某时点应该由人工客服代替机器人客服的程度打出的一个分值。本申请实施例中对信心评估模型所采用的具体算法不做限定，例如，可以是SVR(Support Vector Regression，支持向量回归)算法、LR(Logistic Regression，逻辑回归)算法、或GBDT(Gradient Boosting Decision Tree，迭代决策树)算法等。
在信心评估模型训练完成后，可以在实时的机器人客服与用户的会话过程中，利用信心评估模型来判断在会话中用户发言后的时点，是否需要切换为人工客服。
步骤110，从机器人客服与用户的至少一轮会话中获取有效特征。
本申请的实施例中，以一次用户发言、或者一次机器人客服发言和一次用户发言为一个轮次。通常机器人客服与用户的会话的第一轮是用户发言， 第二轮及后续轮次是一次机器人客服发言和一次用户发言。每个轮次以用户发言来结束，该时点也即是可以由人工客服代替机器人客服的时点。
本步骤中获取的有效特征将用来作为信心评估模型的输入，用来得到当前时点的信心评估值。可以将当前时点前预定轮次数目的会话作为获取有效特征的基础，如果当前时点已经进行的会话轮次小于预定轮次数目，则整个会话作为获取有效特征的基础；也可以始终将已经进行的整个会话作为获取有效特征的基础；不做限定。
在一种实现方式中，可以直接从机器人客服与用户的一轮到多轮会话中，按照一定的规则提取出有效特征。例如，对文本会话，可以参照语义分析、关键词匹配、业务内容匹配等现有技术设置各个有效特征的识别规则，从而自动从上述会话中得到有效特征；对语音会话，可以先采用语音识别技术将其转换为文本会话，在利用上述方法来得到有效特征。
在另一种实现方式中，可以先从机器人客服与用户的一轮到多轮会话中提取原始特征，再按照特征预处理规则对原始特征进行拆分、组合、分类、和/或删除后得到有效特征。类似的，对文本会话，可以参考语义分析、关键词匹配、业务内容匹配等现有技术设置各个原始特征的发现规则，从而自动从一轮到多轮会话中提取出原始特征；对语音会话，可以先采用语音识别技术将语音会话转换为文本会话，在利用原始特征的发现规则来提取出原始特征。
特征预处理规则描述了从原始特征到有效特征的映射关系，包括以下的一种到多种：哪个或哪些原始特征可以删除、哪个原始特征可以拆分为哪几个有效特征、哪个或哪些原始特征可以归属为代表某个类别的有效特征、哪些原始特征可以组合为哪个有效特征等等。
例如，在一种应用场景中，在业务和用户反馈两个维度上的原始特征如表1所示：
Figure PCTCN2017090322-appb-000001
Figure PCTCN2017090322-appb-000002
表1
在表1中的原始特征中，如果机器人客服的一个回复是兜底答案，则该回复的答案匹配度必然不高(如果机器人客服能够查询到匹配度较高的答案，会以该答案而不是兜底答案来回复用户)，假设答案匹配度和是否兜底答案都是有效特征，则可以在特征预处理规则中可以包括这样一条：如果一个回复可提取出是否兜底答案原始特征和答案匹配度原始特征，则将这两个原始特征合并为是否兜底答案原始特征；以免同时采用这两个有效特征导致同一事实对信心评估值的双重影响。
由于信心评估模型基于有效特征进行训练，因此对训练后信心评估模型的使用也以有效特征为输入。由于训练信心评估模型需要一定数量人工标注的会话样本，而实际客户服务中的会话情况可能因业务的增加、业务流程的变更、流行语言的变化而不断变动。在这种实现方式中，可以采用概括性和抽象化的特征作为有效特征，而根据业务发展和变化的具体情况来设置原始特征、原始特征发现规则和特征预处理规则，这样不需因为实际业务情况的变化不断的重新生成会话样本和训练信心评估模型，而仍然可以保持信心评估模型的准确程度。
原始特征发现规则和/或特征预处理规则可以以代码的方式固化在完成本步骤的程序中，也可以写在配置文件中。对存在配置文件的应用场景，可以在执行本步骤前先获取配置文件，从中读取原始特征发现规则和/或特征预处理规则，再将其应用于原始特征提取和/或从原始特征得到有效特征。
需要说明的是，可以在待定特征中选择原始特征，也可以将在待定特征不存在的其他特征作为原始特征，不做限定。
步骤120，将有效特征输入信心评估模型，得到机器人客服与用户会话的当前信心评估值。
