WO2021073258A1 - 基于情绪分析的任务跟催方法、装置、设备及存储介质 - Google Patents

基于情绪分析的任务跟催方法、装置、设备及存储介质 Download PDF

Info

Publication number
WO2021073258A1
WO2021073258A1 PCT/CN2020/111405 CN2020111405W WO2021073258A1 WO 2021073258 A1 WO2021073258 A1 WO 2021073258A1 CN 2020111405 W CN2020111405 W CN 2020111405W WO 2021073258 A1 WO2021073258 A1 WO 2021073258A1
Authority
WO
WIPO (PCT)
Prior art keywords
follow
task
emotion
customer service
communication content
Prior art date
Application number
PCT/CN2020/111405
Other languages
English (en)
French (fr)
Inventor
许永夫
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 WO2021073258A1 publication Critical patent/WO2021073258A1/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
    • G06QINFORMATION 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
    • G06Q40/00Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/02Banking, e.g. interest calculation or account maintenance

Definitions

  • This application relates to the field of speech recognition technology, and in particular to a task follow-up method, device, equipment and storage medium based on sentiment analysis.
  • the current account and reminder methods are more traditional collection methods, such as phone calls, text messages, emails, letters, etc., and the actual customer reach rate is relatively low.
  • a more intelligent online collection method is currently being produced, the inventor realized that the existing online collection system realizes a relatively single function, and many data still need to be analyzed and judged manually, so the analysis results are very subjective. This in turn affects the follow-up rate of credit institutions.
  • the main purpose of this application is to provide a task follow-up method, device, equipment, and computer-readable storage medium based on sentiment analysis, aiming to solve the problem of weak data analysis capabilities of existing online follow-up methods, resulting in low follow-up contact rates. High technical issues.
  • the first aspect of this application provides a task follow-up method based on sentiment analysis, including: obtaining a follow-up task request initiated by a customer service terminal; extracting task information from the follow-up task request, and The task information generates a follow-up task list; the follow-up task is extracted from the follow-up task list, and the corresponding follow-up task utterance template is called according to the task information of the follow-up task; according to the follow-up task utterance template and Preset the follow-up task strategy, generate the corresponding follow-up task text and send it to the follow-up object; record the content of the follow-up communication between the customer service end and the follow-up object, and exchange content with the follow-up object Perform emotion analysis to obtain the current emotion of the customer service; according to the current emotion of the customer service, send corresponding emotion management prompts to the customer service terminal for the customer service to adjust or maintain the emotion in the process of performing the follow-up task.
  • the second aspect of the present application provides a task follow-up device based on sentiment analysis, including a memory, a processor, and computer-readable instructions stored on the memory and running on the processor, and the processor executes
  • the computer-readable instruction implements the following steps: receiving a query request for legal case information initiated by a client; extracting query keywords in the query request; obtaining a follow-up task request initiated by a customer service terminal; and from the follow-up task Extract task information from the request, and generate a follow-up task list based on the task information; extract follow-up tasks from the follow-up task list, and call a corresponding follow-up task utterance template according to the task information of the follow-up task; According to the follow-up task utterance template and the preset follow-up task strategy, generate the corresponding follow-up task text and send it to the follow-up object; record the content of the follow-up communication between the customer service end and the follow-up object end , And perform sentiment analysis on the content of the follow-up communication to obtain the current sentiment
  • the third aspect of the present application provides a computer-readable storage medium in which computer instructions are stored.
  • the computer executes the following steps: Obtain the information initiated by the customer service terminal Follow-up task request; extract task information from the follow-up task request, and generate a follow-up task list based on the task information; extract follow-up tasks from the follow-up task list, and according to the The task information calls the corresponding follow-up task utterance template; according to the follow-up task utterance template and the preset follow-up task strategy, the corresponding follow-up task text is generated and sent to the follow-up object; record the customer service end and the The content of the follow-up exchanges between the end of the reminder and the sentiment analysis of the content of the follow-up exchanges to obtain the current sentiment of the customer service; according to the current sentiment of the customer service, the corresponding emotion management prompts are sent to the client to provide The customer service adjusts or maintains the situation in the process of performing follow-up tasks.
  • the fourth aspect of the present application provides a task follow-up device based on sentiment analysis, including: an acquisition module, used to obtain a follow-up task request initiated by a customer service terminal; a first generation module, used to obtain a follow-up task request from the Extract task information, and generate a follow-up task list according to the task information; the calling module is used to extract follow-up tasks from the follow-up task list, and call the corresponding follow-up tasks according to the task information of the follow-up tasks Discourse template; a second generation module, used to generate the corresponding follow-up task text according to the follow-up task utterance template and preset follow-up task strategy and send it to the end of the follow-up task; sentiment analysis module, used to record the The follow-up communication content between the customer service terminal and the follow-up object terminal, and the sentiment analysis of the follow-up communication content is performed to obtain the current sentiment of the customer service; the prompt module is used to report the current mood of the customer service to the customer service The terminal sends corresponding emotion management prompts for the
  • this application can use the preset follow-up task utterance template and follow-up task strategy to generate the corresponding follow-up task text and send it to the follow-up object, thereby standardizing the follow-up task and avoiding the follow-up task. Subject feels disgusted.
  • the follow-up makes the follow-up object disgusted. Since the collection process can be monitored and managed, it can prevent the customer service from generating bad emotions and affect the follow-up effect, and increase the reach rate of the task follow-up.
  • FIG. 1 is a schematic structural diagram of an operating environment of a task follow-up device based on sentiment analysis involved in a solution of an embodiment of the application;
  • FIG. 2 is a schematic flowchart of a first embodiment of a task follow-up method based on sentiment analysis in this application;
  • FIG. 3 is a detailed flowchart of an embodiment of step S50 in FIG. 2;
  • FIG. 4 is a schematic flowchart of a second embodiment of a task follow-up method based on sentiment analysis according to this application;
  • FIG. 5 is a schematic diagram of functional modules of an embodiment of a task follow-up device based on sentiment analysis according to the present application.
  • the embodiments of the present application provide a task follow-up method, device, equipment and storage medium based on sentiment analysis, which are used to generate corresponding follow-up task text using preset follow-up task utterance templates and follow-up task strategies, in real time Obtain the follow-up communication content between the customer service and the follower, and analyze the sentiment of the follow-up communication content to determine the current mood of the customer service, avoid the negative emotions of the customer service and affect the follow-up effect, and improve the reach of the task follow-up rate.
  • This application provides a task follow-up device based on sentiment analysis.
  • FIG. 1 is a schematic structural diagram of an operating environment of a task follow-up device based on sentiment analysis involved in the solution of an embodiment of the present application.
  • the task follow-up device based on sentiment analysis includes a processor 1001, such as a CPU, a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005.
  • the communication bus 1002 is used to implement connection and communication between these components.
  • the user interface 1003 may include a display screen (Display) and an input unit such as a keyboard (Keyboard), and the network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface).
  • the memory 1005 may be a high-speed RAM memory, or a non-volatile memory (non-volatile memory), such as a magnetic disk memory.
  • the memory 1005 may also be a storage device independent of the aforementioned processor 1001.
  • the hardware structure of the task tracking device based on sentiment analysis shown in FIG. 1 does not constitute a limitation on the task tracking device based on sentiment analysis, and may include more or less than that shown in the figure. Components, or a combination of certain components, or different component arrangements.
  • the memory 1005 which is a computer-readable storage medium, may include an operating system, a network communication module, a user interface module, and a task follow-up program.
  • the operating system is a program that manages and controls the task follow-up equipment and software resources based on sentiment analysis, and supports the operation of the task follow-up program and other software and/or programs.
  • the network interface 1004 is mainly used to access the network; the user interface 1003 is mainly used to detect confirmation instructions and edit instructions, and the processor 1001 can be used
  • the task follow-up program stored in the memory 1005 is called, and the operations of the following embodiments of the task follow-up method based on sentiment analysis are executed.
  • FIG. 2 is a schematic flowchart of a first embodiment of a task follow-up method based on sentiment analysis in this application.
  • the task follow-up method based on sentiment analysis includes the following steps:
  • Step S10 Obtain the follow-up task request initiated by the customer service terminal
  • the follow-up task request initiated by the customer service terminal may include multiple follow-up tasks of the same or different types.
  • Step S20 extract task information from the follow-up task request, and generate a follow-up task list according to the task information
  • the follow-up tasks can be set with different priority processing levels in advance.
  • account and reminder tasks can be divided into multiple task levels, such as ordinary, urgent, and urgent.
  • the customer service is generally prioritized to perform account and reminder tasks with a task level of "urgent".
  • the task information extracted from the follow-up task request is not limited, for example, it may be the task name, the basic task information, the basic information of the follow-up object, and the task level. It should be noted that the follow-up task list contains at least one follow-up task. When there are multiple tasks in the follow-up task list, each follow-up task needs to be sorted.
  • the sorting rule is that the follow-up tasks with a task level of "urgent" are preferably ranked first, and follow-up tasks of the same level are sorted according to other rules, such as task type, task generation time, and so on.
  • a corresponding follow-up task list is generated, which records the basic information of all follow-up tasks to be executed this time and each follow-up task The order of execution.
  • Step S30 extracting a follow-up task from the follow-up task list, and call a corresponding follow-up task utterance template according to the task information of the follow-up task;
  • the customer service use the preset follow-up task utterance goal to communicate with the follow-up object.
  • the follow-up task discourse template is set based on the specific follow-up task. Different follow-up tasks correspond to different discourse templates.
  • the use of the discourse template can generate standardized and neat professional speech, thereby reflecting the professionalism of the service and improving customer reach To avoid unprofessional follow-up methods that affect or bring disadvantages to customers.
  • the follow-up task utterance template contains basic terms related to the follow-up task, the content of the question and the response method that the customer service needs to use, and so on.
  • the task information of the follow-up task is obtained first, such as obtaining basic information of the task, basic information of the follow-up object, task level, etc., and then based on the obtained task information.
  • the task information searches the database, determines the follow-up task utterance template that matches the task information and calls it.
  • Step S40 according to the follow-up task utterance template and the preset follow-up task strategy, generate a corresponding follow-up task text and send it to the follow-up object terminal;
  • a corresponding follow-up task strategy is preset, which is set according to the specific personal situation of the follower, that is, different
  • the follow-up objects correspond to different follow-up task strategies. For example, if the customer likes to chat, the follow-up task strategy can start with the chat, and then generate some follow-up task text related to the chat. Or customers like direct communication, the follow-up task strategy can be to directly cut into the topic, avoid politeness, and then generate follow-up task text related to the follow-up task.
