Intelligent voice outbound method and device, electronic equipment and storage medium

Abstract

The invention relates to an intelligent voice outbound method, which comprises the following steps that: owner information and a verbal skill template are collected and stored in a system database, and the owner information is encrypted information and at least comprises mobile phone number information and address information; importing owner information, and obtaining an owner information table to create an outbound task; converting the verbal skill template to obtain verbal skill voice streaming media; matching a seat according to the call-out task and calling a mobile phone number of an owner so as to dialogue the owner according to the verbal skill voice streaming media, thereby analyzing voice dialogue content of the owner and generating text dialogue content; and processing the text dialogue content, obtaining a text semantic analysis result, and selecting one execution strategy of continuous communication, hanging up and redialing to complete an outbound task. The invention further relates to an intelligent voice outbound device, electronic equipment and a storage medium. By using the intelligent voice outbound method and device, the electronic equipment and the storage medium provided by the invention, the problem that a voice outbound system is poor in efficiency, intelligence and reliability is solved.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to an intelligent voice outbound calling method, device, electronic equipment and storage medium. Background Art

[0002] With the rapid development of informatization and intelligence, voice outbound calling technology has been widely used in scenarios such as electricity bill collection, service notifications and market promotion. Traditional voice outbound calling usually relies on manual customer service to complete tasks. In recent years, with the development of artificial intelligence, natural language processing (NLP), speech recognition (ASR) and text-to-speech (TTS) technology, automatic voice outbound calling systems based on intelligent algorithms have gradually been used to replace traditional methods.

[0003] Although the existing intelligent voice outbound call system has improved the level of automation to a certain extent, it still faces many problems. For example, in the process of collecting and storing user information, there is a lack of effective data encryption and integrity verification methods, and there is a risk of data leakage. The existing voice outbound call system does not process and classify user information in a detailed manner, making it difficult to arrange personalized outbound call tasks based on the specific needs of different users. The quality of speech generation and recognition is limited: the depth of natural language processing in the speech synthesis and recognition process is not enough, the generated speech content is not smooth enough, and the accuracy of speech recognition cannot meet the requirements of complex interactive scenarios. The existing technology has not yet formed a complete method that integrates data security processing, efficient speech generation and recognition, accurate semantic analysis, and dynamic task adjustment, resulting in a large room for improvement in the efficiency, intelligence, and reliability of the voice outbound call system.

[0004] Therefore, the prior art has defects and needs to be improved and developed. Summary of the invention

[0005] The embodiments of the present invention provide an intelligent voice outbound call method, device, electronic device and storage medium, which are used to solve the existing intelligent voice outbound call system. Although the automation level has been improved to a certain extent, it still faces many problems. For example, in the process of collecting and storing user information, there is a lack of effective data encryption and integrity verification means, and there is a risk of data leakage; the existing voice outbound call system is not detailed enough in processing and classifying user information, and it is difficult to arrange personalized outbound call tasks according to the specific needs of different users; the quality of speech generation and recognition is limited: the depth of natural language processing in the process of speech synthesis and recognition is not enough, the generated speech content is not smooth enough, and the accuracy of speech recognition cannot meet the requirements of complex interactive scenarios. That is, the existing technology has not yet formed a complete method that integrates data security processing, efficient speech generation and recognition, accurate semantic analysis and dynamic task adjustment, resulting in technical problems that there is a large room for improvement in the efficiency, intelligence and reliability of the voice outbound call system.

[0006] According to a first aspect of an embodiment of the present invention, there is provided an intelligent voice outbound calling method, comprising:

[0007] Collecting owner information and speech templates and storing them in the system database, wherein the owner information is encrypted information and includes at least mobile phone number information and address information;

[0008] Importing the owner information, obtaining an owner information table to create an outbound call task, wherein the owner information in the owner information table is arranged in ascending order according to the value of the mobile phone number information;

[0009] Converting the speech template to obtain speech audio streaming media;

[0010] Matching agents according to outbound call tasks and calling the owner's mobile phone number to communicate with the owner based on speech voice streaming media, which is used to analyze the owner's voice conversation content and generate text conversation content;

[0011] Process the text conversation content, obtain the text semantic analysis result, and select an execution strategy among continuing communication, hanging up, and redialing to complete the outbound call task.