步骤130，在当前信心评估值满足预定出人工条件时，将用户转接人工客服。
在获取到机器人客服与用户会话中对应于当前时点的有效特征后，将有效特征输入到训练后的信心评估模型中，即可得到当前时点的信心评估值。如果当前信心评估值满足预定出人工条件，则认为当前时点需要人工客服介入，将用户转接至人工客服。
预定出人工条件可以是当前信心评估值大于或小于预定信心阈值，视实际应用场景中当前信心评估值更高是代表更强的转接人工需求程度，还是更弱的转接人工需求程度。预定信心阈值可以由技术人员综合考虑训练后信心评估模型与会话样本的拟合程度、实际应用场景中的用户会话和人工客服的数量比例等因素来确定。也可以由设置一定的标准，由程序根据所设置的标 准自动确定预定信心阈值。例如，可以将会话样本输入到训练后的信心评估模型中，得到会话样本中对应于适宜出人工点的样本信心评估值；设定一系列不同的预定信心阈值的具体数值，计算当选择不同数值的预定信心阈值时样本信心评估值的覆盖率和准确率，设定针对覆盖率和准确率的评判标准，按照评判标准的评价最好的覆盖率和准确率对应的数值作为预定信心阈值。其中，覆盖率是会话样本中适宜出人工点对应的所有样本信心评估值中，满足预定出人工条件(大于或小于预定信心阈值的具体数值)的样本信心评估值所占的比例；准确率是所有满足预定出人工条件的样本信心评估值中，对应于适宜出人工点的样本信心评估值所占的比例。
另外，在采用配置文件的应用场景中，还可以将预定出人工条件也写入配置文件中，并将从配置文件中读取的预定出人工条件应用于步骤130。
可见，本申请的实施例中，在机器人客服与用户的会话样本中人工标注有效特征和适宜出人工点，采用标记后的会话样本训练信心评估模型，并利用训练后的信心评估模型，对当前的会话中的有效特征进行评估，根据模型的输出来判断当前是否应转接人工客服，从而能够根据当前会话的实际进行情况来确定出人工点，既能提升用户的体验，又能避免人工客服不必要的工作负荷，提高客服系统的服务效率。
在本申请的一个应用示例中，将若干机器人客服与用户的会话的历史记录作为会话样本，由专业客服人员分析会话样本，基于不同的维度(包括业务维度、用户体验维度、服务轨迹维度等)分析会话样本，对影响转接人工需求的因素进行总结、提炼，构造为待定特征。
在会话样本中标记待定特征、适宜出人工点和实际出人工点，按照会话样本中各个会话的用户反馈、用户的问题解决情况及用户的满意度情况，利用会话样本对构造的待定特征进行数据分析，从中评估出有效特征并得出有效特征的权重。
在会话样本中标记有效特征和适宜出人工点，利用标记后的会话样本对LR算法的信心评估模型进行训练。将会话样本中的会话输入到训练后的LR 信心评估模型，得到会话样本中对应于适宜出人工点的样本信心评估值，由技术人员根据LR信心评估模型对会话样本的覆盖率和准确率，确定预定信息阈值。
技术人员在配置文件中写入原始特征发现规则和特征预处理规则，并保存在预定位置。本应用示例在开始运行后，从配置文件中加载原始特征发现规则和特征预处理规则。
当用户启动与机器人客服的会话后，在每个用户发言结束时，按照原始特征发现规则，在从会话开始到当前时点的已进行会话中提取所有原始特征，再应用特征预处理规则，将所有原始特征映射为属于当前时点的一个到多个有效特征。
将属于当前时点的有效特征输入训练后的LR信心评估模型，得到当前信心评估值。设LR信心评估模型以得出的评估值较高表示转接人工需求更强，则在当前信心评估值大于预定信息阈值时，将用户转接到人工客服，否则由机器人客服继续与用户进行对话。
与上述流程实现对应，本申请的实施例还提供了一种机器人客服转人工客服的装置，该装置可以通过软件实现，也可以通过硬件或者软硬件结合的方式实现。以软件实现为例，作为逻辑意义上的装置，是通过所在设备的CPU(Central Process Unit，中央处理器)将对应的计算机程序指令读取到内存中运行形成的。从硬件层面而言，除了图2所示的CPU、内存以及非易失性存储器之外，机器人客服转人工客服的装置所在的设备通常还包括用于进行无线信号收发的芯片等其他硬件，和/或用于实现网络通信功能的板卡等其他硬件。
