AI voice call method and system, program product and storage medium

Abstract

The invention discloses an AI voice call method and system, a program product and a storage medium. In the method, after a user starts an outbound task, customer data is collected in real time to screen out-calling customers and select a proper robot to call; and after the call is started, judging whether the customer is interested in according to the voice characteristics and voice content of the customer, and if the customer is not interested in, interrupting the call in time and calling the next customer. By implementing the technical scheme provided by the invention, the efficiency of executing the outbound task by the AI voice robot is improved.

Description

Technical Field

[0001] The present application relates to the field of mobile voice services, and in particular to an AI voice calling method and an AI voice calling system. Background Art

[0002] With the rapid development of technology, enterprises have a growing demand for efficient marketing and customer service, and AI voice calling technology has emerged. Enterprises usually use it for large-scale outbound calls, set specific scripts and task processes, promote products to customers, etc.

[0003] At present, AI voice calling technology usually sets specific calling tasks and target group characteristics first, and then selects customer numbers that may meet the target from the database. Then, the system will automatically make outbound calls to these numbers through robots according to the preset words and processes. After the customer answers the call, the AI ​​voice robot communicates according to the preset words, uses speech recognition and natural language processing technology to understand the customer's intentions and record data.

[0004] However, due to inaccurate data or untimely updates, some non-target customers are often mistakenly identified as target customers, resulting in wrong calls to non-target customers and failure to proactively interrupt the wrong calls in a timely manner after the wrong calls are made, reducing the efficiency of AI voice robots in performing outbound calling tasks. Summary of the invention

[0005] The present application provides an AI voice calling method and an AI voice calling system for improving the efficiency of an AI voice robot in performing outbound calling tasks.

[0006] In the first aspect, the present application provides an AI voice call method, which is applied to an AI voice call system, and the method includes: after responding to the user's operation of starting a task, collecting the regional data, historical call data, real-time consumption behavior value and real-time browsing record value of each customer in the task in real time, calculating the sum of the real-time consumption behavior value and the real-time browsing value, and then calculating the ratio of the sum to the historical call data as the correlation parameter with the target customer; selecting a robot according to the number and regional data of outbound customers, and calling the outbound customers in turn; the outbound customer is a customer whose correlation parameter exceeds a preset correlation threshold; after the call starts, extracting the voice features of the outbound customer during the call in real time, using voice recognition technology to convert the outbound customer's voice into text, and collecting the real-time speech speed data, real-time intonation data, real-time volume data and Real-time pause frequency data, the number of occurrences of preset keywords of interest and the number of occurrences of preset keywords of no interest in the text; the real-time speech speed data, the real-time tone data, the real-time volume data, the real-time pause frequency data and the number of occurrences of preset keywords of interest are input into a preset dynamic interest value evaluation model to obtain the dynamic interest value of the outbound customer; the preset dynamic interest value evaluation model is a preset dynamic interest value evaluation model with different weights, and the real-time speech speed data, the real-time tone data, the real-time volume data, the real-time pause frequency data and the average of the number of occurrences of preset keywords of interest are calculated; when the duration of the outbound customer in the uninterested state exceeds the preset uninterested duration or the number of occurrences of the preset uninterested keyword exceeds the preset uninterested keyword number threshold, the current call is terminated and the next outbound customer is called; the uninterested state is that the dynamic interest value of the outbound customer is lower than the preset interest threshold.

[0007] In the above embodiment, by screening target customers for calling, customers with high correlation can be screened more accurately for outbound calls, thereby improving outbound call efficiency; during the call, it is determined whether the outbound user is interested. If not, the current call is terminated and the next outbound customer is called. By dynamically evaluating the customer's interest value, calls to uninterested customers are terminated in time, saving time and resources and improving overall outbound call efficiency.

[0008] In combination with some embodiments of the first aspect, in some embodiments, a robot is selected based on the number of outbound customers and regional data, and calls are made to the outbound customers in turn, specifically including: selecting a robot based on the number of outbound customers and regional data, and determining whether the robot is currently idle; if it is not currently idle, waiting for the robot to become idle, and then selecting the robot again; if it is currently idle, calling the outbound customers in turn.

[0009] In the above embodiment, by selecting a robot based on the number and regional data of outbound customers, a suitable robot can be reasonably selected according to the specific situation, ensuring that the robot used in the outbound call task matches the customer's needs, thereby improving the efficiency and accuracy of the outbound call.

