Voice task processing method and device based on artificial intelligence, equipment and medium

Abstract

The invention belongs to the technical field of artificial intelligence, and relates to a voice task processing method and device based on artificial intelligence, computer equipment and a storage medium, and the method comprises the steps: obtaining multi-dimensional voice data; preprocessing the multi-dimensional voice data to obtain original voice data; performing noise superposition processing on the original voice data to obtain first voice data; performing spoken-language feature annotation adding processing on the original voice data to obtain second voice data; based on the construction strategy, training and optimizing a pre-trained deep learning model by using the original voice data, the first voice data and the second voice data to obtain a target voice recognition model meeting a preset performance requirement; and processing the voice interaction task based on the target voice recognition model. In addition, the target speech recognition model can be stored in the block chain. Based on the use of the target speech recognition model, the speech recognition accuracy is improved, and the service quality and the user experience are improved.

Description

Technical Field

[0001] The present application relates to the fields of artificial intelligence development technology, financial technology, and medical health, and in particular to artificial intelligence-based voice task processing methods, devices, computer equipment, and storage media. Background Art

[0002] In today's rapidly developing insurance industry, with the advancement of technology and the continuous improvement of customer needs, many insurance companies are actively exploring and practicing the use of advanced information technology to optimize customer service processes and improve user experience. Among them, voice-based virtual assistants, as an innovative customer service model, have received widespread attention and application in the financial and medical fields in recent years. This type of virtual assistant usually integrates automatic speech recognition (ASR, Automatic Speech Recognition) technology and text-to-speech (TTS, Text-To-Speech) technology, aiming to provide customers with convenient and efficient consulting services and business process guidance in financial service scenarios or medical service scenarios through natural language interaction.

[0003] Although voice-based virtual assistants can greatly improve customer experience and service efficiency in theory, they face many technical challenges and shortcomings in actual application. Specifically, when dealing with complex voice interaction tasks, such as those with complex environmental noise or more colloquial expressions, existing voice recognition technology cannot accurately capture and interpret customer voice commands, resulting in low accuracy of voice recognition, which in turn leads to poor service quality and poor user experience. Summary of the invention

[0004] The purpose of the embodiments of the present application is to propose a voice task processing method, device, computer equipment and storage medium based on artificial intelligence to solve the technical problem that the existing voice recognition technology has low voice recognition accuracy when processing complex voice interaction tasks, which in turn leads to poor service quality and poor user experience.

[0005] In order to solve the above technical problems, the embodiment of the present application provides a voice task processing method based on artificial intelligence, which adopts the following technical solution:

[0006] Obtain pre-built multi-dimensional voice data;

[0007] Preprocessing the multi-dimensional voice data to obtain corresponding original voice data;

[0008] Performing noise superposition processing on the original voice data to obtain corresponding first voice data;

[0009] Performing a process of adding a colloquial feature annotation to the original voice data to obtain corresponding second voice data;

[0010] Call pre-trained deep learning models;

[0011] Based on a preset construction strategy, the deep learning model is trained and optimized using the original voice data, the first voice data, and the second voice data to obtain a target voice recognition model that meets preset performance requirements;

[0012] The speech interaction task to be processed is processed based on the target speech recognition model.

[0013] Furthermore, the step of performing noise superposition processing on the original voice data to obtain the corresponding first voice data specifically includes:

[0014] Obtain pre-collected environmental noise samples;

[0015] Get the preset signal-to-noise ratio;

[0016] superimposing the environmental noise sample onto the original voice data based on the signal-to-noise ratio to obtain superimposed original voice data;

[0017] The superimposed original voice data is used as the first voice data.

[0018] Furthermore, the step of performing a process of adding a label of spoken features to the original voice data to obtain corresponding second voice data specifically includes:

[0019] Call the preset annotation tool;

[0020] Annotating the spoken features in the original speech data based on the annotation tool to obtain corresponding annotation information;

[0021] Adding the annotation information to the original voice data accordingly to obtain the added original voice data;

[0022] The added original voice data is used as the second voice data.

[0023] Furthermore, the step of preprocessing the multi-dimensional voice data to obtain corresponding original voice data specifically includes:

[0024] Performing denoising processing on the multi-dimensional voice data to obtain corresponding first multi-dimensional voice data;

[0025] Segmenting the first multi-dimensional voice data to obtain corresponding second multi-dimensional voice data;

[0026] The original voice data is generated based on the second multi-dimensional voice data.

[0027] Furthermore, the step of obtaining pre-constructed multi-dimensional voice data specifically includes:

[0028] Obtain pre-collected real voice data;

[0029] Get pre-built synthetic speech data;

[0030] Integrate the real voice data and the synthesized voice data to obtain corresponding integrated voice data;

[0031] The integrated voice data is used as the multi-dimensional voice data.

