TWM615272U - Smart voice customer service system - Google Patents

Abstract

A smart voice customer service system, including: a voice guidance module; a voice recognition and text conversion module; a semantic content analysis module; a customer response module; a response recording module; an abnormal screening and judging module ; An optimized training module, and a terminal control module, the terminal control module is respectively connected with the voice guidance module, the voice recognition and text conversion module, the semantic content analysis module, the customer response module, The response recording module, the abnormal screening judgment module and the optimization training module are electrically connected to control the execution procedures between the modules. This creation uses the voice recognition and text conversion module to receive the voice signal from the client and converts the voice recognition into a text record file, and then uses the client response module to search for the corresponding content from a dialogue response database to respond to the client, and Using artificial intelligence to optimize the content of customer responses can provide an automated voice customer service with a simulated human response that is accurate and efficient.

Description

Intelligent Voice Customer Service System

This creation is about a voice customer service system, especially a smart voice customer service system that can automatically recognize the semantics of the client and optimize the content of the response to the customer's conversation.

In the conventional telephone customer service system, most respondents press the button of the transaction code they want to inquire according to the voice guidance method. However, in fact, the inquiries of the customer service staff or the questions that the customer wants to know are often complex or multiple. And the respondent’s answer is limited by the button, and cannot give feedback other than the option. Therefore, in the end, most of the customer service staff or transfer to the customer service staff provide an immediate dialogue to explain or answer the content that both parties want to convey.

In order to solve the aforementioned problems, conventional technologies such as the Republic of China Announcement No. M596944 New Patent discloses an intelligent customer service system that can simulate customer service staff to respond to the customer's questions by voice around the clock.

However, the aforementioned conventional technology does not have the function of judging the abnormal result of the client's response, and the function of automatically optimizing and updating the response content. If the conventional technology wants to improve the accuracy of the answer, it needs to continuously increase the answering question library, which increases the complexity and storage space. And the implementation efficiency is low, in view of this, it is necessary to further improve.

One purpose of this creation is to provide an intelligent voice customer service system with the function of judging the client's response to abnormal results.

Another purpose of this creation is to provide a smart voice customer service system with the function of automatically optimizing and updating response content.

In order to achieve the above-mentioned purpose, an intelligent voice customer service system created by this invention includes: a voice guidance module for telephone dialing and digital voice dialogue guidance; a voice recognition and text conversion module for receiving clients The voice signal is converted into a text record file through voice recognition; a semantic content analysis module is electrically connected with the voice recognition and text conversion module to read and analyze the semantics and content of the text record file; a customer response The module includes a dialogue response database for searching the corresponding content response client from the dialogue response database based on the semantic content of the text record file; a response record module includes a dialogue record database for The text log file is stored in the dialogue record database; an abnormal screening judgment module is used to read the text log file and perform abnormal result judgment. The abnormal screening judgment module marks the abnormal result to provide follow-up monitoring personnel Make adjustments to reply to the reference of the customer dialogue, and update the text log file adjusted by the monitoring personnel into an abnormal adjustment log file and store it in the dialogue log database; an optimization training module, by importing the abnormal adjustment log file as Training samples, optimized by artificial intelligence algorithms, as the voice guidance module, the voice recognition and text conversion module, the semantic content analysis module, and the customer response module for the next batch of voice customer service process Reference model of time; and a terminal control module, the terminal control module and the voice guidance module, the voice recognition and text conversion module, the semantic content analysis module, the customer response module, the response The recording module, the abnormal screening and judging module, and the optimization training module are electrically connected to control the execution program between the modules.

In some embodiments of this creation, the voice guidance module has multiple language selection functions.

In some embodiments of the present creation, the voice recognition and text conversion module has multiple language and text selection functions.

In some embodiments of the present creation, the speech recognition and text conversion module has a dialogue situation classification function.

In some embodiments of the present creation, the semantic content analysis module has the function of setting keywords.

In some embodiments of the present creation, when the call is interrupted, the response recording module records the current status and the customer's phone number, and the terminal control module activates the voice guidance module to make a call after a certain time.

