Method and apparatus for classifying speech information

Abstract

The application provides a voice information classification method and device, relates to the technical field of artificial intelligence, and can be applied to the technical field of finance or other technical fields.The voice information classification method comprises the following steps: obtaining voice information to be recognized; extracting voice feature information from the voice information; inputting the voice feature information into a voice classification model created based on business voice data and general voice data to obtain a voice classification result.The application can increase data credibility and realize accurate recognition in a specific business scenario.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and more specifically, to a method and apparatus for classifying voice information. Background Technology

[0002] Deep learning, with its powerful learning capabilities, is widely used in tasks such as speech recognition, image recognition, and advertising recommendation, significantly improving work efficiency and saving costs. Among these, speech intent recognition technology enables various intelligent devices to understand user intent to the greatest extent, improving the level of robot personalization and thus better serving humans. Currently, speech intent recognition is widely used in finance, commerce, and other fields, especially in telephone customer service, where it can identify negative intent and provide opinions and suggestions for correcting it, thereby improving service quality and increasing work efficiency.

[0003] For deep learning models to achieve good results, they require a large amount of data for training. This leads to insufficient generalization ability, resulting in unsatisfactory results when data is scarce. Currently, training data for deep learning-based speech intent recognition models mostly comes from everyday conversations. This type of data is easy to obtain, and a large amount of data can be collected to form datasets. Using everyday language as a dataset provides a large amount of data and can achieve good training results. However, it is difficult to achieve good results for specific business scenarios, such as bank debt collection and outbound marketing calls. On the other hand, data for specific business scenarios is more difficult to obtain, with smaller amounts of data and poor training results. A method is needed to organically combine the two types of data to achieve better training results. Summary of the Invention

[0004] The main objective of this invention is to provide a voice information classification method and apparatus to increase data reliability and achieve accurate recognition in specific business scenarios.

[0005] To achieve the above objectives, embodiments of the present invention provide a voice information classification method, including:

[0006] Obtain the speech information to be recognized;

[0007] Extracting speech feature information from speech information;

[0008] The speech feature information is input into a speech classification model created based on business speech data and general speech data to obtain the speech classification result.

[0009] In one embodiment, the step of creating a speech classification model based on business speech data and general speech data includes:

[0010] Perform the following iterative processing:

[0011] The business voice data and the corresponding general voice data are input into the initial model to obtain the predicted data type;

[0012] The loss function is determined based on the predicted data type, the corresponding true data type, and the corresponding loss function coefficients.

[0013] When the loss function reaches the preset range, the initial model is determined to be a speech classification model; otherwise, the initial model is adjusted according to the loss function.

[0014] In one embodiment, it further includes:

[0015] Obtain the corresponding general voice data based on the business voice data;

[0016] The loss function coefficients are determined based on business voice data and general voice data.

[0017] In one embodiment, obtaining the corresponding general voice data based on the service voice data includes:

[0018] Obtain intent tags based on the business tags in the business voice data;

[0019] Obtain the corresponding general speech data based on the intent label.

[0020] In one embodiment, determining the loss function coefficients based on the service voice data and general voice data includes:

[0021] Determine the service coefficient based on the service voice data;

[0022] The general coefficient is determined based on the intent label and corresponding business label of the general voice data;

[0023] The loss function coefficients are determined based on the business coefficients and general coefficients.

[0024] In one embodiment, determining the service coefficient based on the service voice data includes:

[0025] Input the business voice data into the trained business voice classification model to obtain the business voice prediction result;

[0026] The business coefficient is determined based on the business voice prediction results and the corresponding business labels.

[0027] This invention also provides a voice information classification device, comprising:

[0028] The voice information acquisition module is used to acquire the voice information to be recognized;

[0029] The speech feature information extraction module is used to extract speech feature information from speech information;

[0030] The speech classification module is used to input speech feature information into a speech classification model created based on business speech data and general speech data to obtain speech classification results.

[0031] In one embodiment, it further includes:

[0032] The prediction module is used to input business voice data and corresponding general voice data into the initial model to obtain the predicted data type;

[0033] The loss function module is used to determine the loss function based on the predicted data type, the corresponding true data type, and the corresponding loss function coefficients.

[0034] The speech classification model module is used to determine the initial model as a speech classification model when the loss function reaches a preset range; otherwise, the initial model is adjusted according to the loss function.