  • the follow-up task strategy is a business rule that contains multiple preset rules for generating follow-up task text. According to the rules corresponding to the follow-up task strategy, use the follow-up task utterance template to generate the follow-up task text, and then send the generated follow-up task text to the end of the follow-up task, for example, use email, chat software, etc. to communicate, then The end of the reminder can be a mailbox, a WeChat account, etc.
  • Step S50 recording the follow-up communication content between the customer service terminal and the follow-up object terminal, and perform sentiment analysis on the follow-up communication content to obtain the current sentiment of the customer service;
  • This embodiment is not limited to the implementation manner of sentiment analysis. For example, set the scores corresponding to various emotions in advance, set various emotional keywords and the corresponding scores of each emotional keyword at the same time, then retrieve the obtained follow-up communication content to extract emotional keywords, and finally count the scores of all emotional keywords The total score, and then determine the corresponding emotion according to the total score of the emotional keyword.
  • Step S60 according to the current emotions of the customer service, send corresponding emotion management prompts to the customer service terminal for the customer service to adjust or maintain the emotions in the process of performing the follow-up task.
  • the customer service can be emotionally managed, specifically by sending a reminder to remind the customer service to adjust or maintain the emotion in the process of performing the follow-up task. For example, if the current mood of the customer service becomes worse, the customer is prompted to remain calm and adjust emotions, and serve with a smile; if the current emotion of the customer service is stable and smile, the customer is prompted to maintain the current emotion to perform follow-up tasks.
  • emotional management of customer service is specifically carried out in the following manner:
  • the emotion management prompt is sent to the customer service terminal for the customer service to adjust or maintain the emotions in the process of performing the follow-up task.
  • various emotions are quantified in advance, and then a comparison relationship table between the emotion value and the emotion management prompt is set, that is, any different emotion is provided with a corresponding emotion management prompt. Emotion management tips corresponding to the current emotions of the customer service can be obtained by looking up the table.
  • the preset follow-up task utterance template and follow-up task strategy can be used to generate the corresponding follow-up task text and send it to the follow-up object, thereby standardizing the follow-up task and avoiding follow-up tasks.
  • the object is urged to become disgusted.
  • the follow-up will make the follow-up object disgusted. Since the collection process can be monitored and managed, it can avoid customer service failures.
  • FIG. 3 is a detailed flowchart of an embodiment of step S50 in FIG. 2.
  • the above step S50 further includes:
  • Step S501 Record the follow-up communication content between the customer service terminal and the follow-up object terminal;
  • Step S502 Obtain the communication content sent by the customer service to the follower from the follow-up communication content
  • the customer service end when the follow-up task is executed, the customer service end will communicate with the follow-up object end (that is, the client), that is, the exchange information is sent to each other in a chat mode.
  • the follow-up object end that is, the client
  • the content of the chat can be text or voice.
  • This embodiment is not limited to the way of recording the content of the follow-up communication, for example, information is collected and sorted through a burying method, so as to obtain the follow-up communication content.
  • Step S503 If the exchange content is text information, input the exchange content into a preset emotion classification model for emotion recognition and classification, and obtain an emotion classification result;
  • Step S504 If the communication content is voice information, convert the voice information into text information through voice recognition and input the emotion classification model to perform emotion recognition and classification to obtain an emotion classification result;
  • the communication content is text information
  • the communication content is directly input into the preset emotion classification model to recognize and classify emotions
  • the voice information is first converted into voice information through voice recognition. Text information, and then input the converted text information into the emotion classification model for processing.
  • the training process of the emotion classification model is as follows: first collect multiple chat records of several different customer service and multiple different task objects and perform word segmentation, and then determine the emotional characteristics of each word according to the part of speech of each word, and finally Then train the sentiment classifier according to the sentiment characteristics of each word to construct the sentiment classification model.
  • the emotion classification result contains multiple emotions corresponding to the same text information. For example, a sentence contains both positive emotions and negative emotions. Therefore, it is necessary to further analyze the multiple emotions in the classification result. Perform regression analysis to determine the sentiment that best fits the current customer service.
  • Step S505 Perform regression analysis on the emotion classification result to obtain the regression value of each emotion in the emotion classification result;
  • Step S506 Calculate the score of each emotion according to the regression value of each emotion and use the emotion with the highest score as the current emotion of the customer service.
  • Regression analysis is a statistical analysis method to determine the quantitative relationship between two or more variables. That is, by analyzing the specific forms of correlation between phenomena to determine their causal relationship, and using mathematical models to express their specific relationships.
  • regression analysis is first used to obtain the regression value of each emotion in the emotion classification result, and then the score of each emotion is calculated according to the regression value of each emotion.
  • the score of each emotion is calculated according to the preset emotion weight value.
  • the following formula is used to calculate the score of each emotion:
  • T i represents the score of the i-th emotion
  • Vi ,3 and Vi ,4 respectively represent the third-level regression value and the fourth-level regression value of the i-th emotion obtained by the hierarchical regression method
  • i is Positive integer.
  • the score of each emotion is calculated, and then the emotion with the highest score is used as the current emotion of the customer service.
  • FIG. 4 is a schematic flowchart of a second embodiment of a task follow-up method based on sentiment analysis in this application.
  • the method further includes:
  • Step S70 After the follow-up task is finished, extract the key data related to the follow-up object generated during the process of the customer service performing the follow-up task;
  • Step S80 drawing a task data curve according to the key data, wherein the task data curve uses time as the abscissa and the key data as the ordinate;
  • the key data is specifically generated during the execution of the follow-up task by the customer service, and the specific content and expression form of the key data are not limited in this embodiment.
  • the key data is data about the emotional changes in the customer service follow-up process, and the emotional response of the follow-up object.
  • the task data curve is drawn based on the key data.
  • the task data curve uses time as the abscissa and key data as the ordinate, which can reflect the future Changes in the key data of the user, such as reflecting the emotional changes in the customer service follow-up process at a certain point in the future, and the emotional response of the follow-up object.
  • Step S90 Perform trend prediction on the task data curve by using a preset trend prediction algorithm to obtain a trend prediction result
  • a preset trend prediction algorithm can be used to perform trend prediction. It should be noted that trend prediction is mainly used to predict the emotional response of the follower, so as to facilitate subsequent adjustments to the follow-up task strategy.
  • the trend prediction is performed in the following manner:
  • t represents time
  • Y t represents the predicted value corresponding to time t
  • x t represents the average value of data corresponding to time t
  • Y t-1 represents the predicted value corresponding to time t-1
  • m is a constant whose value range is [ 0.5, 1].
  • the adjacent key data pairs corresponding to each time point in the task data curve are first obtained, and the data of all the key data pairs are calculated Average, then draw the average curve of the data, and finally predict the trend of the average curve of the data.
  • step S100 the trend prediction result is imported into the follow-up task strategy of the follow-up object as a judgment basis when the follow-up task is executed next time for the follow-up object.
  • the trend prediction result is imported into the follow-up task strategy for the next time the follow-up task is executed.
  • the trend prediction results are further screened.
  • the specific implementation is as follows: First, the trend prediction results Import the task data curve, and then generate the trend line of the task data curve according to the least square method, and finally judge whether the slope of the task data curve trend line is greater than the preset threshold, if so, it is judged that the trend prediction result is not referable. Therefore, it cannot be imported into the follow-up task strategy, otherwise it is judged that the trend prediction result is referable, so it can be imported into the follow-up task strategy as the judgment basis for the next time the follow-up task is performed on the follow-up object.
  • the application also provides a task follow-up device based on sentiment analysis.
  • FIG. 5 is a schematic diagram of functional modules of an embodiment of a task follow-up device based on sentiment analysis of the present application.
  • the task follow-up device based on sentiment analysis includes:
  • the obtaining module 10 is used to obtain the follow-up task request initiated by the customer service terminal;
  • the first generating module 20 is configured to extract task information from the follow-up task request, and generate a follow-up task list according to the task information;
  • the calling module 30 is used to extract follow-up tasks from the follow-up task list, and call the corresponding follow-up task utterance template according to the task information of the follow-up tasks;
  • the second generating module 40 is configured to generate a corresponding follow-up task text according to the follow-up task utterance template and a preset follow-up task strategy and send it to the follow-up object;
  • the sentiment analysis module 50 is used to record the follow-up communication content between the customer service and the follow-up object, and perform sentiment analysis on the follow-up communication content to obtain the current sentiment of the customer service;
  • the prompt module 60 is configured to send corresponding emotion management prompts to the customer service according to the current emotion of the customer service, so that the customer service can adjust or maintain the emotion in the process of performing the follow-up task.
  • the task follow-up device when the task follow-up device initiates a follow-up task request, it can use the preset follow-up task utterance template and follow-up task strategy to generate the corresponding follow-up task text and send it to the follow-up object, thereby enabling the follow-up task Standardize the task of reminders to avoid resentment with the reminder.
  • the task follow-up device obtains the content of the follow-up communication between the customer service and the follower in real time, and performs sentiment analysis on the content of the follow-up communication to determine the current sentiment of the customer service, and then adjust the sentiment of the customer service in time ,
  • the collection process can be monitored and managed, which can prevent customer service from generating bad emotions and affect the follow-up effect, and increase the reach rate of task follow-up.
  • the present application also provides a task follow-up device based on sentiment analysis, including: a memory and at least one processor, the memory stores instructions, the memory and the at least one processor are interconnected by wires; the at least one processor A processor calls the instructions in the memory, so that the sentiment analysis-based task follow-up device executes the steps in the above-mentioned sentiment analysis-based task follow-up method.
  • the present application also provides a computer-readable storage medium.
  • the computer-readable storage medium may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium.
  • the computer-readable storage medium stores computer instructions, and when the computer instructions are executed on the computer, the computer executes the following steps:
  • follow-up task utterance template and the preset follow-up task strategy According to the follow-up task utterance template and the preset follow-up task strategy, generate the corresponding follow-up task text and send it to the follow-up object;
  • corresponding emotion management prompts are sent to the customer service terminal for the customer service to adjust or maintain the emotions in the process of performing the follow-up task.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Finance (AREA)
  • Accounting & Taxation (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Human Computer Interaction (AREA)
  • Artificial Intelligence (AREA)
  • Development Economics (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Strategic Management (AREA)
  • Technology Law (AREA)
  • General Business, Economics & Management (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本申请涉及人工智能领域,公开了一种基于情绪分析的任务跟催方法,包括:获取客服端发起的跟催任务请求;从跟催任务请求中提取任务信息并生成跟催任务清单;从跟催任务清单中提取跟催任务并调用对应的跟催任务话语模板;根据跟催任务话语模板和预置跟催任务策略,生成对应的跟催任务文本并发送至跟催对象;记录客服端与跟催对象端之间的跟催交流内容,并对跟催交流内容进行情绪分析,得到客服的当前情绪;根据客服的当前情绪,向客服端发送相应的情绪管理提示,以供客服调整或保持在执行跟催任务过程中的情绪。本申请还公开了一种基于情绪分析的任务跟催装置、设备及计算机可读存储介质。本申请实现了跟催过程的监控管理,提升了跟催触达率。

