Speech recognition processing method and system for electric power safety system

Abstract

The invention provides a voice recognition processing method and system for an electric power safety system, and the method comprises the steps: obtaining original voice data of a consulting object, and generating a corresponding original voice text according to the original voice data; generating a power business information text corresponding to the original voice text according to a corresponding power business generation rule; wherein the electric power business generation rule is a target electric power business generation rule which is determined according to a corresponding electric power business identifier obtained through the original voice text; and converting the power business information text into a corresponding target power business prompt voice through a preset conversion mode, and broadcasting the target power business prompt voice to the consulting object. The voice recognition accuracy and the voice recognition efficiency can be improved, and the consultation object can be helped to timely and accurately acquire information, so that the intelligent effect of an electric power safety system is improved to a certain extent.

Description

Technical Field

[0001] The present invention relates to the technical field of power system automation, and in particular to a speech recognition processing method and system for a power safety system. Background Art

[0002] With the continuous development of computer technology and artificial intelligence, as well as the continuous improvement of people's living needs, intelligent voice recognition is required in many scenarios, such as voice message text conversion scenarios, intelligent voice traffic scenarios, virtual robot interaction scenarios, etc., and these scenarios can be invested in traditional technology industries and high-tech industries, such as power systems, voice recognition systems, etc. In addition, with the continuous progress in natural language understanding, voice recognition processing technology has become an increasingly widespread application direction. For example, combining voice recognition processing technology to make the power system intelligent can better provide more convenient and efficient services to staff and service objects.

[0003] At present, the traditional speech recognition processing technology makes the power safety system intelligent, generally by receiving the voice information input by the user, using the non-autoregressive speech recognition model to perform speech recognition on the voice information to obtain the speech recognition result, and performing the corresponding operation based on the speech recognition result. Although the non-autoregressive speech recognition model has been widely used due to its advantages such as fast speech recognition speed, the non-autoregressive speech recognition model only uses the local information of the speech information at the speech level, and there is a large recognition error, which leads to the disadvantage of low recognition accuracy and reduces the intelligent effect of the power system. Summary of the invention

[0004] The purpose of the present invention is to propose a speech recognition processing method and system for an electric power safety system, so as to solve the technical problem of how to reduce speech recognition errors and thus improve accuracy.

[0005] On the one hand, a speech recognition processing method for an electric power safety system is provided, comprising:

[0006] Obtaining original voice data of the consulting object, and generating corresponding original voice text according to the original voice data;

[0007] Generate an electric power business information text corresponding to the original voice text according to the corresponding electric power business generation rule; wherein the electric power business generation rule is a corresponding electric power business identifier obtained through the original voice text, and a target electric power business generation rule is determined according to the electric power business identifier;

[0008] The power business information text is converted into a corresponding target power business prompt voice through a preset conversion method, and the voice is broadcast to the consulting object.

[0009] Preferably, generating the corresponding original voice text according to the original voice data includes inputting the original voice data into a preset voice recognition model, and identifying the voice text unit corresponding to the original voice data and the text probability distribution corresponding to the voice text unit through the voice recognition model;

[0010] The words corresponding to each speech text unit are determined by the text probability distribution corresponding to the speech text unit and a preset vocabulary, and all the words are combined into the corresponding original speech text.

[0011] Preferably, the speech recognition model is obtained by the following steps: inputting sample speech data into a preset basic speech recognition model, and obtaining a sample speech vector corresponding to the sample speech data, a corresponding sample text unit, and a basic text probability distribution corresponding to each sample text unit through an acoustic network in the basic speech recognition model;

[0012] Calculate a corresponding text recognition loss value based on each of the sample text units, the basic text probability distribution corresponding to each of the sample text units, and the text unit label;

[0013] The parameters of the basic speech recognition model are updated based on the text recognition loss value until the accuracy of the basic speech recognition model reaches a preset threshold, and the basic speech recognition model at this time is output as a speech recognition model.

[0014] Preferably, the corresponding power business identification obtained through the original voice text includes inputting the original voice text into a preset business discrimination model to obtain a basic business score for each power business;

[0015] The basic business score of each power business is calculated according to the preset business weight coefficient to obtain the comprehensive business score corresponding to each power business;

[0016] The preset identifier of the electric power business corresponding to the maximum comprehensive business score is output as the electric power business identifier.

[0017] Preferably, the business discrimination model is obtained by the following steps: inputting the sample speech text into a preset basic business discrimination model to determine the business discrimination probability corresponding to each of the sample speech texts;

[0018] Calculate the business discrimination loss value according to the business discrimination probability corresponding to each of the sample speech texts and the preset business information label;

[0019] The parameters of the basic business discrimination model are updated based on the business discrimination loss value until the accuracy of the basic business discrimination model reaches a preset threshold, and the basic business discrimination model at this time is output as a business discrimination model.