[0035] In one embodiment, it further includes:

[0036] The general voice data module is used to obtain the corresponding general voice data based on the business voice data;

[0037] The loss function coefficient module is used to determine the loss function coefficients based on business voice data and general voice data.

[0038] In one embodiment, the general voice data module includes:

[0039] The intent tagging unit is used to obtain intent tags based on the service tags of the service voice data.

[0040] A general speech data unit is used to obtain the corresponding general speech data based on the intent label.

[0041] In one embodiment, the loss function coefficient module includes:

[0042] The service coefficient unit is used to determine the service coefficient based on the service voice data.

[0043] The general coefficient unit is used to determine the general coefficients based on the intent tags and corresponding business tags of the general voice data.

[0044] The loss function coefficient unit is used to determine the loss function coefficients based on the business coefficients and general coefficients.

[0045] In one embodiment, the business coefficient unit includes:

[0046] The business speech prediction result subunit is used to input business speech data into the trained business speech classification model to obtain business speech prediction results.

[0047] The business coefficient subunit is used to determine the business coefficient based on the business voice prediction results and the corresponding business labels.

[0048] This invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the aforementioned voice information classification method.

[0049] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the aforementioned voice information classification method.

[0050] This invention also provides a computer program product, including a computer program / instructions, wherein the computer program / instructions, when executed by a processor, implement the steps of the aforementioned voice information classification method.

[0051] The speech information classification method and apparatus of this invention first extracts speech feature information from speech information, and then inputs the speech feature information into a speech classification model created based on business speech data and general speech data to obtain speech classification results. This can increase data credibility and achieve accurate recognition in specific business scenarios. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a flowchart of the voice information classification method in an embodiment of the present invention;

[0054] Figure 2 This is a flowchart of creating a speech classification model in an embodiment of the present invention;

[0055] Figure 3 This is a flowchart illustrating the determination of loss function coefficients in an embodiment of the present invention;

[0056] Figure 4 This is a flowchart of S301 in an embodiment of the present invention;

[0057] Figure 5 This is a flowchart of S302 in an embodiment of the present invention;

[0058] Figure 6 This is a structural block diagram of the voice information classification device in an embodiment of the present invention;

[0059] Figure 7This is a structural block diagram of a voice information classification device according to another embodiment of the present invention;

[0060] Figure 8 This is a schematic diagram of the speech segmentation module in an embodiment of the present invention;

[0061] Figure 9 This is a schematic diagram of the dataset module in an embodiment of the present invention;

[0062] Figure 10 This is a schematic diagram of the data processing module in an embodiment of the present invention;

[0063] Figure 11 This is a schematic diagram of the credibility analysis module in an embodiment of the present invention;

[0064] Figure 12 This is a schematic diagram of the deep learning model module in an embodiment of the present invention;

[0065] Figure 13 This is a schematic block diagram illustrating the system configuration of an electronic device 9600 according to an embodiment of this application. Detailed Implementation

[0066] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] Those skilled in the art will recognize that embodiments of the present invention can be implemented as a system, apparatus, device, method, or computer program product. Therefore, this disclosure can be specifically implemented in the following forms: entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0068] The acquisition, storage, use, and processing of data in this application comply with relevant laws and regulations.

[0069] Given the poor generalization ability of deep learning-based speech intent recognition models, this invention provides a speech information classification method and apparatus. During training, different credibility levels are assigned to a general dataset and a specific business dataset. After credibility analysis, the data is used as training samples to achieve better speech intent recognition performance in specific business scenarios. The purpose of this invention is to improve the generalization ability of deep learning-based speech intent recognition models by establishing a speech intent recognition model that can adapt to specific business scenarios through credibility analysis of training data. The invention will be described in detail below with reference to the accompanying drawings.

[0070] Figure 1 This is a flowchart of the voice information classification method in an embodiment of the present invention. Figure 1 As shown, speech information classification methods include:

[0071] S101: Obtain the speech information to be recognized.

[0072] In practice, the speech to be recognized can be obtained through a microphone.

[0073] S102: Extract speech feature information from speech information.

[0074] In one embodiment, speech features in the speech information to be recognized can be determined, silent parts in the speech can be removed, and the speech can be cut into specified lengths.