Description

基于情绪分析的任务跟催方法、装置、设备及存储介质
本申请要求于2019年10月18日提交中国专利局、申请号为201910990755.3、发明名称为“基于情绪分析的任务跟催方法、装置、设备及存储介质”的中国专利申请的优先权,其全部内容通过引用结合在申请中。
技术领域
本申请涉及语音识别技术领域,尤其涉及一种基于情绪分析的任务跟催方法、装置、设备及存储介质。
背景技术
随着国民经济地稳步发展,信用卡等金融信贷技术已经步入规模化发展阶段。为在激烈的市场竞争中生存,各信贷机构在追求客户规模的大比例增长的同时,也在力求提升服务质量,减少逾期坏账。目前国内征信制度存在较多缺陷,导致不良账款居高不下。
目前的账款跟催方式比较传统的催收方式,如电话、短信、邮件、信函等,客户实际触达率比较低。尽管当前也产生了更智能化的线上催收方式,发明人意识到,现有的线上催收系统实现的功能比较单一,很多数据仍旧需要人工进行分析和判断,因而分析结果存在很大的主观性,进而影响信贷机构的跟催触达率。
发明内容
本申请的主要目的在于提供一种基于情绪分析的任务跟催方法、装置、设备及计算机可读存储介质,旨在解决现有线上跟催方式的数据分析能力薄弱而导致跟催触达率不高的技术问题。
为实现上述目的,本申请第一方面提供了一种基于情绪分析的任务跟催方法,包括:获取客服端发起的跟催任务请求;从所述跟催任务请求中提取任务信息,并根据所述任务信息生成跟催任务清单;从所述跟催任务清单中提取跟催任务,并根据所述跟催任务的任务信息调用对应的跟催任务话语模板;根据所述跟催任务话语模板和预置跟催任务策略,生成对应的跟催任务文本并发送至跟催对象端;记录所述客服端与所述跟催对象端之间的跟催交流内容,并对所述跟催交流内容进行情绪分析,得到客服的当前情绪;根据客服的当前情绪,向所述客服端发送相应的情绪管理提示,以供客服调整或保持在执行跟催任务过程中的情绪。
本申请第二方面提供了一种基于情绪分析的任务跟催设备,包括存储器、处理器及存储在所述存储器上并可在所述处理器上运行的计算机可读指令,所述处理器执行所述计算机可读指令时实现如下步骤:接收客户端发起的法律案件信息的查询请求;提取所述查询请求中的查询关键词;获取客服端发起的跟催任务请求;从所述跟催任务请求中提取任务信息,并根据所述任务信息生成跟催任务清单;从所述跟催任务清单中提取跟催任务,并根据所述跟催任务的任务信息调用对应的跟催任务话语模板;根据所述跟催任务话语模板和预置跟催任务策略,生成对应的跟催任务文本并发送至跟催对象端;记录所述客服端与所述跟催对象端之间的跟催交流内容,并对所述跟催交流内容进行情绪分析,得到客服的当前情绪;根据客服的当前情绪,向所述客服端发送相应的情绪管理提示,以供客服调整或保持在执行跟催任务过程中的情。
本申请第三方面提供了一种计算机可读存储介质,所述计算机可读存储介质中存储有计算机指令,当所述计算机指令在计算机上运行时,使得计算机执行如下步骤:获取客服端发起的跟催任务请求;从所述跟催任务请求中提取任务信息,并根据所述任务信息生成跟催任务清单;从所述跟催任务清单中提取跟催任务,并根据所述跟催任务的任务信息调用对应的跟催任务话语模板;根据所述跟催任务话语模板和预置跟催任务策略,生成对应的跟催任务文本并发送至跟催对象端;记录所述客服端与所述跟催对象端之间的跟催交流 内容,并对所述跟催交流内容进行情绪分析,得到客服的当前情绪;根据客服的当前情绪,向所述客服端发送相应的情绪管理提示,以供客服调整或保持在执行跟催任务过程中的情。
本申请第四方面提供了一种基于情绪分析的任务跟催装置,包括:获取模块,用于获取客服端发起的跟催任务请求;第一生成模块,用于从所述跟催任务请求中提取任务信息,并根据所述任务信息生成跟催任务清单;调用模块,用于从所述跟催任务清单中提取跟催任务,并根据所述跟催任务的任务信息调用对应的跟催任务话语模板;第二生成模块,用于根据所述跟催任务话语模板和预置跟催任务策略,生成对应的跟催任务文本并发送至跟催对象端;情绪分析模块,用于记录所述客服端与所述跟催对象端之间的跟催交流内容,并对所述跟催交流内容进行情绪分析,得到客服的当前情绪;提示模块,用于根据客服的当前情绪,向所述客服端发送相应的情绪管理提示,以供客服调整或保持在执行跟催任务过程中的情绪。
本申请在发起跟催任务请求时,能够使用预置的跟催任务话语模板和跟催任务策略,生成对应的跟催任务文本并发送至跟催对象,从而使跟催任务标准化,避免跟催对象产生反感。同时,进一步在跟催过程中,实时获取客服与跟催对象之间的跟催交流内容,并对跟催交流内容进行情绪分析,以确定客服的当前情绪,进而及时调整客服情绪,避免情绪化跟催而使跟催对象产生反感,由于催收过程中可进行监控与管理,因而可避免客服产生不良情绪而影响跟催效果,提升了任务跟催的触达率。
附图说明
图1为本申请实施例方案涉及的基于情绪分析的任务跟催设备运行环境的结构示意图;
图2为本申请基于情绪分析的任务跟催方法第一实施例的流程示意图;
图3为图2中步骤S50一实施例的细化流程示意图;
图4为本申请基于情绪分析的任务跟催方法第二实施例的流程示意图;
图5为本申请基于情绪分析的任务跟催装置一实施例的功能模块示意图。
本申请目的的实现、功能特点及优点将结合实施例,参照附图做进一步说明。
具体实施方式
本申请实施例提供了一种基于情绪分析的任务跟催方法、装置、设备及存储介质,用于使用预置的跟催任务话语模板和跟催任务策略,生成对应的跟催任务文本,实时获取客服与跟催对象之间的跟催交流内容,并对跟催交流内容进行情绪分析,以确定客服的当前情绪,避免客服产生不良情绪而影响跟催效果,提升了任务跟催的触达率。
应当理解,此处所描述的具体实施例仅用以解释本申请,并不用于限定本申请。
本申请提供一种基于情绪分析的任务跟催设备。
参照图1,图1为本申请实施例方案涉及的基于情绪分析的任务跟催设备运行环境的结构示意图。
如图1所示,该基于情绪分析的任务跟催设备包括:处理器1001,例如CPU,通信总线1002、用户接口1003,网络接口1004,存储器1005。其中,通信总线1002用于实现这些组件之间的连接通信。用户接口1003可以包括显示屏(Display)、输入单元比如键盘(Keyboard),网络接口1004可选的可以包括标准的有线接口、无线接口(如WI-FI接口)。存储器1005可以是高速RAM存储器,也可以是稳定的存储器(non-volatile memory),例如磁盘存储器。存储器1005可选的还可以是独立于前述处理器1001的存储装置。
本领域技术人员可以理解,图1中示出的基于情绪分析的任务跟催设备的硬件结构并不构成对基于情绪分析的任务跟催设备的限定,可以包括比图示更多或更少的部件,或者组合某些部件,或者不同的部件布置。
如图1所示,作为一种计算机可读存储介质的存储器1005中可以包括操作系统、网络通信模块、用户接口模块以及任务跟催程序。其中,操作系统是管理和控制基于情绪分析的任务跟催设备和软件资源的程序,支持任务跟催程序以及其它软件和/或程序的运行。
在图1所示的基于情绪分析的任务跟催设备的硬件结构中,网络接口1004主要用于接入网络;用户接口1003主要用于侦测确认指令和编辑指令等,而处理器1001可以用于调用存储器1005中存储的任务跟催程序,并执行以下基于情绪分析的任务跟催方法的各实施例的操作。
基于上述基于情绪分析的任务跟催设备硬件结构,提出本申请基于情绪分析的任务跟催方法的各个实施例。
参照图2,图2为本申请基于情绪分析的任务跟催方法第一实施例的流程示意图。本实施例中,所述基于情绪分析的任务跟催方法包括以下步骤:
步骤S10,获取客服端发起的跟催任务请求;