图3所示为本申请实施例提供的一种机器人客服转人工客服的装置，包括有效特征获取单元、当前信心评估单元和人工客服转接单元，其中：有效特征获取单元用于从机器人客服与用户的至少一轮会话中获取有效特征；当前信心评估单元用于将有效特征输入信心评估模型，得到机器人客服与用户会话的当前信心评估值；所述信心评估模型采用标记有有效特征和适宜出人 工点的机器人客服与用户的会话样本进行训练，所述适宜出人工点为以人工客服代替机器人客服的适当时点；人工客服转接单元用于在当前信心评估值满足预定出人工条件时，将用户转接人工客服。
一个例子中，所述有效特征获取单元具体用于：从机器人客服与用户的至少一轮会话中提取原始特征，按照特征预处理规则对原始特征进行拆分、组合、分类、和/或删除后得到有效特征。
上述例子中，所述装置还包括配置文件获取单元，用于获取配置文件；所述配置文件中包括原始特征发现规则和/或特征预处理规则；所述有效特征获取单元具体用于：从机器人客服与用户的至少一轮会话中，按照原始特征发现规则提取原始特征。
可选的，所述配置文件中还包括：预定出人工条件。
上述例子中，所述原始特征包括：业务相关性、答案匹配度、答案重复次数、是否兜底答案、答案是否为提问、用户明确提出换人工、和用户有潜在换人工倾向、用户的情感倾向、用户在解释自己的问题中的至少一个。
一种实现方式中，所述有效特征由预定数据分析算法根据标记有待定特征、适宜出人工点和实际出人工点的机器人与用户的会话样本以及其中会话的服务效果，对若干待定特征进行评估及整合后得出；所述待定特征能够描述机器人客服的服务质量和用户的转人工意愿。
上述实现方式中，所述有效特征具有各自的权重，由所述预定数据分析算法计算得出；所述信心评估模型根据有效特征的权重进行训练。
可选的，所述预定出人工条件包括：当前信心评估值大于或小于预定信心阈值；所述预定信心阈值根据若干个样本信心评估值的覆盖率和准确率确定；所述样本信心评估值为将会话样本输入信心评估模型后得到的输出；所述覆盖率为会话样本中适宜出人工点对应的所有样本信心评估值中，满足预定出人工条件的样本信心评估值所占的比例；所述准确率为所有满足预定出人工条件的样本信心评估值中，对应于适宜出人工点的样本信心评估值所占的比例。
可选的，所述信心评估模型采用支持向量回归SVR算法、逻辑回归LR算法、或迭代决策树GBDT算法。
以上所述仅为本申请的较佳实施例而已，并不用以限制本申请，凡在本申请的精神和原则之内，所做的任何修改、等同替换、改进等，均应包含在本申请保护的范围之内。
在一个典型的配置中，计算设备包括一个或多个处理器(CPU)、输入/输出接口、网络接口和内存。
内存可能包括计算机可读介质中的非永久性存储器，随机存取存储器(RAM)和/或非易失性内存等形式，如只读存储器(ROM)或闪存(flash RAM)。内存是计算机可读介质的示例。
计算机可读介质包括永久性和非永久性、可移动和非可移动媒体可以由任何方法或技术来实现信息存储。信息可以是计算机可读指令、数据结构、程序的模块或其他数据。计算机的存储介质的例子包括，但不限于相变内存(PRAM)、静态随机存取存储器(SRAM)、动态随机存取存储器(DRAM)、其他类型的随机存取存储器(RAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、快闪记忆体或其他内存技术、只读光盘只读存储器(CD-ROM)、数字多功能光盘(DVD)或其他光学存储、磁盒式磁带，磁带磁磁盘存储或其他磁性存储设备或任何其他非传输介质，可用于存储可以被计算设备访问的信息。按照本文中的界定，计算机可读介质不包括暂存电脑可读媒体(transitory media)，如调制的数据信号和载波。
还需要说明的是，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、商品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、商品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括所述要素的过程、方法、商品或者设备中还存在另外的相同要素。
本领域技术人员应明白，本申请的实施例可提供为方法、系统或计算机程 序产品。因此，本申请可采用完全硬件实施例、完全软件实施例或结合软件和硬件方面的实施例的形式。而且，本申请可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。