[0012] Furthermore, the owner information and speech templates are collected and stored in the system database, wherein the owner information is encrypted information and at least includes mobile phone number information and address information, including:

[0013] Obtain the owner's mobile phone number information and address information through the operator interface, and format them into standardized structured mobile phone number data and standardized structured address data respectively;

[0014] Perform AES-256 encryption on standard structured mobile phone number data and address data;

[0015] Use hash message authentication code to verify the integrity of encrypted mobile phone number and address data;

[0016] The mobile phone number data and address data that have passed data integrity verification are stored in the system database.

[0017] Further, the importing of the owner information to obtain an owner information table to create an outbound call task, wherein the owner information in the owner information table is arranged from small to large according to the value of the mobile phone number information, includes:

[0018] Import said owner information in CSV format;

[0019] Using K-Means based clustering classification on the owner information to obtain the owner information after labeling;

[0020] Using a public-private key encryption mechanism based on elliptic curve to sign the owner information;

[0021] The signed owner information is sorted in ascending order according to the value of the mobile phone number to obtain the owner information table. The order of the mobile phone numbers in the owner information table is the outbound call task order.

[0022] Furthermore, the converting of the speech template to obtain speech voice streaming media includes:

[0023] When the outbound call task is initiated, the address information is embedded into the speech template;

[0024] Matching idle seats to obtain a sound media file by converting the speech template embedded with the address information through a TTS conversion method based on deep learning;

[0025] Search for an idle SIP line to convert the sound media file into voice streaming media.

[0026] Furthermore, matching the idle seats to obtain the voice media file by converting the speech template embedded with the address information through a TTS conversion method based on deep learning includes:

[0027] Preprocess the text of the speech template by performing word segmentation, removing stop words and punctuation marks;

[0028] The preprocessed text of the speech template is converted into a phoneme sequence by phoneme extraction, and the phoneme extraction formula is:

[0029] S=P(T),

[0030] Among them, T is text, P is the phoneme extraction function, and S is the phoneme sequence;

[0031] Generate Mel spectra using the Tacotron2 model;

[0032] The Mel spectrogram is converted into natural speech by using the WaveGlow algorithm to obtain the sound media file.

[0033] Furthermore, the matching of seats according to the outbound call task and calling the owner's mobile phone number to communicate with the owner according to the speech voice streaming media is used to analyze the owner's voice conversation content and generate text conversation content, including:

[0034] Associating the voice streaming media with the mobile phone number information;

[0035] Pushing the voice streaming media and the mobile phone number information to the operator's call system to perform an outbound call task;

[0036] Collecting the owner's voice conversation content and generating conversation audio;

[0037] The audio signal of the conversation audio is framed and windowed using a window function, and the window function formula is:

[0038] W=0.54-0.45cos(2πn / (N-1)),

[0039] Where N is the window length;

[0040] Extracting audio features of the conversation audio through Mel-frequency cepstral coefficients;

[0041] Decoding the conversation audio by connecting the timing classification, and mapping the conversation audio into a conversation text sequence;

[0042] The dialogue text sequence is optimized through a Transformer language model to generate text dialogue content.

[0043] Further, the processing of the text conversation content, obtaining a text semantic analysis result, and selecting an execution strategy of continuing communication, hanging up, and redialing to complete the outbound call task includes:

[0044] Performing word segmentation processing on the text conversation content based on the BERT word segmenter to obtain the conversation text word segmentation;

[0045] The dialogue text segmentation is input into the pre-trained BERT model to generate a dialogue text sentence vector. The dialogue text sentence vector generation formula is:

[0046] V=f(t),

[0047] Where t is the input text, f is the BERT embedding function, and v is the output vector;

[0048] Classifying the conversation text sentence vectors by a Softmax classifier to obtain sentiment analysis content;

[0049] Classify the sentiment analysis content based on the Transformer intent classification model and output the communication strategy;

[0050] Record the communication content of each outbound call task and the owner's response result, and the response result at least includes continue communication, hang up, and call again later;

[0051] A new owner information table is generated according to the communication strategy and the response result to execute the next outbound calling task after the outbound calling task is completed.