[0010] In combination with some embodiments of the first aspect, in some embodiments, after the call starts, the voice features of the outbound customer during the call are extracted in real time, the voice of the outbound customer is converted into text using voice recognition technology, and the voice features and real-time speech speed data, real-time intonation data, real-time volume data, real-time pause frequency data, the number of occurrences of preset keywords of interest and the number of occurrences of preset keywords of no interest in the text are collected in real time, specifically including: after the call starts, start recording, and detect whether the audio quality meets the requirements; if the audio quality does not meet the requirements, pause the recording, adjust the recording parameters and then detect again whether the audio quality meets the requirements; if the audio quality meets the requirements, extract the voice features of the outbound customer during the call in real time, and convert the voice of the outbound customer into text using voice recognition technology; if the text conversion cannot be converted, automatically switch to other voice recognition engines for attempts, and publish prompt information on the system homepage at the same time, the prompt information is used to prompt manual review and correction of abnormal voice; collect voice features and real-time speech speed data, real-time intonation data, real-time volume data, real-time pause frequency data, the number of occurrences of preset keywords of interest and the number of occurrences of preset keywords of no interest in the text.

[0011] In the above embodiment, the accuracy and reliability of speech recognition are improved by starting recording and detecting the audio quality after the call begins. In the event of problems in the speech-to-text conversion process, other speech recognition engines are automatically switched to try, and prompt information is issued at the same time to allow manual review and correction of abnormal speech. When problems occur, the engine can be automatically switched and corrected with the help of manual review, thereby ensuring the integrity of data collection and providing an accurate data basis for subsequent analysis and processing, thereby improving the quality and efficiency of outbound call services.

[0012] In combination with some embodiments of the first aspect, in some embodiments, after selecting a robot based on the number of outbound customers and regional data, and calling the outbound customers in turn, it also includes: during the call, using voice recognition technology to convert the outbound customer's voice into text, and monitoring whether there is a trigger for the preset conversion to manual keyword in the text; when it is monitored whether there is a trigger for the preset conversion to manual keyword in the text, the outbound user with the trigger for the preset conversion to manual keyword is transferred to a manual agent, the robot terminates the current call, and calls the next outbound customer.

[0013] In the above embodiments, on the one hand, the pertinence and efficiency of customer service are improved, and when customers have specific needs, they can be transferred to human agents in a timely manner for more in-depth communication and problem solving.

[0014] In combination with some embodiments of the first aspect, in some embodiments, when it is monitored whether there are keywords in the text that trigger preset transfer to manual service, after transferring to a manual agent, it also includes: the system continuously monitors the connection status and working status of the manual agent, and when it is found that the manual agent is offline or fails, the outbound customer is immediately re-transferred to other suitable manual agents, and a text message is sent to the outbound customer to apologize and explain the situation, and the content of the text message is to apologize to the customer and explain the situation.

[0015] In the above embodiment, it is ensured that customers can always get a timely response when they need manual service, and service interruption will not be caused by problems with individual agents; secondly, sending text messages to apologize and explain the situation can improve customer satisfaction; finally, dynamic management of the status of manual agents optimizes the allocation of service resources and improves the overall outbound call efficiency.

[0016] In combination with some embodiments of the first aspect, in some embodiments, after selecting a robot based on the number of outbound customers and regional data, and calling the outbound customers in turn, it also includes: when the call to the outbound customer is not connected, recalling the unconnected number according to the preset repeated call frequency, recall conditions and time interval.

[0017] In the above embodiment, by setting the preset repeated call frequency, recall conditions and time interval, the unanswered number is recalled when the outbound call to the customer is not connected, avoiding missing the communication opportunity due to a single failure to connect, thereby improving the efficiency and success rate of communication; at the same time, a reasonable repeated call frequency, conditions and time interval can avoid excessive harassment of customers and find a balance between actively contacting customers and respecting customer experience.

[0018] In combination with some embodiments of the first aspect, in some embodiments, after selecting a robot based on the number of outbound customers and regional data, and calling the outbound customers in turn, it also includes: after the call starts, recording the call duration and monitoring the call connection status and whether the outbound customer makes any sound in the call recording; when the call is interrupted and the call duration does not exceed the preset minimum call duration threshold, the number of the interrupted outbound user is arranged in sequence to the last digit of the current outbound call task, and after completing all outbound call tasks, the interrupted outbound user is called again; the interrupted outbound user is an outbound user who did not make any sound during the call.

[0019] In the above embodiment, re-calling is performed for outbound call users whose calls may be interrupted due to various unexpected situations and the call duration is short, and no effective communication is performed, thereby increasing the chances of establishing effective communication with customers and improving the quality of completion of outbound call tasks; secondly, by reasonably arranging the timing of re-calling interrupted outbound call users, the overall progress of the current outbound call task is not affected, thereby improving the efficiency of the AI ​​voice call system.

[0020] In a second aspect, an embodiment of the present application provides an AI voice call system, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the AI ​​voice call system to execute the method described in the first aspect and any possible implementation method of the first aspect.