In some embodiments of the present creation, the abnormal screening and judging module has the function of setting multiple abnormal keywords.

The examples are only used to illustrate the possible implementation of this creation, but they are not intended to limit the scope of protection of this creation. In order to make the purpose, features, and advantages of this creation obvious and easy to understand, the following will give a detailed description of the preferred embodiments of this creation in conjunction with the accompanying drawings.

FIG. 1 is a system architecture diagram of an embodiment of the creative intelligent voice customer service system, and FIG. 2 is a schematic diagram of the system execution flow of an embodiment of the creative intelligent voice customer service system. Please refer to FIGS. 1 and 2. In one embodiment, the presently created intelligent voice customer service system includes: a voice guidance module 10, a voice recognition and text conversion module 20, a semantic content analysis module 30, a customer response module 40, a The response recording module 50, an abnormal screening and judging module 60, an optimization training module 70 and a terminal control module 80, the terminal control module 80 and the voice guidance module 10, the voice recognition and text conversion respectively The module 20, the semantic content analysis module 30, the customer response module 40, the response record module 50, the abnormal screening judgment module 60, and the optimization training module are electrically connected to 70 for controlling each module Execution procedures between groups.

The voice guidance module 10 is used for telephone dialing and digital voice dialogue guidance. When a customer satisfaction survey is about to be dialed to a certain customer, the terminal control module 80 activates the voice guidance module 10 to start. The customer data search is done by setting up the phone number of the customer in advance and then dialing, and then entering the digital voice dialogue of the realistic robot to guide the customer to conduct a satisfaction survey dialogue.

Preferably, the voice guidance module 10 has a multi-language selection function, which can improve the smoothness and efficiency of the customer satisfaction survey process by first understanding the customer’s habitual languages, such as Chinese, English, and Taiwanese.

The voice recognition and text conversion module 20 is used to receive the voice signal of the client and perform voice recognition to convert it into a text record file. When the voice guidance module 10 communicates with the client, the voice recognition and text conversion module 20 starts and begins to recognize and convert the voice sent by the client into text data. In order to clearly confirm the content of each conversation, the voice recognition and text conversion module 20 simultaneously converts the conversation between the client and the client into text data. The pattern is saved as the text log file to facilitate subsequent semantic analysis.

Preferably, the speech recognition and text conversion module 20 has a multi-language text selection function, can convert speech into Chinese, English characters, Japanese, etc., and can be applied to a larger client service range.

Preferably, the voice recognition and text conversion module 20 has a conversational context classification function, which uses voice recognition to understand the customer's semantics or analyze the current background sound to determine the current situation of the customer in response to the customer's inquiry, such as: During the customer satisfaction survey, the customer answered "I can’t hear you clearly" and the background sound noise is too high, indicating that the customer can be outdoors. Therefore, the voice guidance module 10 can turn up the volume or respond with "Sorry , Is it convenient to call you back later?” In this way, invalid or inefficient satisfaction survey reports can be avoided.

The semantic content analysis module 30 is electrically connected to the voice recognition and text conversion module 20 to read and analyze the semantics and content of the text record file. When the customer’s voice signal is recognized and converted into text mode, it will immediately pass The semantic content analysis module 30 analyzes the current semantic content of the customer, such as: "I am not available", "busy", "inconvenient" and "who are you"... etc. to facilitate subsequent access to the terminal control module 80 The following control program and how the voice guidance module 10 guides the client to follow up questions.

Preferably, the semantic content analysis module 30 has the function of setting key words. By setting some key words in advance, the analysis efficiency of the semantic content analysis module 30 can be improved. Some unconventional but still part The vocabulary that people will use is set in advance to quickly judge and summarize the semantics of the client, and improve the efficiency of customer service. For example, "warranty", "oil change", "into the factory" and "maintenance" all belong to the meaning of "maintenance".