[0020] Preferably, the generating of the electric power business information text corresponding to the original voice text includes obtaining the consulting object identifier corresponding to the original voice data, and determining the consulting object number information according to the resource object identifier;

[0021] According to the original voice text and the consulting object number information, the corresponding power business information text is generated according to the power business generation rule.

[0022] Preferably, the determining the consulting object number information according to the resource object identifier includes, if the number information associated with the resource object identifier exists in the information database, outputting the associated number information as the consulting object number information;

[0023] If there is no numbering information associated with the resource object identifier in the information database, temporary numbering information is set for the resource object identifier according to a preset temporary information template, and the temporary numbering information is output as consulting object numbering information.

[0024] On the other hand, a speech recognition processing system for an electric power safety system is also provided, which is used to implement the speech recognition processing method for an electric power safety system, comprising:

[0025] A data acquisition module is used to acquire the original voice data of the consulting object and generate the corresponding original voice text according to the original voice data;

[0026] A text generation module, used to generate an electric power business information text corresponding to the original voice text according to a corresponding electric power business generation rule; wherein the electric power business generation rule is a corresponding electric power business identifier obtained through the original voice text, and a target electric power business generation rule determined according to the electric power business identifier;

[0027] The information prompt module is used to convert the power business information text into the corresponding target power business prompt voice through a preset conversion method, and broadcast it to the consulting object.

[0028] Preferably, the data acquisition module is further used to input the original voice data into a preset voice recognition model, and identify the voice text unit corresponding to the original voice data and the text probability distribution corresponding to the voice text unit through the voice recognition model;

[0029] The words corresponding to each speech text unit are determined by the text probability distribution corresponding to the speech text unit and a preset vocabulary, and all the words are combined into the corresponding original speech text.

[0030] Preferably, the text generation module is further used to input the original speech text into a preset business discrimination model to obtain a basic business score for each power business;

[0031] The basic business score of each power business is calculated according to the preset business weight coefficient to obtain the comprehensive business score corresponding to each power business;

[0032] Outputting the preset identifier of the electric power business corresponding to the maximum comprehensive business score as the electric power business identifier;

[0033] Obtain the consultation object identifier corresponding to the original voice data, and determine the consultation object number information based on the resource object identifier; wherein, if the number information associated with the resource object identifier exists in the information database, the associated number information is output as the consultation object number information; if the number information associated with the resource object identifier does not exist in the information database, temporary number information is set for the resource object identifier according to a preset temporary information template, and the temporary number information is output as the consultation object number information;

[0034] According to the original voice text and the consulting object number information, the corresponding power business information text is generated according to the power business generation rule.

[0035] In summary, the implementation of the embodiments of the present invention has the following beneficial effects:

[0036] The speech recognition processing method and system for the electric power safety system provided by the present invention can not only improve the accuracy and efficiency of speech recognition, but also help the consulting object to obtain information in a timely and accurate manner, thereby improving the intelligent effect of the electric power safety system to a certain extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, without paying creative labor, other drawings obtained based on these drawings still belong to the scope of the present invention.

[0038] Figure 1 The present invention is a schematic diagram of the main process of a speech recognition processing method for an electric power safety system according to an embodiment of the present invention.

[0039] Figure 2 This is an exemplary system architecture diagram in an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings.

[0041] like Figure 1 FIG. 1 is a schematic diagram of an embodiment of a speech recognition processing method for an electric power safety system provided by the present invention. In this embodiment, the method comprises the following steps:

[0042] Step S1, obtaining the original voice data of the consulting object, and generating the corresponding original voice text according to the original voice data; it is understandable that when the server receives the original voice data sent by the consulting object through the designated power business service number, the original voice text corresponding to the original voice data is obtained. It should be noted that the present invention is based on Figure 2 The system architecture 100 is implemented by the system architecture shown in the figure, and the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for a communication link between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc. Users can use terminal devices 101, 102, 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications may be installed on the terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc. The terminal devices 101, 102, 103 may be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 players (Moving Picture Experts Group Audio Layer IV), laptop computers, desktop computers, etc. The server 105 may be a server that provides various services, such as a background server that provides support for the pages displayed on the terminal devices 101, 102, 103.