[0032] Furthermore, the step of obtaining pre-constructed synthesized speech data specifically includes:

[0033] Obtaining a preset initial text containing specified speaker characteristics; wherein the specified speaker characteristics at least include age, gender and accent;

[0034] Setting a corresponding designated speech parameter based on the designated speaker feature and the initial text;

[0035] Based on the specified voice parameters, converting the initial text into corresponding voice data using a preset conversion strategy;

[0036] The synthesized speech data is generated based on the speech data.

[0037] Furthermore, after the step of training and optimizing the deep learning model based on the preset construction strategy using the original voice data, the first voice data, and the second voice data to obtain a target voice recognition model that meets the preset performance requirements, the method further includes:

[0038] Determining a designated storage mode corresponding to the target speech recognition model;

[0039] Performing storage processing on the target speech recognition model based on the specified storage method;

[0040] After completing the storage process of the target speech recognition model, the target speech recognition model is deployed.

[0041] In order to solve the above technical problems, the embodiment of the present application also provides a voice task processing device based on artificial intelligence, which adopts the following technical solution:

[0042] An acquisition module, used to acquire pre-built multi-dimensional voice data;

[0043] A preprocessing module, used to preprocess the multi-dimensional voice data to obtain corresponding original voice data;

[0044] A superposition module, used for performing noise superposition processing on the original voice data to obtain corresponding first voice data;

[0045] An adding module, used for performing a marking and adding process of colloquial features on the original voice data to obtain corresponding second voice data;

[0046] The calling module is used to call the pre-trained deep learning model;

[0047] A first processing module, configured to train and optimize the deep learning model using the original voice data, the first voice data, and the second voice data based on a preset construction strategy to obtain a target voice recognition model that meets preset performance requirements;

[0048] The second processing module is used to process the speech interaction task to be processed based on the target speech recognition model.

[0049] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0050] Obtain pre-built multi-dimensional voice data;

[0051] Preprocessing the multi-dimensional voice data to obtain corresponding original voice data;

[0052] Performing noise superposition processing on the original voice data to obtain corresponding first voice data;

[0053] Performing a process of adding a colloquial feature annotation to the original voice data to obtain corresponding second voice data;

[0054] Call pre-trained deep learning models;

[0055] Based on a preset construction strategy, the deep learning model is trained and optimized using the original voice data, the first voice data, and the second voice data to obtain a target voice recognition model that meets preset performance requirements;

[0056] The speech interaction task to be processed is processed based on the target speech recognition model.

[0057] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0058] Obtain pre-built multi-dimensional voice data;

[0059] Preprocessing the multi-dimensional voice data to obtain corresponding original voice data;

[0060] Performing noise superposition processing on the original voice data to obtain corresponding first voice data;

[0061] Performing a process of adding a colloquial feature annotation to the original voice data to obtain corresponding second voice data;

[0062] Call pre-trained deep learning models;

[0063] Based on a preset construction strategy, the deep learning model is trained and optimized using the original voice data, the first voice data, and the second voice data to obtain a target voice recognition model that meets preset performance requirements;

[0064] The speech interaction task to be processed is processed based on the target speech recognition model.

[0065] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0066] The present application first obtains pre-constructed multi-dimensional voice data; and pre-processes the multi-dimensional voice data to obtain corresponding original voice data; then performs noise superposition processing on the original voice data to obtain corresponding first voice data; and performs annotation and addition processing of spoken features on the original voice data to obtain corresponding second voice data; then calls a pre-trained deep learning model; subsequently, based on a preset construction strategy, uses the original voice data, the first voice data, and the second voice data to train and optimize the deep learning model to obtain a target voice recognition model that meets preset performance requirements; finally, the voice interaction task to be processed is processed based on the target voice recognition model. The present application obtains original voice data by preprocessing pre-constructed multi-dimensional voice data, then performs noise superposition processing on the original voice data to obtain first voice data, and performs annotation and addition processing of spoken features on the original voice data to obtain second voice data. Then, based on the construction strategy, the deep learning model is trained and optimized using the original voice data, the first voice data, and the second voice data to obtain a target voice recognition model that meets preset performance requirements and has powerful noise elimination capabilities and spoken content understanding capabilities, thereby effectively improving the voice recognition accuracy of the virtual assistant in complex environments, and enhancing its understanding and response capabilities to spoken commands. Subsequently, through the use of the target voice recognition model, it is possible to efficiently process complex voice interaction tasks and provide accurate and timely responses, thereby improving the accuracy of voice recognition and helping to improve service quality and user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0068] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0069] Figure 2 A flowchart of an embodiment of a method for processing speech tasks based on artificial intelligence according to the present application;

[0070] Figure 3 is a structural schematic diagram of an embodiment of an artificial intelligence-based speech task processing device according to the present application;

[0071] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0072] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0073] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0074] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0075] like Figure 1As shown, the system architecture 100 may include a terminal device 101, a network 102 and a server 103. The terminal device 101 may be a laptop 1011, a tablet computer 1012 or a mobile phone 1013. The network 102 is used to provide a medium for a communication link between the terminal device 101 and the server 103. The network 102 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0076] The user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0077] The terminal device 101 can be any electronic device with a display screen and supporting web browsing. In addition to the laptop computer 1011, tablet computer 1012 or mobile phone 1013, the terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV), a laptop computer, a desktop computer, etc.