The customer response module 40 includes a dialogue response database 41 for searching the corresponding content response client from the dialogue response database 41 according to the semantic content of the text record file. Because the content or meaning of the client's response is different every time Different, or the same meaning but using different vocabulary to answer, at this time, the customer response module 40 must determine what the semantic meaning of the customer’s answer means, and the customer response module 40 uses artificial intelligence to determine and Search for appropriate corresponding content from the dialogue response database 41 to respond to the client, such as: "Okay, I will disturb you, and I will call you another time", "Sorry, I am a voice robot, I will help you transfer customer service Personnel"...etc.

The response record module 50 includes a dialogue record database 51 for storing the text record file in the dialogue record database 51 for subsequent content analysis of the text record file.

The abnormal screening and judging module 60 is used to read the text log file and perform screening and judging of abnormal results. The abnormal results referred to here are expected conditions that are not set by the system, such as customer satisfaction is not the full score, the best condition, or The abnormal conditions of the system's initial settings can be set according to actual needs. For example: the lowest to highest satisfaction is 1~10 points. If the customer’s satisfaction answer is not "10 points" or there is a customer complaint, then read the text log file to determine the reason, and the client’s response is The content of the words and sentences will vary from person to person. Sometimes the semantic content analysis module 30 may not be able to interpret or the satisfaction result does not match the actual determination. At this time, it is necessary to make corrections and adjustments manually. The response content under these unexpected conditions is first reviewed and corrected by the monitoring personnel and correctly summarized. The abnormal screening judgment module 60 marks the abnormal results to provide the monitoring personnel for interpretation and correction, and The text log file adjusted and corrected by the monitoring personnel is updated into an abnormal adjustment log file and stored in the dialogue record database 51 as a sample of artificial intelligence optimization training to facilitate subsequent responses to customer dialogues and a basis for judgment of satisfaction.

The optimization training module 70 can be used as the voice guidance module 10, the voice recognition, and The reference model of the text conversion module 20, the semantic content analysis module 30 and the customer response module 40 in the next batch of voice customer service process can be more accurate and more realistic.

For example, the optimization training module 70 stores the optimized reference data in the dialog response database 41 as a reference for the client response module 40 to respond to the client dialog next time. In order to make the voice system of the realistic customer service robot closer to receiving and responding to the response content of the client, the optimization training module 70 optimizes the response content in the dialogue response database 41 through each abnormal adjustment record file, so that the The optimized training module 70 learns the content of various client responses, so that the response content in the dialogue response database 41 can be more in line with the usage habits of most clients, improve the efficiency of voice customer service and reduce the use of actual human resources.

Preferably, when the call is interrupted, the response recording module 50 records the current status and the customer's phone number, and the terminal control module 80 activates the voice guidance module 10 to make a call after a certain time. In order to allow customers to have the actual feeling of being respected, the response recording module 50 records along with the voice guidance module 10. When performing customer service or satisfaction surveys, no matter what factors cause the call to be interrupted, the response recording module 50 The module 50 records the current status and automatically sets the callback time, for example: after 30 seconds or one minute, after the callback is connected to the client, it will continue, instead of starting from the beginning, in order to save time and improve efficiency.

Preferably, the abnormality screening and judging module 60 has the function of setting multiple abnormal keywords. In order to increase the accuracy of the screening and judging of abnormal results, the abnormality screening and judging module 60 can set multiple sets of interrelated keywords at the same time. To filter and compare the keyword combinations above the group, it is easier to find abnormal results. For example, the customer service terminal first asks "Is your satisfaction with the company 10 points", the client answers "Yes", but finally the customer service terminal asks " Is there a need for customer complaints", the client replied "Yes", this conflicting result is one of abnormal conditions, or the customer service side asks "The satisfaction of our company is 1~10 points, how would you rate it?" ", and the client answers "no question." This kind of answer that is not in accordance with the expected answer will also be prone to misjudgment. Therefore, the above-mentioned situations need to be interpreted and corrected by the monitoring personnel, and the actual situation can be judged by reading the text record file. After correction and adjustment, the optimization training module 70 is used to train and optimize to reduce the same or similar misjudgments in the future. When the situation occurs again, or when the aforementioned situation occurs, the voice guidance module 10 can re-inquire and confirm whether the customer's previous answer is wrong, and can increase the effective survey samples.