[0075] In practice, Voice Activity Detection (VAD) is performed on the input speech, and features are extracted from the speech data to obtain a speech feature vector sequence. The speech feature vector sequence may include at least one speech object and a feature vector. The speech feature vector can be used to represent at least one of the following: sound intensity features, loudness features, pitch periodicity features, fundamental frequency periodicity features, fundamental frequency features, signal-to-noise ratio features, harmonic noise ratio features, and Mel-Frequency Cepstrum Coefficients (MFCC). After obtaining the speech feature vector, the speech is segmented, removing silent portions, and then divided into frames of a specific length. Silence is determined based on different feature vector methods, commonly using energy-based, frequency domain-based, cepstrum-based, harmonic-based, and long-term information-based methods. The frame length can be set manually, typically 20ms.

[0076] S103: Input the speech feature information into the speech classification model created based on business speech data and general speech data to obtain the speech classification result.

[0077] Figure 2 This is a flowchart illustrating the creation of a speech classification model in an embodiment of the present invention. For example... Figure 2 As shown, the steps for creating a speech classification model based on business speech data and general speech data include:

[0078] Perform the following iterative processing:

[0079] S201: Input the business voice data and the corresponding general voice data into the initial model to obtain the predicted data type.

[0080] General-purpose voice data consists of everyday conversation datasets; this type of data is readily available and abundant. Business-specific voice data is used in designated scenarios. There can be one or more specific business sets (business voice datasets), which, together with the general-purpose dataset, form a dataset unit.

[0081] S202: Determine the loss function based on the predicted data type, the corresponding true data type, and the corresponding loss function coefficients.

[0082] In practice, given speech data X, the model prediction value corresponding to the predicted data type is... If the true value corresponding to the real data type is Y, then the expression for the squared difference loss function is:

[0083] .

[0084] For classification problems, the loss function can be understood as the sum of the loss functions for each class when calculating the total loss function. Therefore, when adding them, multiplying by the loss function coefficient of the corresponding class first allows the results of the confidence analysis to be incorporated into the model. That is, the above expression can be written as:

[0085]

[0086] in, For loss function, Let be the loss function coefficients for the nth class of intent labels. This represents the number of data items corresponding to the nth type of intent label.

[0087] Besides the squared loss function, common loss functions include log loss, hinge loss, exponential loss, cross-entropy loss, and absolute loss, all of which can be processed using the methods described above.

[0088] Figure 3 This is a flowchart illustrating the determination of loss function coefficients in an embodiment of the present invention. For example... Figure 3 As shown, speech information classification methods also include:

[0089] S301: Obtain the corresponding general voice data based on the business voice data.

[0090] Figure 4 This is a flowchart of S301 in an embodiment of the present invention. For example... Figure 4 As shown, S301 includes:

[0091] S401: Obtain intent tags based on the service tags of the service voice data.

[0092] Intent tags (sub-tags) are labels for general voice data and business voice data, including emotions such as happiness, sadness, and anger. Business tags (primary tags) are labels for business voice data.

[0093] A business tag corresponds to multiple intent tags. Therefore, we can filter out intent tags of the same type as a business tag. For example, the intent tags of the telephone banking complaint tag include negative and angry tags. Each business tag is filtered and categorized in this way, forming a tag set together with the intent tags. One business tag can correspond to multiple intent tags, but one intent tag can only correspond to one business tag. Therefore, the number of categories in the tag set is equal to the number of intent tags.

[0094] S402: Obtain the corresponding general speech data based on the intent label.

[0095] S302: Determine the loss function coefficients based on the business voice data and general voice data.

[0096] Figure 5 This is a flowchart of S302 in an embodiment of the present invention. For example... Figure 5 As shown, S302 includes:

[0097] S501: Determine the service coefficient based on the service voice data.

[0098] In one embodiment, S501 includes: inputting business voice data into a trained business voice classification model to obtain business voice prediction results; and determining business coefficients based on the business voice prediction results and corresponding business labels.

[0099] In practical implementation, prediction accuracy can measure whether the data in a general dataset is accurate and effective, and therefore can be used as a business coefficient. A speech classification model can be built using time series LSTM. After the model is trained, the prediction accuracy for each category is obtained. This accuracy is a decimal between 0 and 1, represented by the letter p.

[0100] S502: Determine the general coefficient based on the intent label and corresponding service label of the general voice data.