本实施例对于客服需要跟催的任务内容与表现形式不限,具体根据实际需要进行设置。比如,一般事务跟催、借贷款跟催等。此外,客服端发起的跟催任务请求中可以包含有多个相同或不同类型的跟催任务。
步骤S20,从所述跟催任务请求中提取任务信息,并根据所述任务信息生成跟催任务清单;
本实施例中,为便于对不同类型的任务进行合理有效跟催,因此,可预先将跟催任务设置不同的优先处理等级。例如,账款跟催任务可以划分为普通、加急、紧急等多个任务等级,根据任务等级的不同,一般优先安排客服执行任务等级为“紧急”的账款跟催任务。
本实施例中,对于从跟催任务请求中提取的任务信息不限,例如可以是任务名称、任务基本信息、跟催对象基本信息、任务等级等。需要说明的是,跟催任务清单中至少包含有一个跟催任务,当跟催任务清单中存在多个任务时,需要对各跟催任务进行排序。
在一实施例中,排序规则为优选将任务等级为“紧急”的跟催任务排在前列,而对于同一等级的跟催任务则按照其他规则进行排序,比如任务的类型、任务生成时间等。
本实施例中,在获得跟催任务请求中各任务的任务信息后,生成对应的跟催任务清单,该任务清单上记载有本次待执行的所有跟催任务的基本信息以及各跟催任务的执行顺序。
步骤S30,从所述跟催任务清单中提取跟催任务,并根据所述跟催任务的任务信息调用对应的跟催任务话语模板;
本实施例中,为进一步提升客服执行跟催任务的触达率,因此,优选客服使用预置的跟催任务话语目标与跟催对象进行交流。跟催任务话语模板是基于具体跟催任务而设定的,不同的跟催任务对应不同的话语模板,使用话语模板能够生成规范、工整的专业话术,从而体现服务专业性,提升客户触达率,避免不专业跟催方式对客户造成影响或带来不利。
本实施例对于跟催任务话语模板的具体内容及表现形式不限,比如,跟催任务话语模板中包含了与跟催任务相关的基本术语、客服需要用到的提问内容以及答复方式等。
本实施例中,在从跟催任务清单中提取跟催任务并执行时,先获得跟催任务的任务信息,比如获得任务基本信息、跟催对象基本信息、任务等级等,然后再根据获得的任务信息查找数据库,确定与任务信息匹配的跟催任务话语模板并进行调用。
其中,需要预先在数据库中存储各种跟催任务话语模板,并建立模板与任务之间的关联映射关系。
步骤S40,根据所述跟催任务话语模板和预置跟催任务策略,生成对应的跟催任务文本并发送至跟催对象端;
本实施例中,为了让跟催对象更容易接受客服端发送的跟催任务文本,因此预先设置 有相应的跟催任务策略,该策略具体依据跟催对象的具体个人情况而设置,也即不同的跟催对象对应不同的跟催任务策略。例如,客户喜欢聊天,则跟催任务策略可以是从聊天入手,进而生成一些与聊天相关的跟催任务文本。或者客户喜欢直接沟通,则跟催任务策略可以是直接切入主题,避免客套,进而生成与跟催任务相关的跟催任务文本。
本实施例对于跟催任务策略的具体内容及表现方式不限,具体根据实际需要进行设置。跟催任务策略是一种业务规则,包含有多条预置的用于生成跟催任务文本的规则。根据跟催任务策略对应的规则,使用跟催任务话语模板生成跟催任务文本,然后再将生成的跟催任务文本发送给跟催对象端,例如,使用邮件、聊天软件等方式进行交流,则跟催对象端可以是邮箱、微信号等。
步骤S50,记录所述客服端与所述跟催对象端之间的跟催交流内容,并对所述跟催交流内容进行情绪分析,得到客服的当前情绪;
本实施例中,客服与跟催对象进行交流时,容易产生情绪,因此需要对客服情绪进行监控分析,进而避免客服情绪对跟催对象(也即客户)带来不良影响。本实施例中优选通过对客服端与跟催对象端之间的跟催交流内容进行情绪分析,进而获得客服的当前情绪。
本实施例对于情绪分析的实现方式不限。例如,预先设置各种情绪对应的分值,同时设置各种情绪关键词以及各情绪关键词对应的分值,然后检索获得的跟催交流内容以提取情绪关键词,最后统计所有情绪关键词的总分值,进而根据情绪关键词的总分值确定对应情绪。
步骤S60,根据客服的当前情绪,向所述客服端发送相应的情绪管理提示,以供客服调整或保持在执行跟催任务过程中的情绪。
本实施例中,在确定了客服的当前情绪后,即可对客服进行情绪管理,具体通过发送提示的方式提醒客服调整或保持在执行跟催任务过程中的情绪。比如,客服的当前情绪变差,则提示客服保持冷静并调整情绪,微笑服务;如果客服的当前情绪稳定且保持微笑,则提示客服保持当前情绪执行跟催任务。
可选的,在一实施例中,具体通过如下方式对客服进行情绪管理:
首先,调取预设的情绪值与情绪管理提示之间的对照关系表;
其次,基于客服的当前情绪,查找所述对照关系表,获得与客服的当前情绪相对应的情绪管理提示;
最后,将所述情绪管理提示发送至所述客服端,以供客服调整或保持在执行跟催任务过程中的情绪。
本可选实施例中,预先将各种情绪进行量化,然后设置情绪值与情绪管理提示之间的对照关系表,也即不同的任一种情绪都设有对应的情绪管理提示。通过查表即可与客服的当前情绪相对应的情绪管理提示。
本实施例在发起跟催任务请求时,能够使用预置的跟催任务话语模板和跟催任务策略,生成对应的跟催任务文本并发送至跟催对象,从而使跟催任务标准化,避免跟催对象产生反感。同时,进一步在跟催过程中,实时获取客服与跟催对象之间的跟催交流内容,并对跟催交流内容进行情绪分析,以确定客服的当前情绪,进而及时调整客服情绪,避免情绪化跟催而使跟催对象产生反感,由于催收过程中可进行监控与管理,因而可避免客服产生不良。
参照图3,图3为图2中步骤S50一实施例的细化流程示意图。本实施例中,上述步骤S50进一步包括:
步骤S501,记录所述客服端与所述跟催对象端之间的跟催交流内容;
步骤S502,从所述跟催交流内容中获取客服发送给跟催对象的交流内容;
本实施例中,在执行跟催任务时,客服端会与跟催对象端(也即客户端)进行通信,也即相互之间以聊天方式发送交流信息。比如通过邮件、微信等方式进行聊天,聊天内容可以是文字,也可以是语音。本实施例重点获取客服发送给跟催对象的交流内容,从而用于分析判断客服的当前情绪。
本实施例对于跟催交流内容的记录方式不限,比如通过埋点方式采集信息并进行整理,从而获得跟催交流内容。
步骤S503,若所述交流内容为文本信息,则将所述交流内容输入预置情绪分类模型进行情绪识别与分类,得到情绪分类结果;
步骤S504,若所述交流内容为语音信息,则通过语音识别将所述语音信息转换为文本信息后输入所述情绪分类模型进行情绪识别与分类,得到情绪分类结果;
本实施例中,若交流内容为文本信息,则直接将交流内容输入预置的情绪分类模型进行情绪的识别与分类,而若交流内容为语音信息,则先通过语音识别而将语音信息转换为文本信息,然后再将转换得到的文本信息输入情绪分类模型进行处理。
本实施例中,情绪分类模型的训练过程如下:先采集若干不同客服与多个不同任务对象的多条聊天记录并进行分词,然后根据每个词的词性,确定每个词的情绪特征,最后再根据每个词的情绪特征训练情绪分类器,从而构建情绪分类模型。
本实施例中,情绪分类结果中包含有同一文本信息对应的多种情绪,比如一句话中既包含有积极的情绪,也包含有消极的情绪,因此还需进一步对分类结果中的多种情绪进行回归分析以确定最符合当前客服的情绪。
步骤S505,对所述情绪分类结果进行回归分析,得到所述情绪分类结果中每种情绪的回归值;
步骤S506,根据每种情绪的回归值,计算每种情绪的得分并将得分最高的情绪作为客服的当前情绪。
回归分析是确定两种或两种以上变量间相互依赖的定量关系的一种统计分析方法。也即通过分析现象之间相关的具体形式以确定其因果关系,并用数学模型来表现其具体关系。