[0052] According to a second aspect of an embodiment of the present invention, there is provided an intelligent voice outbound calling device, comprising:

[0053] A collection module, used to collect owner information and speech templates and store them in a system database, wherein the owner information is encrypted information and includes at least mobile phone number information and address information;

[0054] An import module, used to import the owner information, obtain an owner information table to create an outbound call task, wherein the owner information in the owner information table is arranged in ascending order according to the value of the mobile phone number information;

[0055] A conversion module, used for converting the speech template to obtain speech voice streaming media;

[0056] An analysis module is used to match an agent according to an outbound call task and call the owner's mobile phone number to communicate with the owner according to the speech voice streaming media, and is used to analyze the owner's voice conversation content and generate text conversation content;

[0057] The processing module is used to process the text conversation content, obtain the text semantic analysis result and select an execution strategy among continuing communication, hanging up and redialing to complete the outbound call task.

[0058] According to a third aspect of an embodiment of the present invention, there is provided an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program implements the intelligent voice outbound calling method when executed by the processor.

[0059] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the intelligent voice outbound calling method is implemented.

[0060] Beneficial effects:

[0061] It can be seen from the above scheme that the present invention provides an intelligent voice outbound call method, device, electronic device and storage medium. By combining data encryption technology, efficient voice processing algorithm and intelligent task management mechanism, it solves the problems of data security, voice interaction accuracy and insufficient task dynamic adjustment capability in the prior art, and improves the overall performance and application value of the voice outbound call system. Improve data security: By introducing AES-256 encryption algorithm, hash message authentication code and public-private key signature technology, the security of user information during collection, storage and use is ensured to prevent data leakage and tampering. Improve task management efficiency: Use K-Means clustering and sorting algorithm to label user information, realize accurate scheduling and priority allocation of tasks, and improve the execution efficiency of outbound call tasks. Enhance voice processing capabilities: Generate highly natural speech based on deep learning models (such as Tacotron2 and WaveGlow), and combine Mel-frequency cepstral coefficients and window function technology to improve the accuracy and robustness of speech recognition. Realize dynamic task adjustment: Use the BERT model and Transformer intent classification algorithm to perform real-time sentiment analysis and intent judgment on user voice interaction content, and support flexible task strategy adjustment, including continuing communication, hanging up or re-dial, to meet the needs of different scenarios. Through this invention, voice outbound calling technology will be improved in terms of security, intelligence and flexibility, providing an efficient and reliable solution for fields such as electricity bill collection, customer service and marketing, while promoting the application and development of voice technology in more fields.

[0062] It should be appreciated that all combinations of the foregoing concepts, as well as additional concepts described in greater detail below, may be considered to be part of the inventive subject matter of the present disclosure, provided such concepts are not mutually inconsistent.

[0063] The foregoing and other aspects, embodiments and features of the present invention can be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the present invention, such as the features and / or beneficial effects of the exemplary embodiments, will be apparent from the following description or learned from the practice of the specific embodiments according to the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for describing the embodiments are briefly introduced below.

[0065] Figure 1 This is a flow chart of the intelligent voice outbound calling method according to an embodiment of the present application. DETAILED DESCRIPTION

[0066] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Unless otherwise defined, the technical terms or scientific terms used herein should be the common meaning understood by people with general skills in the field to which the present invention belongs.