[0078] The server 103 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal device 101 .

[0079] It should be noted that the artificial intelligence-based voice task processing method provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the artificial intelligence-based voice task processing device is generally arranged in the server / terminal device.

[0080] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0081] Continue to refer Figure 2, showing a flow chart of an embodiment of the voice task processing method based on artificial intelligence according to the present application. According to different requirements, the order of the steps in the flowchart can be changed, and some steps can be omitted. The voice task processing method based on artificial intelligence provided in the embodiment of the present application can be applied to any scenario that requires voice task processing, and the voice task processing method based on artificial intelligence can be applied to products in these scenarios, for example, voice interaction task processing in the field of finance and insurance. The voice task processing method based on artificial intelligence comprises the following steps:

[0082] Step S201, obtaining pre-constructed multi-dimensional voice data.

[0083] In this embodiment, the electronic device (e.g. Figure 1 The server / terminal device shown in the figure) can obtain multi-dimensional voice data through a wired connection or a wireless connection. It should be pointed out that the above-mentioned wireless connection method may include but is not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future. The executor of the present application may be a voice task processing system, or a virtual assistant, which can be referred to as a system for short. The present application can be applied to application scenarios such as the insurance industry, using the system to process voice interaction tasks, thereby providing customers with efficient and high-quality customer service and improving customer satisfaction. Among them, the specific implementation process of obtaining pre-constructed multi-dimensional voice data will be further described in detail in the subsequent specific embodiments of the present application, and will not be elaborated on here.

[0084] Step S202: pre-process the multi-dimensional voice data to obtain corresponding original voice data.

[0085] In this embodiment, the above-mentioned preprocessing of the multi-dimensional voice data to obtain the specific implementation process of the corresponding original voice data will be further described in detail in the subsequent specific embodiments of this application, and will not be elaborated in detail here.

[0086] Step S203: performing noise superposition processing on the original voice data to obtain corresponding first voice data.

[0087] In this embodiment, the above-mentioned specific implementation process of performing noise superposition processing on the original voice data to obtain the corresponding first voice data will be further described in detail in subsequent specific embodiments of the present application, and will not be elaborated in detail here.

[0088] Step S204: annotate and add spoken features to the original voice data to obtain corresponding second voice data.

[0089] In this embodiment, the specific implementation process of the above-mentioned annotation and addition processing of spoken features to the original voice data to obtain the corresponding second voice data will be further described in detail in the subsequent specific embodiments of this application, and will not be elaborated in detail here.

[0090] Step S205, calling the pre-trained deep learning model.

[0091] In this embodiment, the selection of the above-mentioned deep learning model is not specifically limited and can be determined according to actual business needs. For example, a BERT or GPT series model can be used. These models have been pre-trained on a large amount of text data and have strong language comprehension capabilities. Among them, the deep learning model can be pre-trained on a large-scale text corpus to obtain the initial language comprehension ability of the deep learning model.

[0092] Step S206, based on a preset construction strategy, the deep learning model is trained and optimized using the original voice data, the first voice data, and the second voice data to obtain a target speech recognition model that meets preset performance requirements.

[0093] In this embodiment, the deep learning model can be fine-tuned by using the original voice data, and the goal of fine-tuning is to make the model better adapt to specific voice interaction tasks, such as command understanding and response generation. Then, the deep learning model is trained end-to-end using the original voice data, the first voice data, and the second voice data, by constructing an environmental noise elimination task corresponding to the original voice data and the first voice data, and constructing a spoken content understanding task corresponding to the second voice data, and combining the environmental noise elimination task with the spoken content understanding task to train the deep learning model end-to-end. During the training process, the loss functions of the two tasks are optimized at the same time to improve the overall performance of the model. And by introducing a multi-task learning mechanism, the model can better understand and respond to spoken content while processing noisy environment voice data. Subsequently, suitable evaluation indicators are used to evaluate the performance of the model, such as speech recognition accuracy, speech enhancement effect (such as signal-to-noise ratio improvement), spoken content understanding accuracy, etc. According to the evaluation results, the model is further optimized and adjusted. Among them, different model architectures, parameter settings or training methods can be tried to improve the performance and generalization ability of the model. Finally, the performance of the model is continuously optimized through multiple iterative training. In each iteration, adjustments and improvements are made based on the evaluation results and feedback until a satisfactory performance level is reached, and the final model is used as the target speech recognition model.

[0094] The environmental noise elimination task includes: using the noisy environment speech data (first speech data) and the corresponding clean speech data (original speech data) to train the model. By introducing a noise elimination algorithm based on deep learning (such as a masking-based method or a mapping-based method), the model can learn how to extract clean speech signals from a noisy environment. During the training process, the parameters of the model are continuously optimized to improve its noise elimination performance.