本实施例先通过回归分析,以得到情绪分类结果中每种情绪的回归值,然后根据每种情绪的回归值,计算每种情绪的得分。其中,对于具体计算方式不限。比如根据预置的情绪权重值来统计每种情绪的得分。
可选的,在一实施例中,采用如下公式计算每种情绪的得分:
Figure PCTCN2020111405-appb-000001
其中,T i表示表示第i种情绪的得分,V i,3、V i,4分别表示采用分层回归法获得的第i种情绪的第三层回归值与第四层回归值,i为正整数。
本可选实施例中,通过计算每种情绪的得分情况,然后将得分最高的情绪作为客服的当前情绪。
参照图4,图4为本申请基于情绪分析的任务跟催方法第二实施例的流程示意图。本实施例中,在上述步骤S60之后,还包括:
步骤S70,在跟催任务结束后,提取客服执行跟催任务过程中产生的与所述跟催对象相关的关键数据;
步骤S80,根据所述关键数据,绘制任务数据曲线,其中,所述任务数据曲线以时间为横坐标、以所述关键数据为纵坐标;
本实施例中,关键数据具体在客服执行跟催任务过程中产生,本实施例对于关键数据的具体内容及表现形式不限。例如,关键数据是有关客服跟催过程中的情绪变化、跟催对 象的情绪反应等数据。
本实施例中,为便于对下一阶段或后续的跟催任务进行预测,因此基于关键数据,绘制任务数据曲线,该任务数据曲线以时间为横坐标、以关键数据为纵坐标,可以反映将来的关键数据变化情况,比如反映出将来某个时间点客服跟催过程中的情绪变化、跟催对象的情绪反应等趋势。
步骤S90,通过预置趋势预测算法对所述任务数据曲线进行趋势预测,得到趋势预测结果;
本实施例中,在绘制出任务数据曲线后,采用预置的趋势预测算法即可进行趋势预测。需要说明的是,趋势预测主要用于预测跟催对象的情绪反应,从而便于后续调整跟催任务策略。
可选的,在一实施例中,通过以下方式进行趋势预测:
A、依次获取所述任务数据曲线中各时间点对应的相邻的关键数据对;
B、计算所有关键数据对的数据平均值,并根据所述数据平均值绘制数据平均值曲线,其中,所述数据平均值曲线以时间为横坐标、以所述数据平均值为纵坐标;
C、通过如下趋势预测算法对应的计算公式对所述数据平均值曲线进行趋势预测,得到趋势预测结果:
Y t=m*x t+(1-m)*Y t-1
其中,t表示时间,Y t表示时间t对应的预测值,x t表示时间t对应的数据平均值,Y t-1表示时间t-1对应的预测值,m常数,其取值范围为[0.5,1]。
本实施例中,为更好地拟合任务数据曲线的变化趋势,在进行趋势预测之前,先获取任务数据曲线中各时间点对应的相邻的关键数据对,并计算所有关键数据对的数据平均值,然后绘制数据平均值曲线,最后再对数据平均值曲线进行趋势预测。
步骤S100,将所述趋势预测结果导入所述跟催对象的跟催任务策略中,以作为下一次对所述跟催对象执行跟催任务时的判断依据。
本实施例中,在获得趋势预测结果后,将趋势预测结果导入跟催任务策略中,以供下次执行跟催任务时使用。
可选的,在一实施例中,具体采用如下方式实现趋势预测结果的导入:
A、将所述趋势预测结果导入所述任务数据曲线中,并根据最小二乘法生成所述任务数据曲线的趋势线;
B、判断所述趋势线的斜率是否大于预置阈值;
C、若是,则重新对所述任务数据曲线进行趋势预测;
D、若否,则将所述趋势预测结果导入所述跟催对象的跟催任务策略中,以作为下一次对所述跟催对象执行跟催任务时的判断依据。
本可选实施例中,为使得趋势预测结果更具有可参考性,因此在将趋势预测结果导入跟催任务策略中之前,进一步对趋势预测结果进行筛选,具体实现方式如下:首先将趋势预测结果导入任务数据曲线中,然后根据最小二乘法生成任务数据曲线的趋势线,最后再判断任务数据曲线的趋势线的斜率是否大于预置阈值,若是,则判定该趋势预测结果不具有可参考性,因而不能导入跟催任务策略中,否则判定该趋势预测结果具有可参考性,因而能够导入跟催任务策略中,以作为下一次对跟催对象执行跟催任务时的判断依据。
本申请还提供一种基于情绪分析的任务跟催装置。
参照图5,图5为本申请基于情绪分析的任务跟催装置一实施例的功能模块示意图。本实施例中,所述基于情绪分析的任务跟催装置包括:
获取模块10,用于获取客服端发起的跟催任务请求;
第一生成模块20,用于从所述跟催任务请求中提取任务信息,并根据所述任务信息生成跟催任务清单;
调用模块30,用于从所述跟催任务清单中提取跟催任务,并根据所述跟催任务的任务信息调用对应的跟催任务话语模板;
第二生成模块40,用于根据所述跟催任务话语模板和预置跟催任务策略,生成对应的跟催任务文本并发送至跟催对象;
情绪分析模块50,用于记录客服与所述跟催对象之间的跟催交流内容,并对所述跟催交流内容进行情绪分析,得到客服的当前情绪;
提示模块60,用于根据客服的当前情绪,向客服发送相应的情绪管理提示,以供客服调整或保持在执行跟催任务过程中的情绪。
基于与上述本申请基于情绪分析的任务跟催方法相同的实施例说明内容,因此本实施例对基于情绪分析的任务跟催装置的实施例内容不做过多赘述。
本实施例中,任务跟催装置在发起跟催任务请求时,能够使用预置的跟催任务话语模板和跟催任务策略,生成对应的跟催任务文本并发送至跟催对象,从而使跟催任务标准化,避免跟催对象产生反感。同时,进一步在跟催过程中,任务跟催装置实时获取客服与跟催对象之间的跟催交流内容,并对跟催交流内容进行情绪分析,以确定客服的当前情绪,进而及时调整客服情绪,避免情绪化跟催而使跟催对象产生反感,由于催收过程中可进行监控与管理,因而可避免客服产生不良情绪而影响跟催效果,提升了任务跟催的触达率。
本申请还提供一种基于情绪分析的任务跟催设备,包括:存储器和至少一个处理器,所述存储器中存储有指令,所述存储器和所述至少一个处理器通过线路互连;所述至少一个处理器调用所述存储器中的所述指令,以使得所述基于情绪分析的任务跟催设备执行上述基于情绪分析的任务跟催方法中的步骤。
本申请还提供一种计算机可读存储介质,该计算机可读存储介质可以为非易失性计算机可读存储介质,也可以为易失性计算机可读存储介质。计算机可读存储介质存储有计算机指令,当所述计算机指令在计算机上运行时,使得计算机执行如下步骤:
获取客服端发起的跟催任务请求;
从所述跟催任务请求中提取任务信息,并根据所述任务信息生成跟催任务清单;
从所述跟催任务清单中提取跟催任务,并根据所述跟催任务的任务信息调用对应的跟催任务话语模板;
根据所述跟催任务话语模板和预置跟催任务策略,生成对应的跟催任务文本并发送至跟催对象端;
记录所述客服端与所述跟催对象端之间的跟催交流内容,并对所述跟催交流内容进行情绪分析,得到客服的当前情绪;
根据客服的当前情绪,向所述客服端发送相应的情绪管理提示,以供客服调整或保持在执行跟催任务过程中的情绪。
通过以上的实施方式的描述,本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现,当然也可以通过硬件,但很多情况下前者是更佳的实施方式。基于这样的理解,本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来,该计算机软件产品存储在一个存储介质(如ROM/RAM)中,包括若干指令用以使得一台终端(可以是手机,计算机,服务器或者网络设备等)执行本申请各个实施例所述的方法。
上面结合附图对本申请的实施例进行了描述,但是本申请并不局限于上述的具体实施 方式,上述的具体实施方式仅仅是示意性的,而不是限制性的,本领域的普通技术人员在本申请的启示下,在不脱离本申请宗旨和权利要求所保护的范围情况下,还可做出很多形式,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,这些均属于本申请的保护之内。