[0067] The words "first", "second" and similar words used in the patent application specification and claims of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, unless the context clearly indicates otherwise, the singular forms of "a", "an" or "the" and other similar words do not indicate a quantitative limitation, but indicate the existence of at least one. Words such as "include" or "comprise" mean that the elements or objects appearing before "include" or "comprise" cover the features, wholes, steps, operations, elements and / or components listed after "include" or "comprise", and do not exclude the existence or addition of one or more other features, wholes, steps, operations, elements, components and / or their collections. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0068] In the prior art, although the intelligent voice outbound call system has improved the level of automation to a certain extent, it still faces many problems. For example, in the process of collecting and storing user information, there is a lack of effective data encryption and integrity verification methods, and there is a risk of data leakage. The existing voice outbound call system does not process and classify user information in a detailed manner, making it difficult to arrange personalized outbound call tasks based on the specific needs of different users. The quality of speech generation and recognition is limited: the depth of natural language processing in the speech synthesis and recognition process is not enough, the generated speech content is not smooth enough, and the accuracy of speech recognition cannot meet the requirements of complex interactive scenarios. The prior art has not yet formed a complete method that integrates data security processing, efficient speech generation and recognition, accurate semantic analysis, and dynamic task adjustment, resulting in a large room for improvement in the efficiency, intelligence, and reliability of the voice outbound call system.

[0069] In view of this, an embodiment of the present invention provides an intelligent voice outbound calling method, referring to Figure 1 ,include:

[0070] Step S102: collecting owner information and speech templates and storing them in the system database, wherein the owner information is encrypted information and includes at least mobile phone number information and address information;

[0071] Step S104: importing owner information, obtaining an owner information table to create an outbound call task, wherein the owner information in the owner information table is arranged in ascending order according to the value of the mobile phone number information;

[0072] Step S106: converting the speech template to obtain speech voice streaming media;

[0073] Step S108: matching an agent according to the outbound call task and calling the owner's mobile phone number to communicate with the owner according to the speech voice streaming media, so as to analyze the owner's voice conversation content and generate text conversation content;

[0074] Step S110: Process the text conversation content, obtain the text semantic analysis result, and select an execution strategy among continue communication, hang up, and redial to complete the outbound call task.

[0075] By systematically integrating multiple technologies such as data encryption, voice generation and recognition, task management and semantic analysis, the entire process from collecting owner information to completing outbound call tasks is completed.

[0076] The general operation method is as follows: collect and store owner information in the database, complete data standardization and encrypted storage through the operator interface; import owner information, use clustering and sorting algorithms to generate an outbound call task table, and perform signature verification on the data; after starting the task, generate voice streaming media through the TTS module, push it to the operator line, and start the outbound call task; collect audio data during the call, and convert it into text content through the voice recognition module after window function framing and feature extraction; use the semantic analysis module to parse the text, determine the subsequent strategy and update the task status. The advantages of this method include high data security: data is encrypted and integrity verified during collection, storage and transmission to protect user privacy. Improved speech naturalness and recognition accuracy: The application of deep learning models improves the naturalness of speech synthesis and the robustness of speech recognition. High task execution efficiency: Through automated task allocation and dynamic strategy adjustment, the need for human intervention is reduced and the efficiency of task completion is improved. Wide range of application scenarios: It can be widely used in fields such as electricity bill collection and customer service notifications, and supports customized task optimization based on semantic analysis. Multiple security mechanisms: It integrates encryption, signature and integrity verification technologies to achieve data protection throughout the process. Deep learning-driven speech processing brings speech generation and recognition capabilities to an advanced level. Combining semantic analysis with dynamic strategies: The BERT model and Transformer intent classification model are used to improve the accuracy of semantic analysis and support dynamic task adjustment.

[0077] The embodiments of the present invention provide an intelligent voice outbound calling method that supports operations such as customization of speech templates, multi-user multi-concurrency, data analysis, calling, and result recording. In some other embodiments, through AI training, AI interactive expansion functions can also be realized, including simple interactive questions and answers, etc., to meet the later AI interactive deep learning capability function expansion. When executing an outbound call task, if the call is successful and answered, the call ends. If the call fails or is not answered, the system will automatically call again. The number of retries can be customized, and a blacklist can be set to implement filtering, blacklisting, and other operations. Outbound call tasks can also support two methods of acquisition: EXCEL table import and direct acquisition of external system data.