[0043] In this embodiment, the server acts as the receiving end. The server usually has a detection mechanism to continuously check whether new voice data has arrived (i.e., wait for and receive the original voice data from the consulting object). Once the data is received, the server will store it in a temporary storage medium (such as memory or disk) for subsequent processing. The consulting object may send these voice data through a specific power business service number (such as a phone number, a voice input function of an online chat window, or a voice button in a mobile application).

[0044] In one embodiment, the generation of the corresponding original voice text according to the original voice data includes: inputting the original voice data into a preset voice recognition model, identifying the voice text unit corresponding to the original voice data and the text probability distribution corresponding to the voice text unit through the voice recognition model; determining the words corresponding to each voice text unit through the text probability distribution corresponding to the voice text unit and a preset vocabulary, and combining all the words into the corresponding original voice text. It is understandable that when the server receives the original voice data sent by the consulting object through the designated power business service number, the original voice data is input into the voice recognition model, and the voice text unit and the text probability distribution corresponding to each voice text unit are output through the voice recognition model; based on the text probability distribution corresponding to each voice text unit and the vocabulary, the words corresponding to each voice text unit are determined to obtain the original voice text.

[0045] It should be noted that after receiving the original voice data, the original voice data can be converted into the corresponding original voice text. Specifically, before the original voice data is input into the model, some preprocessing may be required for the original voice data, such as denoising, normalizing the volume, segmenting into frames, etc., to improve the recognition accuracy of the model. Then, the preprocessed original voice data can be input into the speech recognition model. The speech recognition model will process the data and output a series of voice text units and their corresponding text probability distributions, and then the voice text units and their corresponding text probability distributions are mapped to the corresponding words. The server will combine these words in order to form the final original voice text. Through this process, the original voice data sent by the consulting object can be converted into a text form that is easy to understand and process, thereby providing a basis for subsequent text analysis, natural language understanding or business processing.

[0046] Among them, the speech recognition model is pre-trained and can recognize and convert speech data into text. These models may be based on deep learning techniques such as recurrent neural networks (RNN), long short-term memory networks (LSTM), or more advanced Transformer architectures. The speech text unit may be a word, phrase, syllable, or character, depending on how the model is trained and the application scenario. The text probability distribution is used to represent the probability that the speech text unit is correctly recognized. This text probability distribution is usually a probability vector, in which each element corresponds to a word or phrase in the vocabulary.

[0047] Specifically, the speech recognition model is obtained through the following steps: inputting sample speech data into a preset basic speech recognition model, obtaining the sample speech vector corresponding to the sample speech data, the corresponding sample text unit, and the basic text probability distribution corresponding to each sample text unit through the acoustic network in the basic speech recognition model; calculating the corresponding text recognition loss value based on each sample text unit, the basic text probability distribution corresponding to each sample text unit, and the text unit label; updating the parameters of the basic speech recognition model based on the text recognition loss value until the accuracy of the basic speech recognition model reaches a preset threshold, and outputting the basic speech recognition model at this time as a speech recognition model. It should be noted that calculating the corresponding text recognition loss value is a conventional calculation method in this field and will not be repeated here.

[0048] Step S2, according to the corresponding power business generation rule, generate the power business information text corresponding to the original voice text; wherein the power business generation rule is the corresponding power business identification obtained through the original voice text, and the target power business generation rule is determined according to the power business identification; it can be understood that based on the original voice text, the power business identification is obtained; based on the power business identification, the target power business generation rule is determined; based on the original voice text, according to the power business generation rule, the power business information text is generated. After obtaining the original voice text, in order to better judge the type of power business that the user (i.e., the consulting object) wants to consult or handle, the original voice text can be preprocessed, such as removing stop words, stemming, word form restoration, etc., to improve the recognition accuracy of the business discrimination model.

[0049] It should be noted that after obtaining the power business identifier, the generation rules that match the power business identifier can be searched in the database based on the power business identifier, that is, the target power business generation rules. Among them, the rules of the power business information text may include grammatical conversion, semantic understanding, information extraction, template matching and other aspects. Their role is to convert the user's natural language input into structured information that can be understood and processed in the power business system.

[0050] When the power business identifier is a power resource consumption query identifier (i.e. used to indicate that the consulting object wants to query the power resource consumption), the target power business generation rule that matches the power resource consumption query identifier is "[number information], the household has currently used [**] degrees of electricity".

[0051] When the electricity business identifier is the electricity resource payment identifier (i.e. used to indicate that the consulting object wants to inquire about the pending electricity charges for electricity resources), the target electricity business generation rule that matches the electricity resource payment identifier is "[Number information], the household's current calculated electricity charge is ** yuan, [penalty for breach of contract is ** yuan, and the current total amount payable is ** yuan.]".