Claims (20)

  1. 一种基于情绪分析的任务跟催方法,其中,包括:
    获取客服端发起的跟催任务请求;
    从所述跟催任务请求中提取任务信息,并根据所述任务信息生成跟催任务清单;
    从所述跟催任务清单中提取跟催任务,并根据所述跟催任务的任务信息调用对应的跟催任务话语模板;
    根据所述跟催任务话语模板和预置跟催任务策略,生成对应的跟催任务文本并发送至跟催对象端;
    记录所述客服端与所述跟催对象端之间的跟催交流内容,并对所述跟催交流内容进行情绪分析,得到客服的当前情绪;
    根据客服的当前情绪,向所述客服端发送相应的情绪管理提示,以供客服调整或保持在执行跟催任务过程中的情绪。
  2. 根据权利要求1所述的基于情绪分析的任务跟催方法,其中,所述记录所述客服端与所述跟催对象端之间的跟催交流内容,并对所述跟催交流内容进行情绪分析,得到客服的当前情绪包括:
    记录所述客服端与所述跟催对象端之间的跟催交流内容;
    从所述跟催交流内容中获取客服发送给跟催对象的交流内容;
    若所述交流内容为文本信息,则将所述交流内容输入预置情绪分类模型进行情绪识别与分类,得到情绪分类结果;
    若所述交流内容为语音信息,则通过语音识别将所述语音信息转换为文本信息后输入所述情绪分类模型进行情绪识别与分类,得到情绪分类结果;
    对所述情绪分类结果进行回归分析,得到所述情绪分类结果中每种情绪的回归值;
    根据每种情绪的回归值,计算每种情绪的得分并将得分最高的情绪作为客服的当前情绪。
  3. 根据权利要求2所述的基于情绪分析的任务跟催方法,其中,采用如下公式计算每种情绪的得分:
    Figure PCTCN2020111405-appb-100001
    其中,T i表示表示第i种情绪的得分,V i,3、V i,4分别表示采用分层回归法获得的第i种情绪的第三层回归值与第四层回归值,i为正整数。
  4. 根据权利要求2所述的基于情绪分析的任务跟催方法,其中,所述根据客服的当前情绪,向所述客服端发送相应的情绪管理提示,以供客服调整或保持在执行跟催任务过程中的情绪包括:
    调取预设的情绪值与情绪管理提示之间的对照关系表;
    基于客服的当前情绪,查找所述对照关系表,获得与客服的当前情绪相对应的情绪管理提示;
    将所述情绪管理提示发送至所述客服端,以供客服调整或保持在执行跟催任务过程中的情绪。
  5. 根据权利要求4所述的基于情绪分析的任务跟催方法,其中,在所述根据客服的当前情绪,向所述客服端发送相应的情绪管理提示,以供客服调整或保持在执行跟催任务过程中的情绪的步骤之后,还包括:
    在跟催任务结束后,提取客服执行跟催任务过程中产生的与所述跟催对象相关的关键数据;
    根据所述关键数据,绘制任务数据曲线,其中,所述任务数据曲线以时间为横坐标、以所述关键数据为纵坐标;
    通过预置趋势预测算法对所述任务数据曲线进行趋势预测,得到趋势预测结果;
    将所述趋势预测结果导入所述跟催对象的跟催任务策略中,以作为下一次对所述跟催对象执行跟催任务时的判断依据。
  6. 根据权利要求4所述的基于情绪分析的任务跟催方法,其中,所述通过预置趋势预测算法对所述任务数据曲线进行趋势预测,得到趋势预测结果包括:
    依次获取所述任务数据曲线中各时间点对应的相邻的关键数据对;
    计算所有关键数据对的数据平均值,并根据所述数据平均值绘制数据平均值曲线,其中,所述数据平均值曲线以时间为横坐标、以所述数据平均值为纵坐标;
    通过如下趋势预测算法对应的计算公式对所述数据平均值曲线进行趋势预测,得到趋势预测结果:
    Y t=m*x t+(1-m)*Y t-1
    其中,t表示时间,Y t表示时间t对应的预测值,x t表示时间t对应的数据平均值,Y t-1表示时间t-1对应的预测值,m常数,其取值范围为[0.5,1]。
  7. 根据权利要求6所述的基于情绪分析的任务跟催方法,其中,所述将所述趋势预测结果导入所述跟催对象的跟催任务策略中,以作为下一次对所述跟催对象执行跟催任务时的判断依据包括:
    将所述趋势预测结果导入所述任务数据曲线中,并根据最小二乘法生成所述任务数据曲线的趋势线;
    判断所述趋势线的斜率是否大于预置阈值;
    若是,则重新对所述任务数据曲线进行趋势预测;
    若否,则将所述趋势预测结果导入所述跟催对象的跟催任务策略中,以作为下一次对所述跟催对象执行跟催任务时的判断依据。
  8. 一种基于情绪分析的任务跟催设备,包括存储器、处理器及存储在所述存储器上并可在所述处理器上运行的计算机可读指令,所述处理器执行所述计算机可读指令时实现如下步骤:
    获取客服端发起的跟催任务请求;
    从所述跟催任务请求中提取任务信息,并根据所述任务信息生成跟催任务清单;
    从所述跟催任务清单中提取跟催任务,并根据所述跟催任务的任务信息调用对应的跟催任务话语模板;
    根据所述跟催任务话语模板和预置跟催任务策略,生成对应的跟催任务文本并发送至跟催对象端;
    记录所述客服端与所述跟催对象端之间的跟催交流内容,并对所述跟催交流内容进行情绪分析,得到客服的当前情绪;
    根据客服的当前情绪,向所述客服端发送相应的情绪管理提示,以供客服调整或保持在执行跟催任务过程中的情绪。
  9. 根据权利要求8所述的基于情绪分析的任务跟催设备,所述处理器执行所述计算机程序时还实现以下步骤:
    记录所述客服端与所述跟催对象端之间的跟催交流内容;
    从所述跟催交流内容中获取客服发送给跟催对象的交流内容;
    若所述交流内容为文本信息,则将所述交流内容输入预置情绪分类模型进行情绪识别 与分类,得到情绪分类结果;
    若所述交流内容为语音信息,则通过语音识别将所述语音信息转换为文本信息后输入所述情绪分类模型进行情绪识别与分类,得到情绪分类结果;
    对所述情绪分类结果进行回归分析,得到所述情绪分类结果中每种情绪的回归值;
    根据每种情绪的回归值,计算每种情绪的得分并将得分最高的情绪作为客服的当前情绪。
  10. 根据权利要求9所述的基于情绪分析的任务跟催设备,所述处理器执行所述计算机程序时还实现以下步骤:
    记录所述客服端与所述跟催对象端之间的跟催交流内容;
    从所述跟催交流内容中获取客服发送给跟催对象的交流内容;
    若所述交流内容为文本信息,则将所述交流内容输入预置情绪分类模型进行情绪识别与分类,得到情绪分类结果;
    若所述交流内容为语音信息,则通过语音识别将所述语音信息转换为文本信息后输入所述情绪分类模型进行情绪识别与分类,得到情绪分类结果;
    对所述情绪分类结果进行回归分析,得到所述情绪分类结果中每种情绪的回归值;