[0078] In some embodiments, the owner information and speech templates are collected and stored in the system database, wherein the owner information is encrypted information and includes at least mobile phone number information and address information, including:

[0079] Step S1021: obtaining the owner's mobile phone number information and address information through the operator interface, and formatting them into standardized structured mobile phone number data and standardized structured address data respectively;

[0080] Step S1022: Perform AES-256 encryption processing on the standard structured mobile phone number data and address data;

[0081] Step S1023: Using a hash message authentication code to verify the data integrity of the encrypted mobile phone number data and address data;

[0082] Step S1024: The mobile phone number data and address data that have passed the data integrity verification are stored in the system database.

[0083] The purpose of steps S1021 to S1023 is to collect and store owner information. The system obtains the owner's mobile phone number and address information through the operator interface and formats it into standardized structured data. The structured data is encrypted using the AES-256 algorithm, and the integrity of the data is verified using a hash message authentication code. The verified data is stored in the system database to ensure data security and reliability.

[0084] In some embodiments, the owner information is imported to obtain an owner information table to create an outbound call task, wherein the owner information in the owner information table is arranged from small to large according to the value of the mobile phone number information, including:

[0085] Step S1041: Importing owner information in CSV format;

[0086] Step S1042: clustering and classifying the owner information based on K-Means to obtain the owner information after labeling;

[0087] Step S1043: using the public-private key encryption mechanism based on elliptic curve to sign the owner information;

[0088] Step S1044: Sort the signed owner information from small to large according to the value of the mobile phone number to obtain an owner information table. The mobile phone number sorting in the owner information table is the outbound call task sorting.

[0089] The purpose of steps S1041 to S1044 is to import and create tasks. The system supports importing owner information files in CSV format, and performs a classification algorithm based on K-Means clustering on the data. The owner information is labeled and classified, and the classified data is signed through the elliptic curve encryption mechanism to ensure the tamper-proof nature of the information. Subsequently, the owner information is sorted from small to large according to the value of the mobile phone number, and an owner information table with the order of task execution is generated.

[0090] In some embodiments, converting the speech template to obtain speech voice streaming media includes:

[0091] Step S1061: When the outbound call task is started, the address information is embedded into the speech template;

[0092] Step S1062: Matching the idle seats to obtain the voice media file by converting the speech template embedded with the address information through the TTS conversion method based on deep learning;

[0093] Step S1063: Search for an idle SIP line to convert the sound media file into voice streaming media.

[0094] The purpose of steps S1061 to S1063 is to convert the speech template. After the outbound call task is started, the system embeds the owner's address information into the speech template to form personalized content, then matches the idle seat, and uses the Tacotron2 model based on deep learning to perform TTS (text-to-speech) conversion on the embedded text to generate natural speech. The WaveGlow algorithm can also be used to further optimize the fluency of the speech. Finally, the generated voice file is converted into voice streaming media through the SIP line to provide voice content for the outbound call task.

[0095] In some embodiments, matching an idle seat to obtain a voice media file by converting a speech template embedded with address information through a TTS conversion method based on deep learning includes:

[0096] Step S10621: pre-processing the text of the dialogue template by performing word segmentation, removing stop words and punctuation marks;

[0097] Step S10622: convert the preprocessed speech template text into a phoneme sequence by phoneme extraction. The phoneme extraction formula is:

[0098] S=P(T),

[0099] Among them, T is text, P is the phoneme extraction function, and S is the phoneme sequence;

[0100] Step S10623: Generate a Mel spectrum using the Tacotron2 model;

[0101] Step S10624: Convert the Mel spectrogram into natural speech using the WaveGlow algorithm to obtain a sound media file.