[0021] In a third aspect, an embodiment of the present application provides a computer program product comprising instructions. When the above-mentioned computer program product runs on an AI voice call system, the above-mentioned AI voice call system executes the method described in the first aspect and any possible implementation method of the first aspect.

[0022] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, comprising instructions. When the instructions are executed on an AI voice call system, the AI ​​voice call system executes the method described in the first aspect and any possible implementation method of the first aspect.

[0023] It can be understood that the AI ​​voice call system provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the AI ​​voice call method provided in the embodiment of the present application. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding method, which will not be repeated here.

[0024] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. This application screens target customers for calls and more accurately selects customers with high correlation for outbound calls, thereby improving outbound call efficiency. During the call, it determines whether the outbound user is interested. If not, the current call is terminated and the next outbound customer is called. By dynamically evaluating the customer's interest value, calls to uninterested customers are terminated in time, saving time and resources and improving overall outbound call efficiency.

[0025] 2. This application continuously monitors the connection status and working status of the manual seats. When it is found that the manual seat is offline or malfunctions, the outbound customer is immediately re-transferred to other suitable manual seats, and a text message is sent to the outbound customer to apologize and explain the situation, ensuring that the customer can always get a timely response when he needs manual service, and the service will not be interrupted due to problems with individual agents; secondly, sending text messages to apologize and explain the situation can improve customer satisfaction; finally, the dynamic management of the status of manual seats optimizes the allocation of service resources and improves the overall outbound efficiency.

[0026] 3. When the call is interrupted and the duration does not exceed the preset minimum call duration threshold, and the outbound user does not make any sound during the call, the number of the interrupted outbound user will be arranged in order to the last place of the current outbound call task, and the user will be called again after all outbound call tasks are completed. For outbound users whose calls may be interrupted due to various unexpected situations and whose calls are short and have not been effectively communicated, the opportunity to establish effective communication with customers is increased, and the quality of completion of outbound call tasks is improved. Secondly, by reasonably arranging the timing of re-calling the interrupted outbound user, the overall progress of the current outbound call task will not be affected, thereby improving the efficiency of the AI ​​voice call system. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a structural diagram of a system architecture to which the AI ​​voice call method in the embodiment of the present application can be applied; Figure 2 This is a flowchart of the AI ​​voice calling method in the embodiment of the present application; Figure 3 This is another flowchart of the AI ​​voice call method in an embodiment of the present application; Figure 4 It is an exemplary hardware structure diagram of the AI ​​voice call system controller in an embodiment of the present application. DETAILED DESCRIPTION

[0028] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to be used as limitations to the present application. As used in the specification and appended claims of the present application, the singular expressions "one", "a kind of", "said", "above", "the" and "this" are intended to also include plural expressions, unless there is a clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to and includes any or all possible combinations of one or more listed items.

[0029] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, unless otherwise specified, "plurality" means two or more.

[0030] Figure 1 It is a structural diagram of a system architecture to which the AI ​​voice call method in the embodiment of the present application can be applied.

[0031] See also Figure 1 The system includes a server 110 , a storage device 120 , an agent device 130 , a communication device 140 and a network device 150 .

[0032] Server 110 mainly includes voice server 111, database server 112 and application server 113, which are key components of the AI ​​voice call system and jointly ensure the efficient operation of the system. Among them, voice server 111 is used to process the input and output of voice signals, including functions such as voice recognition and voice synthesis, to ensure that the system can accurately understand the customer's voice and respond with clear voice; database server 112 is used to store important data such as customer information, speech library, recording data, etc., to provide data support for the operation of the system; application server 113 is used to run the system's business logic software, responsible for the allocation, monitoring and management of outbound call tasks and other core business processes.

[0033] The storage device 120 mainly includes a hard disk array 121 and a backup device 122, which are used to store a large amount of recording data for subsequent quality inspection and data analysis. The hard disk array 121 is composed of multiple hard disks, providing high-capacity storage, and can use independent disk redundant array technology to improve data reliability and read and write speed; the backup device 122 is used to regularly back up system data to prevent data loss. Among them, the backup device 122 can be, but is not limited to, a tape library, an external hard disk, a network attached storage and other devices.

[0034] The human agent device 130 is mainly used to answer calls transferred to human agents, view customer information, record service content, etc., so that the human agent can communicate with the customer. The human agent device 130 can be, but is not limited to, a computer, a mobile phone, a tablet computer, and other devices.

[0035] The communication device 140 mainly includes a voice gateway 141 and a text message gateway 142. The voice gateway 141 is used to connect the system to the telecommunications network, realize the transmission of voice signals, convert digital voice signals into analog signals for making calls through traditional telephone lines, and also convert received analog voice signals into digital signals for system processing; the text message gateway 142 is used to send text messages to customers.