[0095] The task of understanding spoken content includes: training the model using the second speech data annotated with spoken features. By introducing natural language processing tasks (such as named entity recognition, sentiment analysis, etc.), the model can learn the characteristics and rules of spoken content. During the training process, the parameters of the model are continuously optimized to improve its understanding and response accuracy to spoken content. Among them, error tolerance mechanisms are further designed and implemented, such as automatic correction and ignoring errors, so that the model can automatically correct or ignore these errors when it recognizes spoken errors, so as to improve the accuracy of understanding and response.

[0096] Step S207: Process the speech interaction task to be processed based on the target speech recognition model.

[0097] In this embodiment, the speech interaction task to be processed can be input into the target speech recognition model so as to be processed by the target speech recognition model. The target speech recognition model can efficiently process complex speech interaction tasks, provide accurate and timely responses, and continuously optimize to improve user experience.

[0098] The present application first obtains pre-constructed multi-dimensional voice data; and pre-processes the multi-dimensional voice data to obtain corresponding original voice data; then performs noise superposition processing on the original voice data to obtain corresponding first voice data; and performs annotation and addition processing of spoken features on the original voice data to obtain corresponding second voice data; then calls a pre-trained deep learning model; subsequently, based on a preset construction strategy, uses the original voice data, the first voice data, and the second voice data to train and optimize the deep learning model to obtain a target voice recognition model that meets preset performance requirements; finally, the voice interaction task to be processed is processed based on the target voice recognition model. The present application obtains original voice data by preprocessing pre-constructed multi-dimensional voice data, then performs noise superposition processing on the original voice data to obtain first voice data, and performs annotation and addition processing of spoken features on the original voice data to obtain second voice data. Then, based on the construction strategy, the deep learning model is trained and optimized using the original voice data, the first voice data, and the second voice data to obtain a target voice recognition model that meets preset performance requirements and has powerful noise elimination capabilities and spoken content understanding capabilities, thereby effectively improving the voice recognition accuracy of the virtual assistant in complex environments, and enhancing its understanding and response capabilities to spoken commands. Subsequently, through the use of the target voice recognition model, it is possible to efficiently process complex voice interaction tasks and provide accurate and timely responses, thereby improving the accuracy of voice recognition and helping to improve service quality and user experience.

[0099] In some optional implementations, step S203 includes the following steps:

[0100] Get pre-collected samples of ambient noise.

[0101] In this embodiment, various environmental noise samples are collected in advance, such as street noise, office background noise, etc., to ensure the diversity and representativeness of environmental noise samples. The environmental noise samples provide a basis for subsequent noise superposition and model robustness training.

[0102] Get the preset signal-to-noise ratio.

[0103] In this embodiment, there is no specific limitation on the selection of the above signal-to-noise ratio, and it can be set according to actual business needs.

[0104] The environmental noise sample is superimposed on the original voice data based on the signal-to-noise ratio to obtain superimposed original voice data.

[0105] In this embodiment, the collected environmental noise samples are superimposed on the original voice data at different signal-to-noise ratios (SNRs) to simulate voice interaction scenarios in different noise environments.

[0106] The superimposed original voice data is used as the first voice data.

[0107] In this embodiment, in the subsequent training process of the target speech recognition model, the robustness of the model in a noisy environment can be enhanced by inputting speech data samples superimposed with noise into the model.

[0108] The present application obtains pre-collected environmental noise samples; then obtains a preset signal-to-noise ratio; then, based on the signal-to-noise ratio, superimposes the environmental noise samples on the original voice data to obtain superimposed original voice data; and subsequently uses the superimposed original voice data as the first voice data. The present application obtains pre-collected environmental noise samples; then, based on the obtained signal-to-noise ratio, superimposes the environmental noise samples on the original voice data, thereby efficiently and accurately constructing the corresponding first voice data, improving the efficiency of constructing the first voice data, facilitating the subsequent application of the first voice data to the model construction process of the target voice recognition model, and can effectively enhance the robustness of the model in a noisy environment.

[0109] In some optional implementations of this embodiment, step S204 includes the following steps:

[0110] Invoke the preset annotation tool.

[0111] In this embodiment, the above-mentioned marking tool is a tool with information marking function.

[0112] The colloquial features in the original speech data are annotated based on the annotation tool to obtain corresponding annotation information.

[0113] In this embodiment, the spoken features in the original speech data, including features such as repetition, filler words, grammatical errors, etc., can be annotated by using an annotation tool to obtain corresponding annotation information.

[0114] The annotation information is added to the original voice data to obtain the added original voice data.

[0115] In this embodiment, the obtained annotation information can be further added to the original voice data by using an annotation tool, so as to obtain the added original voice data as the final second voice data.

[0116] The added original voice data is used as the second voice data.

[0117] In this embodiment, the annotated spoken features can provide important information for subsequent model training of the target speech recognition model, helping the model to better understand and process spoken content.