[0102] In some embodiments, matching an agent according to an outbound call task and calling the owner's mobile phone number to communicate with the owner according to the speech voice streaming media, for analyzing the owner's voice conversation content and generating text conversation content, including:

[0103] Step S1081: Associating the voice streaming media with the mobile phone number information;

[0104] Step S1082: Pushing voice streaming media and mobile phone number information to the operator's call system to execute an outbound call task;

[0105] Step S1083: collecting the owner's voice conversation content and generating conversation audio;

[0106] Step S1084: Using a window function to perform frame division and windowing processing on the audio signal of the conversation audio, the window function formula is:

[0107] W=0.54-0.45cos(2πn / (N-1)),

[0108] Where N is the window length;

[0109] Step S1085: extracting audio features of the conversation audio through Mel-frequency cepstral coefficients;

[0110] Step S1086: decoding the conversation audio by connecting the time sequence classification, and mapping the conversation audio into a conversation text sequence;

[0111] Step S1087: Optimize the dialogue text sequence through the Transformer language model to generate text dialogue content.

[0112] The purpose of steps S1081 to S1087 is to execute the outbound call task. The system automatically matches the idle SIP line, pushes the generated voice streaming media to the operator's telephone system, and calls the owner's mobile phone number. During the call, the owner's voice conversation content is collected, and the audio signal is processed by window function frame division and Mel frequency cepstral coefficient extraction. The audio signal is decoded by the connection time classification (CTC) algorithm, and finally the text conversation content is generated.

[0113] In some embodiments, processing the text conversation content, obtaining the text semantic analysis result, and selecting an execution strategy of continuing communication, hanging up, and redialing to complete the outbound call task includes:

[0114] Step S1101: performing word segmentation processing on the text conversation content based on the BERT word segmenter to obtain the conversation text word segmentation;

[0115] Step S1102: Input the dialogue text segmentation into the pre-trained BERT model to generate dialogue text sentence vectors. The dialogue text sentence vector generation formula is:

[0116] V=f(t),

[0117] Where t is the input text, f is the BERT embedding function, and v is the output vector;

[0118] Step S1103: classifying the conversation text sentence vectors through a Softmax classifier to obtain sentiment analysis content;

[0119] Step S1104: classifying the sentiment analysis content based on the Transformer intention classification model and outputting a communication strategy;

[0120] Step S1105: recording the communication content of each outbound call task and the owner's response result, where the response result at least includes continuing communication, hanging up, and calling again later;

[0121] Step S1106: Generate a new owner information table according to the communication strategy and the response result to execute the next outbound call task after completing the outbound call task.

[0122] The purpose of steps S1101 to S1106 is to process text conversations and adjust tasks. The conversation content is processed through word segmentation and text embedding by the BERT model to generate sentence vectors, and sentiment analysis is completed in combination with the Softmax classifier. The Transformer-based intent classification model is used to parse text semantics and generate task strategies, including continuing communication, hanging up, or redialing. At the same time, the task results and user responses are stored to update the owner information table to provide support for subsequent task optimization.

[0123] The embodiment of the present invention provides an intelligent voice outbound calling method system with stable operation, smooth voice outbound calling lines, and clear voice notifications. The outbound calling data is recorded completely, accurately and timely, and accurately stored. The concurrent capacity of executing outbound calling tasks can be set to no less than 10, and later hardware upgrades are supported. The voice data converted from text information has a high accuracy rate.

[0124] Another embodiment of the present invention further provides an intelligent voice outbound calling device, comprising:

[0125] A collection module is used to collect owner information and speech templates and store them in the system database, where the owner information is encrypted information and at least includes mobile phone number information and address information;

[0126] The import module is used to import the owner information and obtain the owner information table to create an outbound call task, wherein the owner information in the owner information table is arranged in ascending order according to the value of the mobile phone number information;

[0127] A conversion module is used to convert speech templates to obtain speech audio and video streaming media;

[0128] An analysis module is used to match an agent according to an outbound call task and call the owner's mobile phone number to communicate with the owner according to the speech voice streaming media, and is used to analyze the owner's voice conversation content and generate text conversation content;

[0129] The processing module is used to process the text conversation content, obtain the text semantic analysis result, and select an execution strategy among continuing communication, hanging up, and redialing to complete the outbound call task.