Claims (18)

  1. 一种机器人客服转人工客服的方法，其特征在于，包括：
    从机器人客服与用户的至少一轮会话中获取有效特征；
    将有效特征输入信心评估模型，得到机器人客服与用户会话的当前信心评估值；所述信心评估模型采用标记有有效特征和适宜出人工点的机器人客服与用户的会话样本进行训练，所述适宜出人工点为以人工客服代替机器人客服的适当时点；
    在当前信心评估值满足预定出人工条件时，将用户转接人工客服。
  2. 根据权利要求1所述的方法，其特征在于，所述从机器人客服与用户的至少一轮会话中获取有效特征，包括：从机器人客服与用户的至少一轮会话中提取原始特征，按照特征预处理规则对原始特征进行拆分、组合、分类、和/或删除后得到有效特征。
  3. 根据权利要求2所述的方法，其特征在于，所述方法还包括：获取配置文件；所述配置文件中包括原始特征发现规则和/或特征预处理规则；
    所述从机器人客服与用户的至少一轮会话中提取原始特征，包括：从机器人客服与用户的至少一轮会话中，按照原始特征发现规则提取原始特征。
  4. 根据权利要求3所述的方法，其特征在于，所述配置文件中还包括：预定出人工条件。
  5. 根据权利要求2所述的方法，其特征在于，所述原始特征包括：业务相关性、答案匹配度、答案重复次数、是否兜底答案、答案是否为提问、用户明确提出换人工、用户有潜在换人工倾向、用户的情感倾向、和用户在解释自己的问题中的至少一个。
  6. 根据权利要求1所述的方法，其特征在于，所述有效特征由预定数据分析算法根据标记有待定特征、适宜出人工点和实际出人工点的机器人与用户的会话样本以及其中会话的服务效果，对若干待定特征进行评估及整合后得出；所述待定特征能够描述机器人客服的服务质量和用户的转人工意愿。
  7. 根据权利要求6所述的方法，其特征在于，所述有效特征具有各自的权重，由所述预定数据分析算法计算得出；
    所述信心评估模型根据有效特征的权重进行训练。
  8. 根据权利要求1所述的方法，其特征在于，所述预定出人工条件包括：当前信心评估值大于或小于预定信心阈值；
    所述预定信心阈值根据若干个样本信心评估值的覆盖率和准确率确定；所述样本信心评估值为将会话样本输入信心评估模型后得到的输出；所述覆盖率为会话样本中适宜出人工点对应的所有样本信心评估值中，满足预定出人工条件的样本信心评估值所占的比例；所述准确率为所有满足预定出人工条件的样本信心评估值中，对应于适宜出人工点的样本信心评估值所占的比例。
  9. 根据权利要求1所述的方法，其特征在于，所述信心评估模型采用支持向量回归SVR算法、逻辑回归LR算法、或迭代决策树GBDT算法。
  10. 一种机器人客服转人工客服的装置，其特征在于，包括：
    有效特征获取单元，用于从机器人客服与用户的至少一轮会话中获取有效特征；