[0036] The network device 150 mainly includes a switch 151 and a router 152, which are used to connect various devices in the system and connect to the external network. The switch 151 is used to connect various devices in the system to achieve fast data transmission and ensure that different devices can communicate and share data efficiently; the router 152 is used to connect the system with the external network to achieve Internet access and remote management, provide routing functions for network connections, and ensure that the system can communicate with the outside world.

[0037] This application provides an AI voice calling method, which first screens customers, only calls customers who meet the target user criteria, and judges the customer's interest in real time during the call. If the customer is not interested, the call is hung up in time and the next outbound customer is called. In this way, customers with high relevance can be screened more accurately for outbound calls, and calls to uninterested customers can be terminated in time, saving time and resources and improving the overall outbound call efficiency.

[0038] The following describes an AI voice calling method in one embodiment of the present application: like Figure 2 As shown, it is a flow chart of the AI ​​voice call method provided in the embodiment of the present application, which can be used for Figure 1 The system architecture shown includes the following steps: S201, after responding to the user's operation of starting the task, collecting the regional data, historical call value, real-time consumption behavior value and real-time browsing record value of each customer in the task in real time, and calculating the correlation parameter between each customer and the target customer; It can be understood that regional data refers to the system numbering different provinces with different numbers, and different numbers represent different regional data; the historical call value refers to the number of calls that the customer has made by the AI ​​voice call system to perform outbound call tasks; the real-time consumption behavior value refers to the number of times the customer has purchased the preset product up to now; the real-time browsing record value refers to the number of times the customer has browsed the preset information or products up to now.

[0039] When the user clicks the start task button on the system display page, the outbound call task is started in response to the user operation. First, the system collects various data for each customer in the task in real time. Among them, the regional data is obtained by retrieving the location information of the customer number; the historical call value is obtained by retrieving the system's historical call records; the real-time consumption behavior data is obtained by retrieving the real-time purchase record data of the preset product background; the real-time browsing record data of the preset information or product background is obtained. Real-time browsing record data. Secondly, by calculating the sum of the real-time consumption behavior value and the real-time browsing value, and then calculating the ratio of the above sum to the historical call data, the obtained ratio value is used as the correlation parameter with the target customer.

[0040] S202, selecting a robot according to the number and regional data of outbound customers, and calling the outbound customers in sequence; After screening customers whose correlation parameters exceed the preset correlation threshold as outbound users, first, monitor the number of outbound customers, select the corresponding preset numerical range according to the preset numerical range threshold, and screen the preset corresponding robots within the numerical range; secondly, determine the regional data with the largest number of occurrences among the regional data of all outbound customers of the task, and screen the robots corresponding to the regional data with the largest number of occurrences among the preset corresponding robots within the numerical range as the robots executing the outbound task; finally, use the robots executing the outbound task to call the outbound customers in turn.

[0041] In the above steps, selecting robots based on the number of customers and regional data can improve the targeting and efficiency of outbound calls. Different robots may have different advantages when handling tasks of different sizes and customers in different regions, thereby better meeting outbound call needs.

[0042] S203, after the call starts, extract the voice features of the outbound customer during the call in real time, convert the voice of the outbound customer into text using voice recognition technology, and collect real-time speech speed data, real-time intonation data, real-time volume data and real-time pause frequency data in the voice features, as well as the number of occurrences of preset keywords of interest and the number of occurrences of preset keywords of no interest in the above text; It can be understood that real-time intonation data refers to dividing the possible pitch frequencies of human speech into several intervals, each interval corresponds to a specific value, and the real-time pitch frequency is monitored through audio analysis, the pitch frequency interval is determined and the corresponding value is assigned, which is the real-time intonation data; real-time pause frequency data refers to when a voice interruption is detected, the terminal time is greater than the preset pause time threshold, and the current interruption is judged to be a customer-initiated interruption. After monitoring the customer's active interruption, the number of active interruptions in the current call is counted in real time, and the ratio of the number of active interruptions to the real-time call duration is calculated as the real-time pause frequency data.

[0043] S204, inputting the above-mentioned real-time speech speed data, real-time intonation data, real-time volume data, real-time pause frequency data and the number of occurrences of the preset keyword of interest into a preset dynamic interest value evaluation model to obtain the dynamic interest value of the outbound customer; The dynamic interest value evaluation model uses a pre-set weight allocation mechanism to set different weights for real-time speech speed data, real-time intonation data, real-time volume data, real-time pause frequency data, and the number of occurrences of preset keywords of interest. After the data is input, it is integrated and calculated according to the preset weights, and the resulting value is the dynamic interest value of the outbound customer.