[0130] Another embodiment of the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program implements the intelligent voice outbound calling method when executed by the processor.

[0131] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the data processing device of the gateway, and uses various interfaces and lines to connect various parts of the data processing device of the entire gateway.

[0132] Another embodiment of the present invention further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the intelligent voice outbound calling method is implemented.

[0133] The memory may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required by at least one function; the data storage area may store data created by the processor, etc. In addition, the memory is preferably but not limited to a high-speed random access memory, for example, it may also be a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may also optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0134] Those skilled in the art can understand that the implementation of all or part of the processes in the above-mentioned embodiments can be completed by instructing related hardware through a computer program, which can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above-mentioned types of memory.

[0135] In summary, the intelligent voice outbound calling method, device, electronic device and storage medium provided by the present invention are used to solve the problems of data security, voice interaction accuracy and insufficient task dynamic adjustment capability existing in the prior art by combining data encryption technology, efficient voice processing algorithm and intelligent task management mechanism, thereby improving the overall performance and application value of the voice outbound calling system. Improve data security: By introducing the AES-256 encryption algorithm, hash message authentication code and public-private key signature technology, the security of user information during collection, storage and use is ensured to prevent data leakage and tampering. Improve task management efficiency: Use K-Means clustering and sorting algorithms to label user information, realize accurate scheduling and priority allocation of tasks, and improve the execution efficiency of outbound calling tasks. Enhance voice processing capabilities: Generate highly natural speech based on deep learning models (such as Tacotron2 and WaveGlow), and combine Mel-frequency cepstral coefficients and window function technology to improve the accuracy and robustness of speech recognition. Realize dynamic task adjustment: Use the BERT model and Transformer intent classification algorithm to perform real-time sentiment analysis and intent judgment on user voice interaction content, and support flexible task strategy adjustment, including continuing communication, hanging up or re-dial, to meet the needs of different scenarios. Through this invention, voice outbound calling technology will be improved in terms of security, intelligence and flexibility, providing an efficient and reliable solution for fields such as electricity bill collection, customer service and marketing, while promoting the application and development of voice technology in more fields.

[0136] Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. A person with ordinary knowledge in the technical field to which the present invention belongs may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be determined by the definition of the claims.

Claims

1. An intelligent voice outbound calling method, characterized in that: include: Collecting owner information and speech templates and storing them in the system database, wherein the owner information is encrypted information and includes at least mobile phone number information and address information; Importing the owner information, obtaining an owner information table to create an outbound call task, wherein the owner information in the owner information table is arranged in ascending order according to the value of the mobile phone number information; Converting the speech template to obtain speech audio streaming media; Matching agents according to outbound call tasks and calling the owner's mobile phone number to communicate with the owner based on speech voice streaming media, which is used to analyze the owner's voice conversation content and generate text conversation content; Process the text conversation content, obtain the text semantic analysis result, and select an execution strategy among continuing communication, hanging up, and redialing to complete the outbound call task.

2. The intelligent voice outbound calling method according to claim 1, characterized in that: The owner information and speech templates are collected and stored in the system database, wherein the owner information is encrypted information and at least includes mobile phone number information and address information, including: Obtain the owner's mobile phone number information and address information through the operator interface, and format them into standardized structured mobile phone number data and standardized structured address data respectively; Perform AES-256 encryption on standard structured mobile phone number data and address data; Use hash message authentication code to verify the integrity of encrypted mobile phone number and address data; The mobile phone number data and address data that have passed data integrity verification are stored in the system database.

3. The intelligent voice outbound calling method according to claim 1, characterized in that: The importing of the owner information to obtain an owner information table to create an outbound call task, wherein the owner information in the owner information table is arranged in ascending order according to the value of the mobile phone number information, includes: Import said owner information in CSV format; Using K-Means based clustering classification on the owner information to obtain the owner information after labeling; Using a public-private key encryption mechanism based on elliptic curve to sign the owner information; The signed owner information is sorted in ascending order according to the value of the mobile phone number to obtain the owner information table. The order of the mobile phone numbers in the owner information table is the order of the outbound call tasks.