[0101] In practice, the general coefficient is a decimal between 0 and 1, where 0 represents complete dissimilarity and 1 represents complete similarity, denoted by the letter 'o'. The higher the similarity, the higher the general coefficient. For example, if the business tag is "complaint" and the intent tags are "sadness" and "anger," anger is more relevant to complaints, so anger is assigned a coefficient of 0.8 and sadness is assigned a coefficient of 0.1.

[0102] S503: Determine the loss function coefficients based on the business coefficients and general coefficients.

[0103] In practice, the loss function coefficients can be determined using the following formula:

[0104] a = o × p;

[0105] Where a is the loss function coefficient, o is the general coefficient, and p is the business coefficient.

[0106] S203: Determine whether the loss function has reached the preset range.

[0107] S204: When the loss function reaches the preset range, the initial model is determined to be a speech classification model.

[0108] The goal of deep learning model training is to minimize the loss function. When the loss function stops decreasing, the training has converged and is complete. When used, data is input into the trained model, and the model provides prediction results.

[0109] S205: When the loss function does not reach the preset range, adjust the initial model according to the loss function.

[0110] In practical applications, dialogues that are not in the dataset may also appear. These dialogues can be used to identify intent, and dialogues that are incorrectly identified can be recorded, assigned the correct labels, and added to the dataset to improve the accuracy of the model.

[0111] Figure 1 The speech information classification method shown can be implemented by a computer. Figure 1 As shown in the process, the speech information classification method of this invention first extracts speech feature information from speech information, and then inputs the speech feature information into a speech classification model created based on business speech data and general speech data to obtain speech classification results. This can increase data credibility and achieve accurate recognition in specific business scenarios.

[0112] The specific process of this invention embodiment is as follows:

[0113] 1. Obtain intent tags based on the business tags of the business voice data, and obtain the corresponding general voice data based on the intent tags.

[0114] 2. Input the business voice data into the trained business voice classification model to obtain the business voice prediction results.

[0115] 3. Determine the business coefficient based on the business voice prediction results and the corresponding business tags.

[0116] 4. Determine the general coefficient based on the intent tags and corresponding business tags of the general voice data.

[0117] 5. Determine the loss function coefficients based on the business coefficients and general coefficients.

[0118] 6. Input the business voice data and the corresponding general voice data into the initial model to obtain the predicted data type.

[0119] 7. Determine the loss function based on the predicted data type, the corresponding true data type, and the corresponding loss function coefficients.

[0120] 8. When the loss function reaches the preset range, determine the initial model as a speech classification model; otherwise, adjust the initial model according to the loss function and return to step 6.

[0121] 9. Obtain the speech information to be recognized and extract speech feature information from the speech information.

[0122] 10. Input the speech feature information into the speech classification model to obtain the speech classification result.

[0123] In summary, the speech information classification method of the present invention assigns different levels of credibility to the general dataset and the specific business dataset during training. After data credibility analysis, the data is used as training samples for training. This method can classify speech in specific business scenarios more accurately. Without adding a large number of additional data samples, it improves the generalization ability of the model and makes the model training more accurate and efficient.

[0124] Based on the same inventive concept, this embodiment of the invention also provides a voice information classification device. Since the principle of this device in solving the problem is similar to that of the voice information classification method, the implementation of this device can refer to the implementation of the method, and the repeated parts will not be described again.

[0125] Figure 6 This is a structural block diagram of the voice information classification device in an embodiment of the present invention. Figure 6 As shown, the voice information classification device includes:

[0126] The voice information acquisition module is used to acquire the voice information to be recognized;

[0127] The speech feature information extraction module is used to extract speech feature information from speech information;

[0128] The speech classification module is used to input speech feature information into a speech classification model created based on business speech data and general speech data to obtain speech classification results.

[0129] In one embodiment, it further includes:

[0130] The prediction module is used to input business voice data and corresponding general voice data into the initial model to obtain the predicted data type;

[0131] The loss function module is used to determine the loss function based on the predicted data type, the corresponding true data type, and the corresponding loss function coefficients.

[0132] The speech classification model module is used to determine the initial model as a speech classification model when the loss function reaches a preset range; otherwise, the initial model is adjusted according to the loss function.

[0133] In one embodiment, it further includes:

[0134] The general voice data module is used to obtain the corresponding general voice data based on the business voice data;

[0135] The loss function coefficient module is used to determine the loss function coefficients based on business voice data and general voice data.