    当前信心评估单元，用于将有效特征输入信心评估模型，得到机器人客服与用户会话的当前信心评估值；所述信心评估模型采用标记有有效特征和适宜出人工点的机器人客服与用户的会话样本进行训练，所述适宜出人工点为以人工客服代替机器人客服的适当时点；
    人工客服转接单元，用于在当前信心评估值满足预定出人工条件时，将用户转接人工客服。
  11. 根据权利要求10所述的装置，其特征在于，所述有效特征获取单元具体用于：从机器人客服与用户的至少一轮会话中提取原始特征，按照特征预处理规则对原始特征进行拆分、组合、分类、和/或删除后得到有效特征。
  12. 根据权利要求11所述的装置，其特征在于，所述装置还包括：配置文件获取单元，用于获取配置文件；所述配置文件中包括原始特征发现规则 和/或特征预处理规则；
    所述有效特征获取单元具体用于：从机器人客服与用户的至少一轮会话中，按照原始特征发现规则提取原始特征。
  13. 根据权利要求12所述的装置，其特征在于，所述配置文件中还包括：预定出人工条件。
  14. 根据权利要求11所述的装置，其特征在于，所述原始特征包括：业务相关性、答案匹配度、答案重复次数、是否兜底答案、答案是否为提问、用户明确提出换人工、用户有潜在换人工倾向、用户的情感倾向、和用户在解释自己的问题中的至少一个。
  15. 根据权利要求10所述的装置，其特征在于，所述有效特征由预定数据分析算法根据标记有待定特征、适宜出人工点和实际出人工点的机器人与用户的会话样本以及其中会话的服务效果，对若干待定特征进行评估及整合后得出；所述待定特征能够描述机器人客服的服务质量和用户的转人工意愿。
  16. 根据权利要求15所述的装置，其特征在于，所述有效特征具有各自的权重，由所述预定数据分析算法计算得出；
    所述信心评估模型根据有效特征的权重进行训练。
  17. 根据权利要求10所述的装置，其特征在于，所述预定出人工条件包括：当前信心评估值大于或小于预定信心阈值；
    所述预定信心阈值根据若干个样本信心评估值的覆盖率和准确率确定；所述样本信心评估值为将会话样本输入信心评估模型后得到的输出；所述覆盖率为会话样本中适宜出人工点对应的所有样本信心评估值中，满足预定出人工条件的样本信心评估值所占的比例；所述准确率为所有满足预定出人工条件的样本信心评估值中，对应于适宜出人工点的样本信心评估值所占的比例。
  18. 根据权利要求10所述的装置，其特征在于，所述信心评估模型采用支持向量回归SVR算法、逻辑回归LR算法、或迭代决策树GBDT算法。
PCT/CN2017/090322 2016-07-08 2017-06-27 机器人客服转人工客服的方法和装置 WO2018006727A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201610538674.6 2016-07-08
CN201610538674.6A CN107590159A (zh) 2016-07-08 2016-07-08 机器人客服转人工客服的方法和装置