Through the intelligent voice customer service system created by this creation, a more comprehensive, self-learning and optimized voice customer service system can be provided for the robot voice customer service system. Through artificial intelligence learning, each conversation between the customer service terminal and the client is analyzed and carried out. Optimization, as a reference for the next implementation of voice customer service, can improve the simulation, execution efficiency, accuracy and reduce the human burden of actual customer service personnel.

The above-mentioned embodiments are only to illustrate the technical ideas and features of this creation, and their purpose is to enable those who are familiar with the art to understand the content of this creation and implement them accordingly. However, the patent scope of this creation cannot be limited by this. , All equal changes or modifications made in accordance with the spirit of this creation and the contents of the manual shall be covered by the scope of this creation patent.

10: Voice guidance module 20: Voice recognition and text conversion module 30: Semantic content analysis module 40: Customer Response Module 41: Dialogue Response Database 50: Response record module 51: Dialogue Record Database 60: Abnormal screening judgment module 70: Optimize the training module 80: Terminal control module

Figure 1 is a system architecture diagram of an embodiment of the creative intelligent voice customer service system; Fig. 2 is a schematic diagram of the system execution flow of an embodiment of the creative intelligent voice customer service system.

10: Voice guidance module

20: Voice recognition and text conversion module

30: Semantic content analysis module

40: Customer Response Module

41: Dialogue Response Database

50: Response record module

51: Dialogue Record Database

60: Abnormal screening judgment module

70: Optimize the training module

80: Terminal control module

Claims (7)

A smart voice customer service system, including: a voice guidance module for telephone dialing and digital voice dialogue guidance; a voice recognition and text conversion module for receiving voice signals from the client and performing voice recognition conversion A text log file; a semantic content analysis module electrically connected with the speech recognition and text conversion module for reading and analyzing the semantics and content of the text log file; a customer response module containing a dialogue response data The database is used to search for the corresponding content response client from the dialogue response database based on the semantic content of the text record file; a response record module includes a dialogue record database for storing the text record file in the Dialogue record database; an abnormal screening and judging module for reading the text record file and judging the abnormal result. The abnormal screening and judging module marks the abnormal result to provide the follow-up monitoring staff with a reference for adjusting and replying to the customer dialogue. The text log file adjusted by the monitoring personnel is updated into an abnormal adjustment log file and stored in the dialogue record database; an optimized training module, by importing the abnormal adjustment log file as a training sample, and calculating it through artificial intelligence The method is optimized to serve as a reference model for the voice guidance module, the voice recognition and text conversion module, the semantic content analysis module, and the customer response module in the next batch of voice customer service processes; and a terminal Control module, the terminal control module is respectively connected with the voice guidance module, the voice recognition and text conversion module, the semantic content analysis module, the customer response module, the response record module, and the abnormal screening judgment The module and the optimized training module are electrically connected to control the execution program between the modules. For example, the intelligent voice customer service system described in item 1 of the scope of patent application, wherein the voice guidance module has a multi-language selection function. For example, the intelligent voice customer service system described in item 1 of the scope of patent application, wherein the voice recognition and text conversion module has multiple language and text selection functions. For example, the intelligent voice customer service system described in item 1 of the scope of patent application, wherein the voice recognition and text conversion module has a dialogue situation classification function. For example, the intelligent voice customer service system described in item 1 of the scope of patent application, wherein the semantic content analysis module has the function of setting keywords. For example, the intelligent voice customer service system described in item 1 of the scope of patent application, wherein when the call is interrupted, the response recording module records the current status and the customer phone number, and the terminal control module activates the voice guidance after a certain time Lead the module for telephone dialing. For example, in the intelligent voice customer service system described in item 1 of the scope of patent application, the abnormality screening and judging module has the function of setting multiple abnormal keywords.

Images