[0136] In one embodiment, the general voice data module includes:

[0137] The intent tagging unit is used to obtain intent tags based on the service tags of the service voice data.

[0138] A general speech data unit is used to obtain the corresponding general speech data based on the intent label.

[0139] In one embodiment, the loss function coefficient module includes:

[0140] The service coefficient unit is used to determine the service coefficient based on the service voice data.

[0141] The general coefficient unit is used to determine the general coefficients based on the intent tags and corresponding business tags of the general voice data.

[0142] The loss function coefficient unit is used to determine the loss function coefficients based on the business coefficients and general coefficients.

[0143] In one embodiment, the business coefficient unit includes:

[0144] The business speech prediction result subunit is used to input business speech data into the trained business speech classification model to obtain business speech prediction results.

[0145] The business coefficient subunit is used to determine the business coefficient based on the business voice prediction results and the corresponding business labels.

[0146] Figure 7 This is a structural block diagram of a voice information classification device according to another embodiment of the present invention. Figure 7As shown, in practical applications, the speech information classification device includes a speech data acquisition module, a speech segmentation module, a dataset module, a data processing module, a credibility analysis module, a deep learning model module, and a result analysis module. Among these, the dataset module, data processing module, credibility analysis module, and deep learning model module are required for modeling and training the speech intent recognition model. The overall process can be understood as follows: inputting speech data, segmenting the speech data, and then inputting it into the established speech intent recognition model; the model provides the intent recognition result; and after analyzing the result, the model is further optimized.

[0147] The data acquisition module includes a voice information acquisition module, which is used to acquire the voice to be recognized through a microphone and input it as input data into subsequent modules.

[0148] Figure 8 This is a schematic diagram of the speech segmentation module in an embodiment of the present invention. For example... Figure 8 As shown, the speech segmentation module includes a speech feature information extraction module. This module performs Voice Activity Detection (VAD) on the input speech data, which is obtained by the data acquisition module. The speech feature extraction module extracts features from the speech data to obtain a speech feature vector sequence. The speech feature vector sequence may include at least one speech element and a feature vector. The speech feature vector can be used to represent at least one of the following: sound intensity feature, loudness feature, pitch periodicity feature, fundamental frequency periodicity feature, fundamental frequency feature, signal-to-noise ratio feature, harmonic noise ratio feature, and Mel-Frequency Cepstrum Coefficients (MFCC). The speech segmentation module segments the speech after obtaining the speech feature vectors, removing silent portions and dividing the speech into frames of a specific length. Silence is determined based on different feature vector methods, commonly energy-based, frequency domain-based, cepstrum-based, harmonic-based, and long-term information-based methods. The frame length can be set manually, typically 20ms.

[0149] Figure 9 This is a schematic diagram of the dataset module in an embodiment of the present invention. Figure 9 As shown, the dataset module contains a general dataset and specific business datasets. The general dataset consists of everyday dialogue datasets, which are readily available and voluminous. Specific business datasets are used in specific scenarios. There can be one or more specific business datasets, which, together with the general dataset, form a dataset unit.

[0150] Figure 10 This is a schematic diagram of the data processing module in an embodiment of the present invention. For example... Figure 10As shown, prior to credibility analysis, the existing data needs to be processed. In the intent recognition dataset, each label, such as happy, sad, and angry, corresponds to a data category. The data processing module is used to classify the labels and their subordinate data. The primary label is the name of a label in a specific business dataset; the secondary label is a label in a general dataset. Secondary labels of the same category as the primary label are filtered out. For example, for the label "telephone banking complaint," similar secondary labels include "negative" and "angry." Each primary label is filtered and classified in this way, forming a label set together with the secondary labels. It is important to note that within a specific business dataset, one primary label can correspond to multiple secondary labels, but one secondary label can only correspond to one primary label. Therefore, the number of categories in the label set is equal to the number of secondary labels.

[0151] The credibility analysis module includes a general speech data module and a loss function coefficient module. Figure 11 This is a schematic diagram of the credibility analysis module in an embodiment of the present invention. For example... Figure 11 As shown, it comprises three parts: a subjective analysis unit, an objective analysis unit, and a data fusion unit. The objective analysis unit primarily analyzes whether the data in the general dataset is accurate and effective, using prediction accuracy as a metric. A speech classification model is built using a time-series LSTM. After the model is trained, the prediction accuracy for each category is obtained; this accuracy is a decimal between 0 and 1.