[0118] The present application calls a preset annotation tool; then annotates the spoken features in the original voice data based on the annotation tool to obtain corresponding annotation information; then adds the annotation information to the original voice data to obtain the added original voice data; and subsequently uses the added original voice data as the second voice data. The present application annotates the spoken features in the original voice data based on the use of an annotation tool to obtain corresponding annotation information, and then adds the annotation information to the original voice data, thereby achieving rapid and accurate completion of the annotation and adding processing of the spoken features of the original voice data, improving the processing efficiency and processing intelligence of the annotation and adding processing, facilitating the use of the second voice data based on the annotated spoken features, and providing important information for the subsequent model training of the target voice recognition model, thereby helping the model to better understand and process the spoken content.

[0119] In some optional implementations, step S202 includes the following steps:

[0120] The multi-dimensional voice data is subjected to denoising processing to obtain corresponding first multi-dimensional voice data.

[0121] In this embodiment, the background noise in the multi-dimensional voice data can be removed by using signal processing techniques such as bandpass filtering, spectral subtraction, etc., and the denoising effect can be evaluated to ensure that the voice quality is not affected. The denoised voice data helps to improve the accuracy of the model in voice recognition and language understanding.

[0122] The first multi-dimensional voice data is segmented to obtain corresponding second multi-dimensional voice data.

[0123] In this embodiment, the long speech segment can be segmented into shorter speech segments according to the pauses and intonation changes of the first multi-dimensional speech data, and each segment can be ensured to contain complete speech information. The segmented speech segments are convenient for model training and evaluation, and also help improve the recognition efficiency of the model.

[0124] The original voice data is generated based on the second multi-dimensional voice data.

[0125] The present application performs denoising on the multi-dimensional voice data to obtain corresponding first multi-dimensional voice data; then performs segmentation on the first multi-dimensional voice data to obtain corresponding second multi-dimensional voice data; and subsequently generates the original voice data based on the second multi-dimensional voice data. The present application performs denoising and segmentation on the multi-dimensional voice data, thereby achieving efficient and accurate pre-processing of the multi-dimensional voice data, and effectively ensuring the data quality of the obtained original voice data.

[0126] In some optional implementations, step S201 includes the following steps:

[0127] Get pre-collected real speech data.

[0128] In this embodiment, voice samples in real environments, i.e., real voice data, are collected in advance, and these real voice data can come from customer service recordings or public voice databases. Among them, the real voice data provides voice samples in real environments for the model, which helps the model better adapt to actual application scenarios.

[0129] Get pre-built synthetic speech data.

[0130] In this embodiment, the specific implementation process of obtaining the pre-constructed synthesized speech data will be further described in detail in the subsequent specific embodiments of this application, and will not be elaborated on here. In addition, the above-mentioned synthesized speech data provides a rich voice sample for subsequent model training, which helps the model learn the differences in the characteristics of different speakers.

[0131] The real voice data and the synthesized voice data are integrated to obtain corresponding integrated voice data.

[0132] In this embodiment, the integrated voice data includes the real voice data and the synthesized voice data.

[0133] The integrated voice data is used as the multi-dimensional voice data.

[0134] The present application obtains pre-collected real voice data; and obtains pre-constructed synthetic voice data; and then integrates and processes the real voice data and the synthetic voice data to obtain corresponding integrated voice data; and then uses the integrated voice data as the multi-dimensional voice data. The present application obtains pre-collected real voice data; and obtains pre-constructed synthetic voice data; and then integrates and processes the real voice data and the synthetic voice data, so that the corresponding multi-dimensional voice data can be constructed efficiently and accurately, which is beneficial for providing rich voice samples for the subsequent model training of the target voice recognition model based on the obtained multi-dimensional voice data.

[0135] In some optional implementations of this embodiment, the obtaining of pre-constructed synthesized speech data includes the following steps:

[0136] Obtain a preset initial text containing specified speaker characteristics; wherein the specified speaker characteristics at least include age, gender and accent.

[0137] In this embodiment, the above-mentioned specified speaker characteristics refer to diversified speaker characteristics, which may at least include age, gender and accent. Exemplary, age includes: children, teenagers, adults (youth, middle-aged, elderly). Gender includes: male, female. Accents include: standard Mandarin, dialects (such as Cantonese, Sichuanese, Hakka), international accents (such as Indian English, British English, American English, etc.). In addition, representative and diverse initial texts can be written based on the determined speaker characteristics. The text content can be dialogues, monologues, stories, news, etc., ensuring that different language styles and emotional expressions are covered. For example, write a story about animals for children, write a monologue about memories for the elderly, write daily conversations for people with different accents, etc.

[0138] A designated speech parameter corresponding to the initial text is set based on the designated speaker feature.

[0139] In this embodiment, the above-mentioned specified voice parameters specifically refer to TTS parameters. Appropriate TTS parameters can be set according to the text content of the initial text and the speaker characteristics. For example, a faster speech rate and a higher pitch are set for children's stories, and a slower speech rate and a lower pitch are set for the monologue of the elderly. In addition, different speaker IDs are selected to generate voices with different accents and gender characteristics.

[0140] Based on the specified voice parameters, the initial text is converted into corresponding voice data using a preset conversion strategy.