Publications (1)

Publication Number Publication Date
WO2018006727A1 true WO2018006727A1 (zh) 2018-01-11

Family

ID=60904586

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/090322 WO2018006727A1 (zh) 2016-07-08 2017-06-27 机器人客服转人工客服的方法和装置

Country Status (3)

Country Link
CN (1) CN107590159A (zh)
TW (1) TWI677846B (zh)
WO (1) WO2018006727A1 (zh)

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109308320A (zh) * 2018-07-20 2019-02-05 北京智能点科技有限公司 一种机器人多轮对话流程化配置方法
CN110163655A (zh) * 2019-04-15 2019-08-23 中国平安人寿保险股份有限公司 基于梯度提升树的坐席分配方法、装置、设备及存储介质
CN110457578A (zh) * 2019-07-11 2019-11-15 阿里巴巴集团控股有限公司 一种客服服务需求识别方法及装置
CN110602335A (zh) * 2019-08-14 2019-12-20 中国平安财产保险股份有限公司 设置终端优先级的方法、装置、计算机设备和存储介质
KR20200095516A (ko) * 2018-01-26 2020-08-10 알리바바 그룹 홀딩 리미티드 로봇 고객 서비스로부터 사람 고객 서비스로 전환하기 위한 방법 및 장치
CN111583023A (zh) * 2020-05-07 2020-08-25 中国工商银行股份有限公司 业务处理方法、装置和计算机系统
CN111625918A (zh) * 2019-02-27 2020-09-04 阿里巴巴集团控股有限公司 一种工艺参数推荐方法、装置及电子设备
CN112433598A (zh) * 2019-08-07 2021-03-02 科沃斯商用机器人有限公司 人机交互方法、设备及存储介质
CN112543185A (zh) * 2020-11-23 2021-03-23 建信金融科技有限责任公司 一种客户服务方法、装置和系统
CN113051386A (zh) * 2021-04-30 2021-06-29 中国银行股份有限公司 转人工服务的调整方法、装置、电子设备以及存储介质
CN113239170A (zh) * 2021-06-01 2021-08-10 平安科技（深圳）有限公司 基于相互角色感知的对话生成方法、装置、设备及介质
CN113315876A (zh) * 2021-05-27 2021-08-27 中国银行股份有限公司 电话银行服务控制方法、装置、服务器及存储介质
CN113434630A (zh) * 2021-06-25 2021-09-24 平安科技（深圳）有限公司 客服服务评估方法、装置、终端设备及介质
CN114117157A (zh) * 2021-11-19 2022-03-01 招联消费金融有限公司 会话处理方法、装置、计算机设备和存储介质

Families Citing this family (22)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110232573A (zh) * 2018-03-06 2019-09-13 广州供电局有限公司 基于交互式的智能应答系统
CN108418981B (zh) * 2018-03-16 2019-10-29 苏州思必驰信息科技有限公司 为客户提供对话服务的方法及装置
CN110392168B (zh) * 2018-04-16 2021-06-01 华为技术有限公司 呼叫处理方法、装置、服务器、存储介质和系统
CN108712583B (zh) * 2018-06-05 2021-03-23 上海智蕙林医疗科技有限公司 一种基于机器人的人工服务方法及系统
CN110674385A (zh) * 2018-06-14 2020-01-10 阿里巴巴集团控股有限公司 客服升级场景下匹配客服的方法及装置
CN109087175A (zh) * 2018-08-15 2018-12-25 深圳追科技有限公司 客服会话切换的方法、装置及系统
CN109446305A (zh) * 2018-10-10 2019-03-08 长沙师范学院 智能旅游客服系统的构建方法以及系统
CN109685550A (zh) * 2018-12-04 2019-04-26 北京猎户星空科技有限公司 智能设备控制方法、装置、电子设备及存储介质
CN109887525B (zh) * 2019-01-04 2023-04-07 平安科技（深圳）有限公司 智能客服方法、装置及计算机可读存储介质
CN109787885A (zh) * 2019-01-16 2019-05-21 中民乡邻投资控股有限公司 一种问答服务方法
CN109949830B (zh) * 2019-03-12 2021-03-30 中国联合网络通信集团有限公司 用户意图识别方法及设备
CN110472023A (zh) * 2019-07-10 2019-11-19 深圳追一科技有限公司 客服切换方法、装置、计算机设备和存储介质
CN110838014A (zh) * 2019-09-20 2020-02-25 北京智齿博创科技有限公司 在线客服系统中人工智能路由策略
CN112580917B (zh) * 2019-09-30 2024-04-05 深圳无域科技技术有限公司 一种客户特征的评估方法及装置
CN111754061A (zh) * 2019-11-22 2020-10-09 北京沃东天骏信息技术有限公司 控制人机分流的方法、装置、服务器设备及存储介质
CN110990553A (zh) * 2019-12-18 2020-04-10 上海智勘科技有限公司 智能音箱语音交互系统与保险代理人的耦合方法及系统
CN111143537A (zh) * 2019-12-30 2020-05-12 税友软件集团股份有限公司 一种基于智能客服系统的服务方法、装置、设备及介质
TWI751504B (zh) * 2020-02-27 2022-01-01 中華電信股份有限公司 人機協作對話系統與方法
CN111526253B (zh) * 2020-03-09 2022-02-08 深圳追一科技有限公司 呼叫控制方法、装置、计算机设备和存储介质
CN111405129A (zh) * 2020-03-12 2020-07-10 中国建设银行股份有限公司 智能外呼风险监控方法及装置
CN111510563A (zh) * 2020-04-16 2020-08-07 中国银行股份有限公司 智能外呼方法及装置、存储介质及电子设备
CN111985248A (zh) * 2020-06-30 2020-11-24 联想(北京)有限公司 一种信息交互方法以及装置