[0152] The subjective analysis unit needs to provide corresponding coefficients, which are decimals between 0 and 1, where 0 represents complete dissimilarity and 1 represents complete similarity. The higher the similarity, the higher the coefficient. For example, if the main label is "complaint" and the sublabels are "sadness" and "anger," then "anger" is more relevant to "complaint," so a coefficient of 0.8 can be assigned to "anger," and a coefficient of 0.1 can be assigned to "sadness."

[0153] The data fusion unit is used to combine the results obtained from objective and subjective analysis. For each sublabel, a coefficient is obtained using this method.

[0154] Figure 12 This is a schematic diagram of the deep learning model module in an embodiment of the present invention. Figure 12As shown, the deep learning model module includes a speech classification module, a prediction module, a loss function module, and a speech classification model module. Deep model learning can be divided into an input layer (61), a hidden layer (62), and an output layer (63). The input layer contains the input data; one circle represents one data point, and k circles represent k-dimensional input data. The hidden layer contains the specific algorithm of the model; one circle represents one neuron. Deep learning includes multiple hidden layers, which may employ operations such as convolution, pooling, and non-linear transformations depending on the algorithm. Each neuron has its own weight w. The goal of neural network training is to find the optimal w parameters to maximize the network's prediction accuracy. The output layer contains the output data; one circle represents one data point, and l circles represent l-dimensional output data. Speech intent recognition model algorithms can include Support Vector Machine (SVM), Gaussian Mixture Model (GMM), Hidden Markov Model (HMM), Bag-of-Words Model (BOW), Convolutional Neural Network (CNN), and Decision Tree models, among others.

[0155] In practical applications, dialogues that are not present in the dataset may occur. The results analysis module performs intent recognition on these dialogues, records those that are incorrectly identified, assigns them the correct labels, and adds them to the dataset, thereby improving the model's accuracy.

[0156] In summary, the speech information classification device of this invention first extracts speech feature information from speech information, and then inputs the speech feature information into a speech classification model created based on business speech data and general speech data to obtain speech classification results. This can increase data credibility and achieve accurate recognition in specific business scenarios.

[0157] Figure 13 This is a schematic block diagram illustrating the system configuration of the electronic device 9600 according to an embodiment of this application. Figure 13 As shown, the electronic device 9600 may include a central processing unit 9100 and a memory 9140; the memory 9140 is coupled to the central processing unit 9100. It is worth noting that... Figure 13 This is an example; other types of structures can also be used to supplement or replace this structure to achieve telecommunications functions or other functions.

[0158] In one embodiment, the voice information classification method function can be integrated into the central processing unit 9100. The central processing unit 9100 can be configured to perform the following control:

[0159] Obtain the speech information to be recognized;

[0160] Extracting speech feature information from speech information;

[0161] The speech feature information is input into a speech classification model created based on business speech data and general speech data to obtain the speech classification result.

[0162] As can be seen from the above description, the speech information classification method provided in this application first extracts speech feature information from speech information, and then inputs the speech feature information into a speech classification model created based on business speech data and general speech data to obtain speech classification results. This can increase data credibility and achieve accurate recognition in specific business scenarios.

[0163] In another embodiment, the voice information classification device can be configured separately from the central processing unit 9100. For example, the voice information classification device can be configured as a chip connected to the central processing unit 9100, and the voice information classification method can be implemented through the control of the central processing unit.

[0164] like Figure 13 As shown, the electronic device 9600 may further include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worth noting that the electronic device 9600 does not necessarily need to include these components. Figure 13 All components shown; in addition, the electronic device 9600 may also include Figure 13 For components not shown, please refer to existing technologies.

[0165] like Figure 13 As shown, the central processing unit 9100, sometimes also referred to as a controller or operating control, may include a microprocessor or other processor device and / or logic device, which receives inputs and controls the operation of various components of the electronic device 9600.

[0166] The memory 9140 may be, for example, one or more of a cache, flash memory, hard drive, removable media, volatile memory, non-volatile memory, or other suitable devices. It may store the aforementioned failure-related information, and also store a program for executing that information. The central processing unit 9100 may execute the program stored in the memory 9140 to perform information storage or processing, etc.