[0044] S205: When the duration of the outbound customer being in the uninterested state exceeds a preset uninterested duration or the number of occurrences of the preset uninterested keyword exceeds a preset uninterested keyword number threshold, terminate the current call and call the next outbound customer.

[0045] During the call, the dynamic interest value of the outbound customer is calculated in real time. When the dynamic interest value is lower than the preset interest threshold, the outbound customer is judged to be in a state of disinterest, and the disinterest state is timed. When the duration of the disinterest state exceeds the preset disinterest duration, the customer is judged not to be the target user. At the same time, the number of occurrences of the preset disinterested keywords is monitored after the call starts. When the number of occurrences of the preset disinterested keywords exceeds the preset disinterested keyword number threshold, the customer is judged not to be the target user. After judging that the customer is not the target user, the call is terminated in time and the next customer is called.

[0046] By monitoring the customer's interest in real time, we can determine whether the customer is the target user. After determining that the customer is not the target user, we can terminate the call in time and call the next customer, avoiding wasting too much time on uninterested customers and turning to customers who are more likely to generate positive feedback in time, thereby improving the overall outbound call effect. Secondly, we can improve customer experience. For uninterested customers, terminating the call in time can reduce their troubles and avoid causing resentment.

[0047] In the above embodiment, customers with high relevance to the target customers are screened out through historical data, and a suitable robot is selected to make outbound calls to them. During the call, it is monitored in real time whether the customer is interested. If the customer is not interested for a long time, the call is terminated in time, which can save time and resources and improve the overall outbound call efficiency.

[0048] In some embodiments, when the AI ​​voice call system performs outbound calling tasks, it may encounter some special situations, such as the call is not connected or the customer disconnects the call due to signal interruption or inconvenience in answering the call after the call is connected. The AI ​​voice call system determines that the customer is in different situations and uses different methods to re-call different customers, thereby improving the contact success rate and information transmission rate with the customer. Figure 3As shown, it is another flowchart of the AI ​​voice call method provided in the embodiment of the present application, which can be used Figure 1 The AI ​​voice call system architecture shown includes the following steps: S301, after responding to the user's operation of starting the task, collecting the regional data, historical call data, real-time consumption behavior value and real-time browsing record value of each customer in the task in real time, and calculating the correlation parameter between each customer and the target customer; S302, selecting a robot according to the number of outbound customers and regional data, and determining whether the robot is currently idle; If it is not in idle state, execute the following step S303; If it is in idle state, execute the following step S304; Customers whose correlation parameters exceed the preset correlation threshold are selected as outbound users, and the appropriate robot is selected based on the number of outbound customers and regional data. After selecting a robot, determine whether the robot is currently idle. If it is not idle, wait until the robot is idle before selecting it; if it is idle, use the robot directly to call the outbound customer.

[0049] S303, after waiting for the robot to be in an idle state, execute the following step S304; When it is detected that the selected robot is not currently idle, wait until the robot becomes idle before using the robot to make calls to outbound customers in turn.

[0050] S304, using the robot to call outbound customers in turn; When it is detected that the selected robot is currently idle, the robot is directly used to call outbound customers one by one.

[0051] In the above steps, firstly, by screening customers with high correlation for outbound calls, the success rate and return rate of outbound calls are improved; secondly, robots are selected according to the number of customers and regional data, which enhances the targeting and adaptability of outbound calls; finally, the judgment and processing of the idle state of the robot ensures the efficient execution of outbound call tasks and avoids waste of resources and task conflicts.

[0052] S305, start recording and detect whether the audio quality meets the requirements; If the requirements are not met, the following step S306 is executed; If the requirements are met, the following step S307 is executed; After starting recording, the system detects whether the audio quality meets the requirements to avoid affecting the accuracy of speech feature extraction and speech recognition.

[0053] In some embodiments of the present application, the audio quality is detected by monitoring the clarity of the audio and whether there is noise interference; in other embodiments of the present application, the audio quality can also be detected by detecting whether the audio volume is within a preset volume range threshold, etc., which is not limited here.

[0054] S306, pausing the recording, adjusting the recording parameters and executing the above step S305 again; If the audio quality does not meet the requirements, the system will pause the recording and adjust the recording parameters. After the adjustment, step S305 is executed again, that is, whether the audio quality meets the requirements.

[0055] In some embodiments of the present application, the recording parameters are adjusted by using a method of noise reduction of the recording; in other embodiments of the present application, the recording parameters may also be adjusted by adjusting the microphone volume, etc., which is not limited here.

[0056] S307, extracting the voice features of the outbound customer during the call in real time, and converting the voice of the outbound customer into text using voice recognition technology; When the audio quality meets the requirements, the system will extract the voice features of the outbound customer during the call in real time, including speaking speed, intonation, volume, etc., and use speech recognition technology to convert the outbound customer's voice into text.