[0052] When the power business identifier is the power resource fault power outage identifier (that is, it is used to indicate that the consulting object wants to query the fault power outage information of the power resources in a certain area), the target power business generation rule that matches the power resource fault power outage identifier is "[Power usage address], affected by [power outage reason], a fault power outage occurred at [fault time]. Currently, the processing progress is [progress information], and the estimated power restoration time is [*day*hour*minute]. If you have been affected by the power outage, please wait patiently, our company is carrying out emergency repairs.".

[0053] When the power business identifier is the power resource power outage plan identifier (that is, it is used to indicate that the consulting object wants to query the planned power outage information of power resources in a certain area), the target power business generation rule that matches the power resource payment identifier is "[start and end time], [line name], a planned power outage for maintenance will be arranged, and the affected area is [power outage range]".

[0054] The electric power service identifier and the generation rule matching the electric power service identifier may also be expressed as other situations, which are not specifically limited here.

[0055] In one embodiment, the corresponding power business identification obtained through the original voice text includes: inputting the original voice text into a preset business discrimination model to obtain the basic business score of each power business; calculating the basic business score of each power business according to the preset business weight coefficient to obtain the comprehensive business score corresponding to each power business; outputting the preset identification of the power business corresponding to the largest comprehensive business score as the power business identification. It is understandable that the original voice data can be first converted into text information, that is, the original voice text, using voice recognition technology, and then, in the original voice text, natural language processing technology (such as keyword extraction algorithm) is used to identify keywords or phrases related to the consulting object, and then the identified keywords or phrases are matched with the preset consulting object identification library to determine the corresponding consulting object identification. Among them, the consulting object identification may be a unique identifier, name or code, which is used to locate a specific consulting object in the power business system.

[0056] Among them, the basic business score is a score given by the business discrimination model for each possible type of power business, which reflects the degree of match between the text and the business type. It can be understood that the higher the score, the stronger the correlation between the text and the power business. It can be understood that the higher the comprehensive business score, the higher the degree of match between the power business and the text input by the user, and it also reflects the priority or importance of the power business in the overall business processing. Among them, the business weight coefficient is set according to factors such as the priority, importance or frequency of user needs of the power business. It can be understood that these business weight coefficients can be used to reflect the relative importance of different business types in the overall business processing. The comprehensive business score can be specifically obtained by multiplying the basic business score by the corresponding business weight coefficient. In this way, even if the basic business scores of two businesses are the same, their comprehensive business scores will be different due to different weight coefficients.

[0057] Specifically, the business discrimination model is obtained through the following steps: inputting the sample speech text into the preset basic business discrimination model to determine the business discrimination probability corresponding to each of the sample speech texts; calculating the business discrimination loss value according to the business discrimination probability corresponding to each of the sample speech texts and the preset business information label; updating the parameters of the basic business discrimination model based on the business discrimination loss value until the accuracy of the basic business discrimination model reaches the preset threshold, and outputting the basic business discrimination model at this time as the business discrimination model. It should be noted that calculating the corresponding loss value is a conventional calculation method in this field and will not be repeated here.

[0058] In one embodiment, the step of generating the electric power business information text corresponding to the original voice text includes obtaining the consulting object identifier corresponding to the original voice data, and determining the consulting object number information according to the resource object identifier; generating the corresponding electric power business information text according to the electric power business generation rule based on the original voice text and the consulting object number information. It is understandable that before the step of generating the electric power business information text, the consulting object identifier corresponding to the original voice data is obtained; the consulting object number information is determined based on the resource object identifier; and the electric power business information text is generated according to the electric power business generation rule based on the original voice text and the consulting object number information.

[0059] In one embodiment, the step of determining the consulting object number information according to the resource object identifier includes: if the number information associated with the resource object identifier exists in the information database, the associated number information is output as the consulting object number information; if the number information associated with the resource object identifier does not exist in the information database, temporary number information is set for the resource object identifier according to a preset temporary information template, and the temporary number information is output as the consulting object number information. The consulting object identifier is used as a query condition to search for the corresponding number information (e.g., an electricity account, an electricity meter, a service request, an electricity address, etc.) in the database of the power business system, that is, the consulting object number information is extracted from the database query result, and the information will be used to generate the power business information text in the subsequent steps.

[0060] According to the power business generation rules, the power business information text is generated. According to the results of the power business generation rules analysis, the information related to the power business (such as power query, electricity bill payment, fault repair, etc.) is extracted from the original voice text, and combined with the consulting object number information, this information is integrated into structured power business information. Finally, the natural language generation technology (such as template filling, text splicing, etc.) is used to convert the structured power business information into a readable text form, that is, the power business information text. The text may contain information such as the user's specific needs, power business status, and operation instructions, and is used to provide users with clear and accurate power business feedback.