[0141] In this embodiment, the above conversion strategy specifically refers to the processing of converting the initial text into speech by a TTS model based on TTS (text to speech) technology. The TTS model can be a model based on deep learning that can generate speech with specific speaker characteristics based on tags in the text. The initial text can be input into the TTS model, and then the TT model can generate speech data with different speaker characteristics according to the set TTS parameters.

[0142] The synthesized speech data is generated based on the speech data.

[0143] In this embodiment, the generated voice data can be listened to to check whether the sound quality, speaking speed, pitch, etc. meet expectations. If they meet expectations, the voice data is used as the final synthesized voice data.

[0144] The present application obtains a preset initial text containing specified speaker features; wherein the specified speaker features include at least age, gender and accent; then, based on the specified speaker features and the initial text, the specified speech parameters corresponding to the settings are set; then, based on the specified speech parameters, the initial text is converted into corresponding speech data using a preset conversion strategy; and then, based on the speech data, the synthesized speech data is generated. The present application obtains a preset initial text containing specified speaker features, then, based on the specified speaker features and the initial text, the specified speech parameters corresponding to the settings are set, and based on the specified speech parameters, the initial text is converted into corresponding speech data using a preset conversion strategy, and then, based on the obtained speech data, the corresponding synthesized speech data with diversified speaker features can be generated quickly and accurately, thereby providing rich speech samples for the subsequent model training of the target speech recognition model. The synthesized speech data helps the target speech recognition model to better understand and process the speech inputs of different speakers, thereby improving the performance and user experience of the target speech recognition.

[0145] In some optional implementations of this embodiment, after step S206, the electronic device may further perform the following steps:

[0146] Determine a designated storage method corresponding to the target speech recognition model.

[0147] In this embodiment, there is no limitation on the determination of the above-mentioned designated storage method, which can be selected according to actual business needs. For example, local database storage, file system storage, cloud server storage, blockchain storage, etc. can be used.

[0148] The target speech recognition model is stored based on the specified storage method.

[0149] In this embodiment, the storage process of the target speech recognition model is completed by determining a designated storage medium corresponding to the above-mentioned designated storage method and then storing the target speech recognition model in the designated storage medium.

[0150] After completing the storage process of the target speech recognition model, the target speech recognition model is deployed.

[0151] In this embodiment, the target speech recognition model can be deployed in the system to handle complex speech interaction tasks. In application scenarios such as the insurance industry, the target speech recognition model can be used to provide efficient and high-quality customer service and improve customer satisfaction.

[0152] The present application determines a designated storage method corresponding to the target speech recognition model; then stores the target speech recognition model based on the designated storage method; and after completing the storage process of the target speech recognition model, deploys the target speech recognition model. After completing the construction of the target speech recognition model, the present application will also intelligently use the designated storage method to store the target speech recognition model, thereby ensuring the security of the generated target speech recognition model. In addition, the target speech recognition model will also be deployed so that the subsequent system can use the target speech recognition model to process the speech interaction tasks, thereby providing customers with efficient and high-quality customer service, which is conducive to improving customer satisfaction.

[0153] In some optional implementations, the user information obtained is subject to the user's consent and complies with relevant laws and policies.

[0154] In addition, the non-Company software tools or components that appear in the embodiments of the present application are merely examples and do not represent actual use.

[0155] In addition, this application simulates voice interaction scenarios in the real world to comprehensively evaluate and optimize the ability of virtual assistants to handle diverse speaker characteristics, environmental noise, and spoken content. It not only improves the accuracy of speech recognition of virtual assistants in complex environments, but also enhances their ability to understand and respond to spoken instructions, thereby significantly improving the efficiency and quality of customer service. In the insurance business, especially in the fields of property insurance and health insurance, customers often consult and make claims through telephone or voice assistants. This application can significantly improve the service capabilities of these virtual assistants in noisy environments or when facing customers with different accents and language habits. By improving the accuracy of speech recognition and the ability to understand spoken instructions, this application not only enhances the efficiency and quality of customer service, but also improves customer satisfaction and reduces the workload of manual customer service.

[0156] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0157] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned target speech recognition model, the above-mentioned target speech recognition model can also be stored in a node of a blockchain.

[0158] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0159] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0160] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0161] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions 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 aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0162] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0163] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a speech task processing device based on artificial intelligence, and the device embodiment is similar to Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0164] like Figure 3 As shown, the artificial intelligence-based speech task processing device 300 described in this embodiment includes: an acquisition module 301, a preprocessing module 302, a superposition module 303, an adding module 304, a calling module 305, a first processing module 306 and a second processing module 307.

[0165] in:

[0166] An acquisition module 301 is used to acquire pre-built multi-dimensional voice data;

[0167] A preprocessing module 302 is used to preprocess the multi-dimensional voice data to obtain corresponding original voice data;

[0168] A superposition module 303 is used to perform noise superposition processing on the original voice data to obtain corresponding first voice data;

[0169] An adding module 304 is used to perform a labeling and adding process of colloquial features on the original voice data to obtain corresponding second voice data;

[0170] A calling module 305 is used to call a pre-trained deep learning model;

[0171] A first processing module 306 is used to train and optimize the deep learning model using the original voice data, the first voice data, and the second voice data based on a preset construction strategy to obtain a target voice recognition model that meets preset performance requirements;

[0172] The second processing module 307 is used to process the speech interaction task to be processed based on the target speech recognition model.