[0167] Input unit 9120 provides input to central processing unit 9100. Input unit 9120 may be, for example, a keypad or touch input device. Power supply 9170 provides power to electronic device 9600. Display 9160 displays images and text. Display may be, for example, an LCD display, but is not limited thereto.

[0168] The memory 9140 can be a solid-state memory, such as a read-only memory (ROM), random access memory (RAM), a SIM card, etc. It can also be a memory that retains information even when power is off, can be selectively erased, and contains more data; examples of this type of memory are sometimes referred to as EPROMs, etc. The memory 9140 can also be some other type of device. The memory 9140 includes a buffer 9141 (sometimes referred to as a buffer memory). The memory 9140 may include an application / function storage unit 9142 for storing application programs and function programs or processes for executing the operation of the electronic device 9600 via the central processing unit 9100.

[0169] The memory 9140 may also include a data storage unit 9143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 may include various drivers for the electronic device for communication functions and / or for performing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0170] The communication module 9110 is a transmitter / receiver 9110 that transmits and receives signals via the antenna 9111. The communication module (transmitter / receiver) 9110 is coupled to the central processing unit 9100 to provide input signals and receive output signals, which can be the same as in a conventional mobile communication terminal.

[0171] Based on different communication technologies, multiple communication modules 9110 can be configured in the same electronic device, such as cellular network modules, Bluetooth modules, and / or wireless LAN modules. The communication module (transmitter / receiver) 9110 is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide audio output via the speaker 9131 and receive audio input from the microphone 9132, thereby realizing typical telecommunications functions. The audio processor 9130 may include any suitable buffer, decoder, amplifier, etc. Additionally, the audio processor 9130 is also coupled to a central processing unit 9100, enabling on-device recording via the microphone 9132 and on-device playback of stored audio via the speaker 9131.

[0172] This invention also provides a computer-readable storage medium capable of implementing all steps of the voice information classification method in the above embodiments, where the execution subject is a server or a client. The computer-readable storage medium stores a computer program that, when executed by a processor, implements all steps of the voice information classification method in the above embodiments. For example, when the processor executes the computer program, it implements the following steps:

[0173] Obtain the speech information to be recognized;

[0174] Extracting speech feature information from speech information;

[0175] The speech feature information is input into a speech classification model created based on business speech data and general speech data to obtain the speech classification result.

[0176] In summary, the computer-readable storage medium of this invention first extracts speech feature information from speech information, and then inputs the speech feature information into a speech classification model created based on business speech data and general speech data to obtain speech classification results. This can increase data credibility and achieve accurate recognition in specific business scenarios.

[0177] This invention also provides a computer program product capable of implementing all steps of the voice information classification method in the above embodiments, where the execution subject is a server or a client. The computer program product includes a computer program / instruction, which, when executed by a processor, implements all steps of the voice information classification method in the above embodiments. For example, when the processor executes the computer program, it implements the following steps:

[0178] Obtain the speech information to be recognized;

[0179] Extracting speech feature information from speech information;

[0180] The speech feature information is input into a speech classification model created based on business speech data and general speech data to obtain the speech classification result.

[0181] In summary, the computer program product of this invention first extracts speech feature information from speech information, and then inputs the speech feature information into a speech classification model created based on business speech data and general speech data to obtain speech classification results. This can increase data credibility and achieve accurate recognition in specific business scenarios.

[0182] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. In particular, hardware + program embodiments are relatively simple in description because they are fundamentally similar to method embodiments; relevant parts can be referred to the descriptions in the method embodiments.

[0183] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0184] While this application provides the method operation steps as described in the embodiments or flowcharts, more or fewer operation steps may be included based on conventional or non-inventive labor. The order of steps listed in the embodiments is merely one possible execution order among many and does not represent the only execution order. In actual device or client product execution, the methods shown in the embodiments or drawings can be executed sequentially or in parallel (e.g., in a parallel processor or multi-threaded processing environment).

[0185] While this specification provides method operation steps as described in the embodiments or flowcharts, more or fewer operation steps may be included based on conventional or non-inventive means. The order of steps listed in the embodiments is merely one possible execution order among many and does not represent the only execution order. In actual device or end product execution, the methods shown in the embodiments or drawings may be executed sequentially or in parallel (e.g., in a parallel processor or multi-threaded processing environment, or even a distributed data processing environment). The terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, product, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, product, or apparatus. Without further limitations, the presence of other identical or equivalent elements in the process, method, product, or apparatus that includes said elements is not excluded.