[0057] S308, when text conversion fails, automatically switch to other speech recognition engines for attempts, and publish a prompt message on the system homepage; If text conversion fails, the system will automatically switch to other speech recognition engines for attempts to avoid the inability to obtain text content due to problems with a single speech recognition engine. At the same time, prompt information will be posted on the system homepage to allow administrators or relevant personnel to be aware of the problem in a timely manner so that further measures can be taken.

[0058] In the above steps, first start recording and test the audio quality. If the requirements are not met, pause the recording and adjust the parameters. Once the requirements are met, extract the voice features and convert them into text. When there is a problem with the text conversion, automatically switch to the voice recognition engine and issue a prompt message to ensure the reliability of basic data, improve system stability and management efficiency, so as to fully understand customer expressions and emotions, and make the voice processing of outbound call tasks more accurate and efficient.

[0059] S309, real-time collection of the above-mentioned voice features and real-time speech speed data, real-time intonation data, real-time volume data, real-time pause frequency data, the number of occurrences of preset keywords of interest and the number of occurrences of preset keywords of no interest in the text; S310, inputting the above-mentioned real-time speech speed data, real-time intonation data, real-time volume data, real-time pause frequency data and the number of occurrences of preset keywords of interest into a preset dynamic interest value evaluation model to obtain the dynamic interest value of the outbound customer; S311, when the duration of the outbound customer being in a state of disinterest exceeds a preset disinterest duration or the number of occurrences of a preset disinterest keyword exceeds a preset disinterest keyword number threshold, terminate the current call and call the next outbound customer; S312, in the case where the outbound call to the customer is not connected, re-calling the unconnected number according to the preset repeated call frequency, re-call condition and time interval; When an outbound call is not connected, the system will retry the unanswered number according to the preset retry frequency, retry conditions and time interval. The preset retry frequency determines the number of times to try to call a number that has not been connected. The retry conditions may include different situations, such as how long after the first call if the call is not connected before retrying. The time interval specifies the time interval between each retry.

[0060] In the above steps, by recalling in case of non-connection, potential opportunities can be avoided from being missed due to a single non-connection, and outbound call resources can be fully utilized to improve the efficiency and success rate of outbound call tasks.

[0061] S313. After the call starts, record the call duration and monitor the call connection status and whether the outbound customer makes a sound in the call recording; After the call begins, the system records the duration of the call for subsequent analysis of call efficiency and other situations. At the same time, it continuously monitors the call connection to ensure the stability of the call so that any connection problems can be handled in a timely manner. In addition, it also monitors whether the outbound customer makes any sound in the call recording, which helps to determine whether the customer is actively participating in the call or there are abnormal situations such as silence.

[0062] S314, when the call duration does not exceed the preset minimum call duration threshold when the call is interrupted, the number of the interrupted outbound user is arranged in order to the last position of the current outbound call task, and the interrupted outbound user is called again after all outbound call tasks are completed; It is understandable that if a call is interrupted and the call duration at this time does not exceed the preset minimum call duration threshold, it means that the call may be interrupted due to an unexpected situation instead of ending normally.

[0063] When a call is interrupted and the call duration does not exceed the preset minimum call duration threshold, the system will arrange the number of the interrupted outbound user in order to the last digit of the current outbound call task. After completing all outbound call tasks, the system will call the interrupted outbound user again to ensure complete communication with the customer.

[0064] In the above steps, the stability and integrity of the outbound call task are improved by recording the call duration, monitoring the call connection and customer voice, and properly handling the interrupted calls. On the one hand, problems in the call can be discovered in time and corresponding measures can be taken. On the other hand, the interrupted call can be called again, which increases the chances of successful communication with the customer and improves the efficiency and quality of the outbound call task.

[0065] S315, monitoring whether there are triggering keywords for preset conversion to manual in the text converted from the voice of the outbound customer; The system continuously monitors the text content converted from the outbound customer's voice to check whether there are any triggering keywords for preset transfer to manual service. The preset transfer to manual service keywords are pre-set, indicating that the customer may have more complex questions or special needs that require manual service to better handle.

[0066] S316: If the text contains a keyword that triggers the preset call transfer to a human agent, the robot transfers the outbound user to a human agent, terminates the current call, and calls the next outbound customer; If the system detects that there are keywords in the text that trigger the preset call to a human agent, the system will transfer the outbound caller to a human agent. At this time, the robot terminates the current call so that the human agent can communicate with the customer, and the robot starts calling the next outbound customer.