    根据每种情绪的回归值,计算每种情绪的得分并将得分最高的情绪作为客服的当前情绪。
  11. 根据权利要求9所述的基于情绪分析的任务跟催设备,所述处理器执行所述计算机程序时还实现以下步骤:
    调取预设的情绪值与情绪管理提示之间的对照关系表;
    基于客服的当前情绪,查找所述对照关系表,获得与客服的当前情绪相对应的情绪管理提示;
    将所述情绪管理提示发送至所述客服端,以供客服调整或保持在执行跟催任务过程中的情绪。
  12. 根据权利要求11所述的基于情绪分析的任务跟催设备,所述处理器执行所述计算机程序时还实现以下步骤:
    在跟催任务结束后,提取客服执行跟催任务过程中产生的与所述跟催对象相关的关键数据;
    根据所述关键数据,绘制任务数据曲线,其中,所述任务数据曲线以时间为横坐标、以所述关键数据为纵坐标;
    通过预置趋势预测算法对所述任务数据曲线进行趋势预测,得到趋势预测结果;
    将所述趋势预测结果导入所述跟催对象的跟催任务策略中,以作为下一次对所述跟催对象执行跟催任务时的判断依据。
  13. 根据权利要求11所述的基于情绪分析的任务跟催设备,所述处理器执行所述计算机程序时还实现以下步骤:
    依次获取所述任务数据曲线中各时间点对应的相邻的关键数据对;
    计算所有关键数据对的数据平均值,并根据所述数据平均值绘制数据平均值曲线,其中,所述数据平均值曲线以时间为横坐标、以所述数据平均值为纵坐标;
    通过如下趋势预测算法对应的计算公式对所述数据平均值曲线进行趋势预测,得到趋势预测结果:
    Y t=m*x t+(1-m)*Y t-1
    其中,t表示时间,Y t表示时间t对应的预测值,x t表示时间t对应的数据平均值,Y t-1 表示时间t-1对应的预测值,m常数,其取值范围为[0.5,1]。
  14. 根据权利要求13所述的基于情绪分析的任务跟催设备,所述处理器执行所述计算机程序时还实现以下步骤:
    将所述趋势预测结果导入所述任务数据曲线中,并根据最小二乘法生成所述任务数据曲线的趋势线;
    判断所述趋势线的斜率是否大于预置阈值;
    若是,则重新对所述任务数据曲线进行趋势预测;
    若否,则将所述趋势预测结果导入所述跟催对象的跟催任务策略中,以作为下一次对所述跟催对象执行跟催任务时的判断依据。
  15. 一种计算机可读存储介质,所述计算机可读存储介质中存储计算机指令,当所述计算机指令在计算机上运行时,使得计算机执行如下步骤:
    获取客服端发起的跟催任务请求;
    从所述跟催任务请求中提取任务信息,并根据所述任务信息生成跟催任务清单;
    从所述跟催任务清单中提取跟催任务,并根据所述跟催任务的任务信息调用对应的跟催任务话语模板;
    根据所述跟催任务话语模板和预置跟催任务策略,生成对应的跟催任务文本并发送至跟催对象端;
    记录所述客服端与所述跟催对象端之间的跟催交流内容,并对所述跟催交流内容进行情绪分析,得到客服的当前情绪;
    根据客服的当前情绪,向所述客服端发送相应的情绪管理提示,以供客服调整或保持在执行跟催任务过程中的情绪。
  16. 根据权利要求15所述的计算机可读存储介质,当所述计算机指令在计算机上运行时,使得计算机还执行以下步骤:
    记录所述客服端与所述跟催对象端之间的跟催交流内容;
    从所述跟催交流内容中获取客服发送给跟催对象的交流内容;
    若所述交流内容为文本信息,则将所述交流内容输入预置情绪分类模型进行情绪识别与分类,得到情绪分类结果;
    若所述交流内容为语音信息,则通过语音识别将所述语音信息转换为文本信息后输入所述情绪分类模型进行情绪识别与分类,得到情绪分类结果;
    对所述情绪分类结果进行回归分析,得到所述情绪分类结果中每种情绪的回归值;
    根据每种情绪的回归值,计算每种情绪的得分并将得分最高的情绪作为客服的当前情绪。
  17. 根据权利要求16所述的计算机可读存储介质,当所述计算机指令在计算机上运行时,使得计算机还执行以下步骤:
    采用如下公式计算每种情绪的得分:
    Figure PCTCN2020111405-appb-100002
    其中,T i表示表示第i种情绪的得分,V i,3、V i,4分别表示采用分层回归法获得的第i种情绪的第三层回归值与第四层回归值,i为正整数。
  18. 根据权利要求16所述的计算机可读存储介质,当所述计算机指令在计算机上运行时,使得计算机还执行以下步骤:
    调取预设的情绪值与情绪管理提示之间的对照关系表;
    基于客服的当前情绪,查找所述对照关系表,获得与客服的当前情绪相对应的情绪管 理提示;
    将所述情绪管理提示发送至所述客服端,以供客服调整或保持在执行跟催任务过程中的情绪。
  19. 根据权利要求18所述的计算机可读存储介质,当所述计算机指令在计算机上运行时,使得计算机还执行以下步骤:
    在跟催任务结束后,提取客服执行跟催任务过程中产生的与所述跟催对象相关的关键数据;
    根据所述关键数据,绘制任务数据曲线,其中,所述任务数据曲线以时间为横坐标、以所述关键数据为纵坐标;
    通过预置趋势预测算法对所述任务数据曲线进行趋势预测,得到趋势预测结果;
    将所述趋势预测结果导入所述跟催对象的跟催任务策略中,以作为下一次对所述跟催对象执行跟催任务时的判断依据。
  20. 一种基于情绪分析的任务跟催装置,其中,所述基于情绪分析的任务跟催包括:
    获取模块,用于获取客服端发起的跟催任务请求;
    第一生成模块,用于从所述跟催任务请求中提取任务信息,并根据所述任务信息生成跟催任务清单;
    调用模块,用于从所述跟催任务清单中提取跟催任务,并根据所述跟催任务的任务信息调用对应的跟催任务话语模板;
    第二生成模块,用于根据所述跟催任务话语模板和预置跟催任务策略,生成对应的跟催任务文本并发送至跟催对象端;
    情绪分析模块,用于记录所述客服端与所述跟催对象端之间的跟催交流内容,并对所述跟催交流内容进行情绪分析,得到客服的当前情绪;
    提示模块,用于根据客服的当前情绪,向所述客服端发送相应的情绪管理提示,以供客服调整或保持在执行跟催任务过程中的情绪。
PCT/CN2020/111405 2019-10-18 2020-08-26 基于情绪分析的任务跟催方法、装置、设备及存储介质 WO2021073258A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910990755.3 2019-10-18
CN201910990755.3A CN110929005A (zh) 2019-10-18 2019-10-18 基于情绪分析的任务跟催方法、装置、设备及存储介质