4. The intelligent voice outbound calling method according to claim 1, characterized in that: The converting the speech template to obtain speech voice streaming media includes: When the outbound call task is initiated, the address information is embedded into the speech template; Matching idle seats to obtain a sound media file by converting the speech template embedded with the address information through a TTS conversion method based on deep learning; Search for an idle SIP line to convert the sound media file into voice streaming media.

5. The intelligent voice outbound calling method according to claim 4, characterized in that: The matching of idle seats to obtain a sound media file by converting the speech template embedded with the address information through a TTS conversion method based on deep learning includes: Preprocess the text of the speech template by performing word segmentation, removing stop words and punctuation marks; The preprocessed text of the speech template is converted into a phoneme sequence by phoneme extraction, and the phoneme extraction formula is: S=P(T), Among them, T is text, P is the phoneme extraction function, and S is the phoneme sequence; Generate Mel spectra using the Tacotron2 model; The Mel spectrogram is converted into natural speech by using the WaveGlow algorithm to obtain the sound media file.

6. The intelligent voice outbound calling method according to claim 1, characterized in that: The method of matching an agent according to an outbound call task and calling the owner's mobile phone number to communicate with the owner according to the speech voice streaming media is used to analyze the owner's voice conversation content and generate text conversation content, including: Associating the voice streaming media with the mobile phone number information; Pushing the voice streaming media and the mobile phone number information to the operator's call system to perform an outbound call task; Collecting the owner's voice conversation content and generating conversation audio; The audio signal of the conversation audio is framed and windowed using a window function, and the window function formula is: W=0.54-0.45cos(2πn / (N-1)), Where N is the window length; Extracting audio features of the conversation audio through Mel-frequency cepstral coefficients; Decoding the conversation audio by connecting the timing classification, and mapping the conversation audio into a conversation text sequence; The dialogue text sequence is optimized through a Transformer language model to generate text dialogue content.

7. The intelligent voice outbound calling method according to claim 1, characterized in that: The processing of the text conversation content, obtaining a text semantic analysis result and selecting an execution strategy of continuing communication, hanging up, and redialing to complete the outbound call task includes: Performing word segmentation processing on the text conversation content based on the BERT word segmenter to obtain the conversation text word segmentation; The dialogue text segmentation is input into the pre-trained BERT model to generate a dialogue text sentence vector. The dialogue text sentence vector generation formula is: V=f(t), Where t is the input text, f is the BERT embedding function, and v is the output vector; Classifying the conversation text sentence vectors by a Softmax classifier to obtain sentiment analysis content; Classify the sentiment analysis content based on the Transformer intent classification model and output the communication strategy; Record the communication content of each outbound call task and the owner's response result, and the response result at least includes continue communication, hang up, and call again later; A new owner information table is generated according to the communication strategy and the response result to execute the next outbound calling task after the outbound calling task is completed.

8. An intelligent voice outbound calling device, characterized in that: include: A collection module, used to collect owner information and speech templates and store them in a system database, wherein the owner information is encrypted information and includes at least mobile phone number information and address information; An import module, used to import the owner information, obtain an owner information table to create an outbound call task, wherein the owner information in the owner information table is arranged in ascending order according to the value of the mobile phone number information; A conversion module, used for converting the speech template to obtain speech voice streaming media; An analysis module is used to match an agent according to an outbound call task and call the owner's mobile phone number to communicate with the owner according to the speech voice streaming media, and is used to analyze the owner's voice conversation content and generate text conversation content; The processing module is used to process the text conversation content, obtain the text semantic analysis result and select an execution strategy among continuing communication, hanging up and redialing to complete the outbound call task.

9. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the intelligent voice outbound calling method as claimed in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the intelligent voice outbound calling method as described in any one of claims 1 to 7 is implemented.

Images