[0067] A manual seat refers to a work position occupied by a dedicated customer service staff in a customer service or outbound call system. The customer service staff communicates directly with the customer here to provide more humane and flexible services and better understand the customer's emotions and needs.

[0068] S317, the system continuously monitors the connection status and working status of the human agent; The system continuously monitors the connection status and working status of the human agent to ensure that the human agent can normally receive the transferred customers and is in a state where it can provide services. S318. When it is found that the human agent is offline or malfunctioning, the outbound customer will be immediately transferred to another suitable human agent, and a text message will be sent to apologize to the outbound customer and explain the situation.

[0069] If a human agent is found to be offline or malfunctioning, the system will immediately re-route the outbound customer to another suitable human agent to avoid the customer waiting for a long time or being unattended. At the same time, the system will send a text message to the outbound customer, apologizing to the outbound customer and explaining the situation.

[0070] In the above steps, by monitoring keywords, we can accurately judge customer needs and transfer them to manual services in a timely manner, ensuring that customers can receive more professional treatment when they have special needs. At the same time, we continuously monitor the status of manual agents and respond quickly when problems arise, which improves the stability and reliability of the service, reduces customer waiting time and dissatisfaction, and improves the overall quality of outbound service.

[0071] Steps S301, S309-S311 and Figure 2 Steps S201 to S205 in the illustrated embodiment are similar, and the descriptions of steps S201 to S205 may be referred to, and will not be repeated here.

[0072] In the above embodiment, first, the correlation parameters are calculated according to the customer's historical call data, real-time consumption and browsing records to screen outbound customers, and the appropriate robot is selected in combination with the customer quantity and regional data, and the idle state of the robot is judged to ensure efficient outbound calls; the recording is turned on and the audio quality is detected. If it is not satisfied, the parameters are adjusted. If it is satisfied, the voice features are extracted and converted into text. When there is a problem with the text conversion, the voice recognition engine is switched and a prompt is issued; a variety of voice and text data are collected and input into the evaluation model to obtain a dynamic interest value, and the call can be terminated when the customer's uninterested time exceeds the threshold or the keyword exceeds the number of times; the unreachable customer is re-called according to the preset frequency, conditions and time interval; the call duration is recorded, the connection and the customer's voice are monitored, and if the interruption duration does not exceed the threshold, it is arranged to be re-called at the end of the task; the preset transfer to manual keywords in the monitoring text are transferred to the manual seat if they exist, and the status of the manual seat is continuously monitored. In the event of a failure, the call is transferred and an apology is given to the customer, thereby improving the efficiency, accuracy and customer satisfaction of outbound calls, ensuring the smooth progress of outbound call tasks and the improvement of service quality.

[0073] The following introduces an exemplary AI voice call system controller provided in an embodiment of the present application.

[0074] Figure 4 4 is a schematic diagram of an exemplary hardware structure of the AI ​​voice call system controller 400 provided in an embodiment of the present application.

[0075] In some embodiments, the AI ​​voice call system controller 400 is a computer device or the AI ​​voice call system controller 400 includes a computer device. The computer device includes a processor, a memory, and a network interface connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data. The network interface of the computer device is used to communicate with other external terminals or servers through a network connection. In some embodiments, the network interface can be a wired network interface, and in some embodiments, the network interface can also be a wireless network interface. When the computer program is executed by the processor, the method in the embodiment of the present application is implemented.

[0076] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0077] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0078] As used in the above embodiments, the term "when..." may be interpreted to mean "if..." or "after..." or "in response to determining..." or "in response to detecting...", depending on the context. Similarly, the phrases "upon determining..." or "if (the stated condition or event) is detected" may be interpreted to mean "if determining..." or "in response to determining..." or "upon detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)", depending on the context.

[0079] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk), etc.

[0080] Those skilled in the art can understand that to implement all or part of the processes in the above-mentioned embodiments, the processes can be completed by computer programs to instruct related hardware, and the programs can be stored in computer-readable storage media. When the programs are executed, they can include the processes of the above-mentioned method embodiments. The aforementioned storage media include: ROM or random access memory RAM, magnetic disk or optical disk and other media that can store program codes.