Publications (1)

Publication Number Publication Date
WO2021073258A1 true WO2021073258A1 (zh) 2021-04-22

Family

ID=69849016

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/111405 WO2021073258A1 (zh) 2019-10-18 2020-08-26 基于情绪分析的任务跟催方法、装置、设备及存储介质

Country Status (2)

Country Link
CN (1) CN110929005A (zh)
WO (1) WO2021073258A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110929005A (zh) * 2019-10-18 2020-03-27 平安科技(深圳)有限公司 基于情绪分析的任务跟催方法、装置、设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20190018666A (ko) * 2019-02-18 2019-02-25 네이버 주식회사 기계의 자동 활성을 위한 방법 및 시스템
CN109767787A (zh) * 2019-01-28 2019-05-17 腾讯科技(深圳)有限公司 情绪识别方法、设备及可读存储介质
CN109784414A (zh) * 2019-01-24 2019-05-21 出门问问信息科技有限公司 一种电话客服中客户情绪检测方法、装置及电子设备
CN109949103A (zh) * 2019-03-29 2019-06-28 联想(北京)有限公司 一种数据处理方法、装置及电子设备
CN110929005A (zh) * 2019-10-18 2020-03-27 平安科技(深圳)有限公司 基于情绪分析的任务跟催方法、装置、设备及存储介质

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103970864B (zh) * 2014-05-08 2017-09-22 清华大学 基于微博文本的情绪分类和情绪成分分析方法及系统
CN108427916A (zh) * 2018-02-11 2018-08-21 上海复旦通讯股份有限公司 一种用于客服坐席情绪的监控系统及监控方法
CN109064315A (zh) * 2018-08-02 2018-12-21 平安科技(深圳)有限公司 逾期账单智能催收方法、装置、计算机设备及存储介质
CN109767765A (zh) * 2019-01-17 2019-05-17 平安科技(深圳)有限公司 话术匹配方法及装置、存储介质、计算机设备
CN109859032A (zh) * 2019-01-22 2019-06-07 深圳壹账通智能科技有限公司 基于智能语音的账款催收方法、装置、设备和存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109784414A (zh) * 2019-01-24 2019-05-21 出门问问信息科技有限公司 一种电话客服中客户情绪检测方法、装置及电子设备
CN109767787A (zh) * 2019-01-28 2019-05-17 腾讯科技(深圳)有限公司 情绪识别方法、设备及可读存储介质
KR20190018666A (ko) * 2019-02-18 2019-02-25 네이버 주식회사 기계의 자동 활성을 위한 방법 및 시스템
CN109949103A (zh) * 2019-03-29 2019-06-28 联想(北京)有限公司 一种数据处理方法、装置及电子设备
CN110929005A (zh) * 2019-10-18 2020-03-27 平安科技(深圳)有限公司 基于情绪分析的任务跟催方法、装置、设备及存储介质

Also Published As

Publication number Publication date
CN110929005A (zh) 2020-03-27

Similar Documents

Publication Publication Date Title
US11004013B2 (en) Training of chatbots from corpus of human-to-human chats
CN107818798B (zh) 客服服务质量评价方法、装置、设备及存储介质
CN108962282B (zh) 语音检测分析方法、装置、计算机设备及存储介质
CN110619506B (zh) 一种岗位画像生成方法、岗位画像生成装置及电子设备
US20160117624A1 (en) Intelligent meeting enhancement system
CN108416375B (zh) 工单分类方法及装置
US20170068976A1 (en) System using automatically triggered analytics for feedback data
US20190266254A1 (en) System to process electronic records using a request orchestration platform
CN113468296B (zh) 可配置业务逻辑的模型自迭代式智能客服质检系统与方法
EP3685245B1 (en) Method, apparatus, and computer-readable media for customer interaction semantic annotation and analytics
CN109710766B (zh) 一种工单数据的投诉倾向分析预警方法及装置
CN110475032A (zh) 多业务界面切换方法、装置、计算机装置及存储介质
CN110516057A (zh) 一种信访问题答复方法及装置
CN113297365B (zh) 一种用户意向判定方法、装置、设备及存储介质
WO2019214046A1 (zh) 资产走势分析的方法、装置、计算机设备和存储介质
WO2022267322A1 (zh) 会议纪要的生成方法、装置、终端设备及计算机存储介质
WO2021073258A1 (zh) 基于情绪分析的任务跟催方法、装置、设备及存储介质
CN111061948A (zh) 一种用户标签推荐方法、装置、计算机设备及存储介质
CN110931002B (zh) 人机交互方法、装置、计算机设备和存储介质
CN113111157B (zh) 问答处理方法、装置、计算机设备和存储介质
KR20220099690A (ko) 문서를 요약하는 장치, 방법 및 컴퓨터 프로그램
Bianchi et al. A machine learning based help desk approach for units involved in official surveys
US20230067687A1 (en) System and method and apparatus for integrating conversational signals into a dialog
CN111666469B (zh) 语句库构建方法、装置、设备和存储介质
US11693911B2 (en) System and method for managing relationships by identifying relevant content and generating correspondence based thereon

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: 20876911

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: 20876911

Country of ref document: EP

Kind code of ref document: A1