[0186] For ease of description, the above devices are described in terms of function, divided into various modules. Of course, in implementing the embodiments of this specification, the functions of each module can be implemented in one or more software and / or hardware components, or a module that performs the same function can be implemented by a combination of multiple sub-modules or sub-units. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms.

[0187] Those skilled in the art will also know that, besides implementing the controller using purely computer-readable program code, the same functions can be achieved by logically programming the method steps, making the controller function as logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers (PLCs), and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the devices within it used to implement various functions can also be considered structures within that hardware component. Alternatively, the devices used to implement various functions can be considered as both software modules implementing the method and structures within a hardware component.

[0188] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0189] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0190] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0191] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0192] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0193] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0194] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, the embodiments of this specification can take the form of computer program products implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0195] The embodiments described in this specification can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. The embodiments of this specification can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0196] The embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on its differences from other embodiments. In particular, system embodiments are basically similar to method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. In the description of this specification, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the embodiments in this specification. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described can be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0197] The above description is merely an embodiment of the present specification and is not intended to limit the embodiments of the present specification. For those skilled in the art, various modifications and variations can be made to the embodiments of the present specification. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the embodiments of the present specification should be included within the scope of the claims of the embodiments of the present specification.

Claims

1. A method for classifying voice information, characterized in that, include: Obtain the speech information to be recognized; Extract speech feature information from the speech information; The speech feature information is input into a speech classification model created based on business speech data and general speech data to obtain the speech classification result; The steps to create a speech classification model based on business speech data and general speech data include: Perform the following iterative processing: The business voice data and the corresponding general voice data are input into the initial model to obtain the predicted data type; The loss function is determined based on the predicted data type, the corresponding real data type, and the corresponding loss function coefficients. When the loss function reaches a preset range, the initial model is determined to be a speech classification model; otherwise, the initial model is adjusted according to the loss function. Also includes: Obtain the intent tag based on the service tag of the service voice data; The corresponding general voice data is obtained based on the intent label; the intent label is a label for the general voice data and the business voice data. Determine the service coefficients based on the aforementioned service voice data; The general coefficient is determined based on the intent label and corresponding service label of the general voice data; The loss function coefficients are determined based on the business coefficients and the general coefficients; The loss function coefficients are determined according to the following formula: a = o × p; Where a is the loss function coefficient, o is the general coefficient, and p is the business coefficient.

2. The voice information classification method according to claim 1, characterized in that, Determining the service coefficients based on the aforementioned service voice data includes: The business voice data is input into the trained business voice classification model to obtain the business voice prediction result; The service coefficient is determined based on the service voice prediction result and the corresponding service label.

3. A voice information classification device, characterized in that, include: The voice information acquisition module is used to acquire the voice information to be recognized; A speech feature information extraction module is used to extract speech feature information from the speech information; The speech classification module is used to input the speech feature information into a speech classification model created based on business speech data and general speech data to obtain the speech classification result. The prediction module is used to input the business voice data and the corresponding general voice data into the initial model to obtain the predicted data type; The loss function module is used to determine the loss function based on the predicted data type, the corresponding true data type, and the corresponding loss function coefficients. The speech classification model module is used to determine that the initial model is a speech classification model when the loss function reaches a preset range; otherwise, the initial model is adjusted according to the loss function. The general speech data module includes an intent labeling unit and a general speech data unit; The intent tagging unit is used to obtain intent tags based on the service tags of the service voice data; The general voice data unit is used to obtain corresponding general voice data according to the intent label; the intent label is a label for the general voice data and the business voice data. The loss function coefficient module is used to determine the loss function coefficients based on the service voice data and the general voice data; The loss function coefficient module includes: A service coefficient unit is used to determine service coefficients based on the service voice data. A general coefficient unit is used to determine a general coefficient based on the intent label and the corresponding service label of the general voice data; A loss function coefficient unit is used to determine the loss function coefficients based on the business coefficients and the general coefficients; The loss function coefficients are determined according to the following formula: a = o × p; Where a is the loss function coefficient, o is the general coefficient, and p is the business coefficient.

4. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the speech information classification method according to any one of claims 1 to 2.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the speech information classification method according to any one of claims 1 to 2.

6. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the speech information classification method according to any one of claims 1 to 2.

Images