Claims

1. An AI voice calling method, characterized in that: include: After responding to the user's operation of starting a task, the regional data, historical call value, real-time consumption behavior value and real-time browsing record value of each customer in the task are collected in real time, and the sum of the real-time consumption behavior value and the real-time browsing value is calculated, and then the ratio of the sum to the historical call data is calculated as the correlation parameter with the target customer; Select a robot according to the number of outbound customers and the regional data, and call the outbound customers in turn; the outbound customers are customers whose correlation parameters exceed a preset correlation threshold; After the call begins, the voice features of the outbound customer during the call are extracted in real time, the voice of the outbound customer is converted into text using voice recognition technology, and the real-time speech speed data, real-time intonation data, real-time volume data and real-time pause frequency data in the voice features, the number of occurrences of preset keywords of interest and the number of occurrences of preset keywords of no interest in the text are collected in real time; The real-time speech speed data, the real-time intonation data, the real-time volume data, the real-time pause frequency data and the number of occurrences of the preset keyword of interest are input into a preset dynamic interest value evaluation model to obtain a dynamic interest value of an outbound customer; the preset dynamic interest value evaluation model is a preset dynamic interest value evaluation model that calculates the mean of the real-time speech speed data, the real-time intonation data, the real-time volume data, the real-time pause frequency data and the number of occurrences of the preset keyword of interest with different weights; When the time that the outbound customer is in the uninterested state exceeds the preset uninterested time or the number of occurrences of the preset uninterested keyword exceeds the preset uninterested keyword number threshold, the current call is terminated and the next outbound customer is called; the uninterested state is that the outbound customer's dynamic interest value is lower than the preset interest threshold.

2. The method according to claim 1, characterized in that The selecting of robots according to the number of outbound customers and the regional data, and calling the outbound customers in sequence, specifically includes: Selecting a robot according to the number of outbound customers and the regional data, and determining whether the robot is currently in an idle state; If the robot is not in an idle state, wait until the robot is in an idle state and then select the robot again; If the current state is idle, calls will be made to outbound customers one by one.

3. The method according to claim 1, characterized in that: After the call starts, the voice features of the outbound customer during the call are extracted in real time, the voice of the outbound customer is converted into text using voice recognition technology, and the voice features and real-time speech speed data, real-time intonation data, real-time volume data, real-time pause frequency data, the number of occurrences of preset interested keywords and the number of occurrences of preset uninterested keywords in the text are collected in real time, specifically including: After the call starts, start recording to check whether the audio quality meets the requirements; If the audio quality does not meet the requirements, pause the recording, adjust the recording parameters and check again whether the audio quality meets the requirements; If the audio quality meets the requirements, the voice features of the outbound customer during the call are extracted in real time, and the voice of the outbound customer is converted into text using speech recognition technology; If the text conversion fails, it will automatically switch to other speech recognition engines for further attempts, and at the same time, a prompt message will be posted on the system homepage, which will prompt a human to review and correct the abnormal speech. The speech features and real-time speech speed data, real-time intonation data, real-time volume data, real-time pause frequency data, preset occurrence times of interested keywords and preset occurrence times of uninterested keywords in the text are collected in real time.

4. The method according to claim 1, characterized in that: After selecting the robot according to the number of outbound customers and the regional data and calling the outbound customers in sequence, the method further includes: During the call, voice recognition technology is used to convert the outbound customer's voice into text, and the text is monitored for keywords that trigger the preset manual conversion; When it is detected whether there is a keyword in the text that triggers the preset transfer to manual call, the outbound caller with the keyword is transferred to a manual agent, the robot terminates the current call, and calls the next outbound customer.

5. The method according to claim 4, characterized in that In the case where it is detected whether there is a preset keyword in the text that triggers the manual transfer, after the call is transferred to the manual agent, the method further includes: The system continuously monitors the connection status and working status of the manual agent. When it is found that the manual agent is offline or malfunctions, the system immediately re-transfers the outbound customer to other suitable manual agents, and sends a text message to the outbound customer to apologize and explain the situation; the content of the text message is to apologize to the customer and explain the situation.

6. The method according to claim 1, characterized in that After selecting the robot according to the number of outbound customers and the regional data and calling the outbound customers in sequence, the method further includes: In the case that the outbound call to the customer is not connected, the unconnected number is called back according to the preset repeated call frequency, call back conditions and time interval.

7. The method according to claim 1, characterized in that After selecting the robot according to the number of outbound customers and the regional data and calling the outbound customers in sequence, the method further includes: After the call starts, record the call duration and monitor the call connection status and whether the outbound customer makes a sound in the call recording; When a call is interrupted and the call duration does not exceed the preset minimum call duration threshold, the number of the interrupted outbound user will be arranged in order to the last digit of the current outbound call task, and the interrupted outbound user will be called again after all outbound call tasks are completed; the interrupted outbound user is an outbound user who did not make any sound during the call.

8. An AI voice calling system, characterized in that: The AI ​​voice call system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the AI ​​voice call system to execute the method described in any one of claims 1-7.

9. A computer program product comprising instructions, characterized in that When the computer program product runs on an AI voice call system, the AI ​​voice call system is enabled to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium comprising instructions, characterized in that: When the instruction is executed on the AI ​​voice call system, the AI ​​voice call system is caused to execute the method as claimed in any one of claims 1 to 7.

Images