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8379830B1 (en) * 2006-05-22 2013-02-19 Convergys Customer Management Delaware Llc System and method for automated customer service with contingent live interaction
CN104991887A (zh) * 2015-06-18 2015-10-21 北京京东尚科信息技术有限公司 提供信息的方法及装置
CN105072173A (zh) * 2015-08-03 2015-11-18 谌志群 自动客服和人工客服自动切换的客服方法及系统
CN105591882A (zh) * 2015-12-10 2016-05-18 北京中科汇联科技股份有限公司 一种智能机器人与人混合客服的方法及系统
CN105592237A (zh) * 2014-10-24 2016-05-18 中国移动通信集团公司 一种会话切换的方法、装置及智能客服机器人

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2013019200A1 (en) * 2011-07-31 2013-02-07 Hewlett-Packard Development Company, L.P. Systems and methods of knowledge transfer
TWI546768B (zh) * 2011-12-13 2016-08-21 詹偉強 多網站自動辨識之客服裝置
CN105701088B (zh) * 2016-02-26 2018-12-28 北京京东尚科信息技术有限公司 从机器对话切换到人工对话的方法和装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8379830B1 (en) * 2006-05-22 2013-02-19 Convergys Customer Management Delaware Llc System and method for automated customer service with contingent live interaction
CN105592237A (zh) * 2014-10-24 2016-05-18 中国移动通信集团公司 一种会话切换的方法、装置及智能客服机器人
CN104991887A (zh) * 2015-06-18 2015-10-21 北京京东尚科信息技术有限公司 提供信息的方法及装置
CN105072173A (zh) * 2015-08-03 2015-11-18 谌志群 自动客服和人工客服自动切换的客服方法及系统
CN105591882A (zh) * 2015-12-10 2016-05-18 北京中科汇联科技股份有限公司 一种智能机器人与人混合客服的方法及系统

Cited By (24)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10977664B2 (en) * 2018-01-26 2021-04-13 Advanced New Technologies Co., Ltd. Method and apparatus for transferring from robot customer service to human customer service
KR20200095516A (ko) * 2018-01-26 2020-08-10 알리바바 그룹 홀딩 리미티드 로봇 고객 서비스로부터 사람 고객 서비스로 전환하기 위한 방법 및 장치
KR102445992B1 (ko) * 2018-01-26 2022-09-21 어드밴스드 뉴 테크놀로지스 씨오., 엘티디. 로봇 고객 서비스로부터 사람 고객 서비스로 전환하기 위한 방법 및 장치
CN109308320A (zh) * 2018-07-20 2019-02-05 北京智能点科技有限公司 一种机器人多轮对话流程化配置方法
CN111625918B (zh) * 2019-02-27 2023-05-09 阿里巴巴集团控股有限公司 一种工艺参数推荐方法、装置及电子设备
CN111625918A (zh) * 2019-02-27 2020-09-04 阿里巴巴集团控股有限公司 一种工艺参数推荐方法、装置及电子设备
CN110163655A (zh) * 2019-04-15 2019-08-23 中国平安人寿保险股份有限公司 基于梯度提升树的坐席分配方法、装置、设备及存储介质
CN110163655B (zh) * 2019-04-15 2024-03-05 中国平安人寿保险股份有限公司 基于梯度提升树的坐席分配方法、装置、设备及存储介质
CN110457578B (zh) * 2019-07-11 2023-07-18 创新先进技术有限公司 一种客服服务需求识别方法及装置
CN110457578A (zh) * 2019-07-11 2019-11-15 阿里巴巴集团控股有限公司 一种客服服务需求识别方法及装置
CN112433598A (zh) * 2019-08-07 2021-03-02 科沃斯商用机器人有限公司 人机交互方法、设备及存储介质
CN110602335A (zh) * 2019-08-14 2019-12-20 中国平安财产保险股份有限公司 设置终端优先级的方法、装置、计算机设备和存储介质
CN110602335B (zh) * 2019-08-14 2022-11-18 中国平安财产保险股份有限公司 设置终端优先级的方法、装置、计算机设备和存储介质
CN111583023A (zh) * 2020-05-07 2020-08-25 中国工商银行股份有限公司 业务处理方法、装置和计算机系统
CN112543185A (zh) * 2020-11-23 2021-03-23 建信金融科技有限责任公司 一种客户服务方法、装置和系统
CN113051386B (zh) * 2021-04-30 2024-02-02 中国银行股份有限公司 转人工服务的调整方法、装置、电子设备以及存储介质
CN113051386A (zh) * 2021-04-30 2021-06-29 中国银行股份有限公司 转人工服务的调整方法、装置、电子设备以及存储介质
CN113315876A (zh) * 2021-05-27 2021-08-27 中国银行股份有限公司 电话银行服务控制方法、装置、服务器及存储介质
CN113239170A (zh) * 2021-06-01 2021-08-10 平安科技（深圳）有限公司 基于相互角色感知的对话生成方法、装置、设备及介质
CN113239170B (zh) * 2021-06-01 2023-11-28 平安科技（深圳）有限公司 基于相互角色感知的对话生成方法、装置、设备及介质
CN113434630A (zh) * 2021-06-25 2021-09-24 平安科技（深圳）有限公司 客服服务评估方法、装置、终端设备及介质
CN113434630B (zh) * 2021-06-25 2023-07-25 平安科技（深圳）有限公司 客服服务评估方法、装置、终端设备及介质
CN114117157A (zh) * 2021-11-19 2022-03-01 招联消费金融有限公司 会话处理方法、装置、计算机设备和存储介质
CN114117157B (zh) * 2021-11-19 2024-04-09 招联消费金融股份有限公司 会话处理方法、装置、计算机设备和存储介质