[0173] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based voice task processing method in the aforementioned implementation mode, and are not described in detail here.

[0174] In some optional implementations of this embodiment, the superposition module 303 includes:

[0175] A first acquisition submodule is used to acquire pre-collected environmental noise samples;

[0176] A second acquisition submodule is used to acquire a preset signal-to-noise ratio;

[0177] A superposition submodule, configured to superimpose the environmental noise sample onto the original voice data based on the signal-to-noise ratio to obtain the superimposed original voice data;

[0178] The first determining submodule is configured to use the superimposed original voice data as the first voice data.

[0179] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based voice task processing method in the aforementioned implementation mode, and are not described in detail here.

[0180] In some optional implementations of this embodiment, the adding module 304 includes:

[0181] Call submodule, used to call the preset annotation tool;

[0182] A labeling submodule, used for labeling the colloquial features in the original speech data based on the labeling tool to obtain corresponding labeling information;

[0183] An adding submodule, used for adding the annotation information to the original voice data to obtain the added original voice data;

[0184] The second determining submodule is used to use the added original voice data as the second voice data.

[0185] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based voice task processing method in the aforementioned implementation mode, and are not described in detail here.

[0186] In some optional implementations of this embodiment, the preprocessing module 302 includes:

[0187] A denoising submodule, used for performing denoising processing on the multi-dimensional voice data to obtain corresponding first multi-dimensional voice data;

[0188] A segmentation submodule, configured to segment the first multi-dimensional voice data to obtain corresponding second multi-dimensional voice data;

[0189] A generating submodule is used to generate the original voice data based on the second multi-dimensional voice data.

[0190] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based voice task processing method in the aforementioned implementation mode, and are not described in detail here.

[0191] In some optional implementations of this embodiment, the acquisition module 301 includes:

[0192] The third acquisition submodule is used to acquire the real voice data collected in advance;

[0193] The fourth acquisition submodule is used to acquire pre-constructed synthetic speech data;

[0194] An integration submodule, used for integrating the real voice data and the synthesized voice data to obtain corresponding integrated voice data;

[0195] The third determining submodule is configured to use the integrated voice data as the multi-dimensional voice data.

[0196] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based voice task processing method in the aforementioned implementation mode, and are not described in detail here.

[0197] In some optional implementations of this embodiment, the fourth acquisition submodule includes:

[0198] An acquisition unit, configured to acquire a preset initial text containing specified speaker characteristics; wherein the specified speaker characteristics at least include age, gender and accent;

[0199] A setting unit, configured to set corresponding designated speech parameters based on the designated speaker characteristics and the initial text;

[0200] A conversion unit, configured to convert the initial text into corresponding voice data using a preset conversion strategy based on the specified voice parameters;

[0201] A generating unit is used to generate the synthesized speech data based on the speech data.

[0202] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based voice task processing method in the aforementioned implementation mode, and are not described in detail here.

[0203] In some optional implementations of this embodiment, the artificial intelligence-based voice task processing device further includes:

[0204] A determination module, used to determine a designated storage mode corresponding to the target speech recognition model;

[0205] A storage module, used for storing the target speech recognition model based on the specified storage method;

[0206] A deployment module is used to deploy the target speech recognition model after completing the storage processing of the target speech recognition model.

[0207] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based voice task processing method in the aforementioned implementation mode, and are not described in detail here.

[0208] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0209] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with components 41-43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0210] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0211] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the computer device 4. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions of the speech task processing method based on artificial intelligence, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0212] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run computer-readable instructions or process data stored in the memory 41, such as computer-readable instructions for running the artificial intelligence-based voice task processing method.

[0213] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0214] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0215] In an embodiment of the present application, original voice data is obtained by preprocessing pre-constructed multi-dimensional voice data, and then the original voice data is subjected to noise superposition processing to obtain first voice data, and the original voice data is subjected to spoken feature annotation and addition processing to obtain second voice data. Then, based on the construction strategy, the deep learning model is trained and optimized using the original voice data, the first voice data, and the second voice data to obtain a target voice recognition model that meets preset performance requirements and has powerful noise elimination capabilities and spoken content understanding capabilities, thereby effectively improving the voice recognition accuracy of the virtual assistant in complex environments, and enhancing its understanding and response capabilities to spoken commands. Subsequently, through the use of the target voice recognition model, it is possible to efficiently process complex voice interaction tasks, and provide accurate and timely responses, thereby improving the accuracy of voice recognition, and helping to improve service quality and user experience.

[0216] The present application also provides another implementation, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the above-mentioned artificial intelligence-based speech task processing method.