Also Published As

Publication number Publication date
CN107590159A (zh) 2018-01-16
TWI677846B (zh) 2019-11-21
TW201804420A (zh) 2018-02-01

Similar Documents

Publication Publication Date Title
WO2018006727A1 (zh) 机器人客服转人工客服的方法和装置
CN110334201B (zh) 一种意图识别方法、装置及系统
US11249774B2 (en) Realtime bandwidth-based communication for assistant systems
WO2020143844A1 (zh) 意图分析方法、装置、显示终端及计算机可读存储介质
US10977664B2 (en) Method and apparatus for transferring from robot customer service to human customer service
US9990923B2 (en) Automated software execution using intelligent speech recognition
CN110032641B (zh) 计算机执行的、利用神经网络进行事件抽取的方法及装置
WO2021159632A1 (zh) 智能问答方法、装置、计算机设备及计算机存储介质
US11704500B2 (en) Techniques to add smart device information to machine learning for increased context
US20220164683A1 (en) Generating a domain-specific knowledge graph from unstructured computer text
WO2020147395A1 (zh) 基于情感的文本分类处理方法、装置和计算机设备
JP2020521210A (ja) 情報処理方法及び端末、コンピュータ記憶媒体
CN110019729B (zh) 智能问答方法及存储介质、终端
Díaz-Morales Cross-device tracking: Matching devices and cookies
WO2022227162A1 (zh) 问答数据处理方法、装置、计算机设备及存储介质
CN110597965B (zh) 文章的情感极性分析方法、装置、电子设备及存储介质
CN110633475A (zh) 基于计算机场景的自然语言理解方法、装置、系统和存储介质
US20210256220A1 (en) Method and system for automated autonomous intent mining
CN112667792B (zh) 人机对话数据处理方法、装置、计算机设备和存储介质
AU2019298610B2 (en) Search device and program
CN111241236B (zh) 任务导向性问答方法、系统、电子设备、可读存储介质
CN113515642A (zh) 电话营销关键断点确定方法及装置
CN110399482B (zh) 文本分类方法、模型和装置
CN113157863B (zh) 问答数据处理方法、装置、计算机设备及存储介质
US20240152871A1 (en) Intelligent generation of job profiles

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: 17823539

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: 17823539

Country of ref document: EP

Kind code of ref document: A1