[0217] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0218] In an embodiment of the present application, original voice data is obtained by preprocessing pre-constructed multi-dimensional voice data, and then the original voice data is subjected to noise superposition processing to obtain first voice data, and the original voice data is subjected to spoken feature annotation and addition processing to obtain second voice data. Then, based on the construction strategy, the deep learning model is trained and optimized using the original voice data, the first voice data, and the second voice data to obtain a target voice recognition model that meets preset performance requirements and has powerful noise elimination capabilities and spoken content understanding capabilities, thereby effectively improving the voice recognition accuracy of the virtual assistant in complex environments, and enhancing its understanding and response capabilities to spoken commands. Subsequently, through the use of the target voice recognition model, it is possible to efficiently process complex voice interaction tasks, and provide accurate and timely responses, thereby improving the accuracy of voice recognition, and helping to improve service quality and user experience.

[0219] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0220] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A speech task processing method based on artificial intelligence, characterized in that: The steps include: Obtain pre-built multi-dimensional voice data; Preprocessing the multi-dimensional voice data to obtain corresponding original voice data; Performing noise superposition processing on the original voice data to obtain corresponding first voice data; Performing a process of adding a colloquial feature annotation to the original voice data to obtain corresponding second voice data; Call pre-trained deep learning models; Based on a preset construction strategy, the deep learning model is trained and optimized using the original voice data, the first voice data, and the second voice data to obtain a target voice recognition model that meets preset performance requirements; The speech interaction task to be processed is processed based on the target speech recognition model.

2. The method for processing speech tasks based on artificial intelligence according to claim 1, characterized in that: The step of performing noise superposition processing on the original voice data to obtain the corresponding first voice data specifically includes: Obtain pre-collected environmental noise samples; Get the preset signal-to-noise ratio; superimposing the environmental noise sample onto the original voice data based on the signal-to-noise ratio to obtain superimposed original voice data; The superimposed original voice data is used as the first voice data.

3. The method for processing speech tasks based on artificial intelligence according to claim 1, characterized in that: The step of performing a process of adding and labeling spoken features to the original voice data to obtain corresponding second voice data specifically includes: Call the preset annotation tool; Annotating the spoken features in the original speech data based on the annotation tool to obtain corresponding annotation information; Adding the annotation information to the original voice data accordingly to obtain the added original voice data; The added original voice data is used as the second voice data.

4. The method for processing speech tasks based on artificial intelligence according to claim 1, characterized in that: The step of preprocessing the multi-dimensional voice data to obtain corresponding original voice data specifically includes: Performing denoising processing on the multi-dimensional voice data to obtain corresponding first multi-dimensional voice data; Segmenting the first multi-dimensional voice data to obtain corresponding second multi-dimensional voice data; The original voice data is generated based on the second multi-dimensional voice data.

5. The method for processing speech tasks based on artificial intelligence according to claim 1, characterized in that: The step of obtaining pre-constructed multi-dimensional voice data specifically includes: Obtain pre-collected real voice data; Get pre-built synthetic speech data; Integrate the real voice data and the synthesized voice data to obtain corresponding integrated voice data; The integrated voice data is used as the multi-dimensional voice data.

6. The method for processing speech tasks based on artificial intelligence according to claim 5, characterized in that: The step of obtaining pre-constructed synthesized speech data specifically includes: Obtaining a preset initial text containing specified speaker characteristics; wherein the specified speaker characteristics at least include age, gender and accent; Setting a corresponding designated speech parameter based on the designated speaker feature and the initial text; Based on the specified voice parameters, converting the initial text into corresponding voice data using a preset conversion strategy; The synthesized speech data is generated based on the speech data.

7. The method for processing speech tasks based on artificial intelligence according to claim 1, characterized in that: After the step of training and optimizing the deep learning model based on the preset construction strategy using the original voice data, the first voice data, and the second voice data to obtain a target voice recognition model that meets the preset performance requirements, the method further includes: Determining a designated storage mode corresponding to the target speech recognition model; Performing storage processing on the target speech recognition model based on the specified storage method; After completing the storage process of the target speech recognition model, the target speech recognition model is deployed.

8. A speech task processing device based on artificial intelligence, characterized in that: include: An acquisition module, used to acquire pre-built multi-dimensional voice data; A preprocessing module, used to preprocess the multi-dimensional voice data to obtain corresponding original voice data; A superposition module, used for performing noise superposition processing on the original voice data to obtain corresponding first voice data; An adding module, used for performing a marking and adding process of colloquial features on the original voice data to obtain corresponding second voice data; The calling module is used to call the pre-trained deep learning model; A first processing module, configured to train and optimize the deep learning model using the original voice data, the first voice data, and the second voice data based on a preset construction strategy to obtain a target voice recognition model that meets preset performance requirements; The second processing module is used to process the speech interaction task to be processed based on the target speech recognition model.

9. A computer device, characterized in that: It comprises a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the speech task processing method based on artificial intelligence as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the steps of the artificial intelligence-based speech task processing method as described in any one of claims 1 to 7 are implemented.

Images