[0061] To sum up, according to the power business generation rules, the original voice text is converted into processable information in the power business system, and the automated process of generating specific power business information text can rely on accurate speech recognition, effective information extraction and integration, precise rule parsing and natural language generation technology to ensure that efficient and accurate power business services can be provided to the consulting objects.

[0062] Step S3, converting the power business information text into the corresponding target power business prompt voice through a preset conversion method, and broadcasting it to the consulting object. It is understandable that according to specific needs and resource conditions, a suitable speech synthesis technology (for example, splicing-based synthesis technology, parameter-based synthesis technology, and deep learning-driven synthesis technology, etc.) can be selected to convert the power business information text into the target power business prompt voice to achieve high-quality voice broadcasting.

[0063] In a specific embodiment, the received original voice data is first converted into corresponding text information, namely the original voice text, and then the associated power business identifier can be matched based on the original voice text, and the power business generation rule adapted to the power business identifier can be obtained, so that the power business information text can be generated more quickly and accurately in combination with the power business generation rule. This not only improves the accuracy of business identification or query, but also can be flexibly adjusted for different businesses. Then, the power business information text is converted into the target power business prompt voice and broadcast to the consulting object in real time. This can not only improve the accuracy and efficiency of voice recognition, but also help the consulting object to obtain information in a timely and accurate manner, thereby improving the intelligent effect of the power safety system to a certain extent.

[0064] An embodiment of the present invention further provides a speech recognition processing system for an electric power safety system, which is used to implement the speech recognition processing method for an electric power safety system, comprising:

[0065] A data acquisition module is used to acquire the original voice data of the consulting object and generate the corresponding original voice text according to the original voice data;

[0066] A text generation module, used to generate an electric power business information text corresponding to the original voice text according to a corresponding electric power business generation rule; wherein the electric power business generation rule is a corresponding electric power business identifier obtained through the original voice text, and a target electric power business generation rule determined according to the electric power business identifier;

[0067] The information prompt module is used to convert the power business information text into the corresponding target power business prompt voice through a preset conversion method, and broadcast it to the consulting object.

[0068] In a specific embodiment, the data acquisition module is further used to input the original voice data into a preset voice recognition model, and identify the voice text unit corresponding to the original voice data and the text probability distribution corresponding to the voice text unit through the voice recognition model;

[0069] The words corresponding to each speech text unit are determined by the text probability distribution corresponding to the speech text unit and a preset vocabulary, and all the words are combined into the corresponding original speech text.

[0070] The text generation module is also used to input the original voice text into a preset business discrimination model to obtain a basic business score for each power business;

[0071] The basic business score of each power business is calculated according to the preset business weight coefficient to obtain the comprehensive business score corresponding to each power business;

[0072] Outputting the preset identifier of the electric power business corresponding to the maximum comprehensive business score as the electric power business identifier;

[0073] Obtain the consultation object identifier corresponding to the original voice data, and determine the consultation object number information based on the resource object identifier; wherein, if the number information associated with the resource object identifier exists in the information database, the associated number information is output as the consultation object number information; if the number information associated with the resource object identifier does not exist in the information database, temporary number information is set for the resource object identifier according to a preset temporary information template, and the temporary number information is output as the consultation object number information;

[0074] According to the original voice text and the consulting object number information, the corresponding power business information text is generated according to the power business generation rule.

[0075] It should be noted that the system described in the above embodiment corresponds to the method described in the above embodiment. Therefore, the undetailed parts of the system described in the above embodiment can be obtained by referring to the contents of the method described in the above embodiment, and will not be repeated here.

[0076] In summary, the implementation of the embodiments of the present invention has the following beneficial effects:

[0077] The speech recognition processing method and system for the electric power safety system provided by the present invention can not only improve the accuracy and efficiency of speech recognition, but also help the consulting object to obtain information in a timely and accurate manner, thereby improving the intelligent effect of the electric power safety system to a certain extent.

[0078] The above disclosure is only the preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

Claims

1. A speech recognition processing method for an electric power safety system, characterized in that: include: Obtaining original voice data of the consulting object, and generating corresponding original voice text according to the original voice data; Generate an electric power business information text corresponding to the original voice text according to the corresponding electric power business generation rule; wherein the electric power business generation rule is a corresponding electric power business identifier obtained through the original voice text, and a target electric power business generation rule is determined according to the electric power business identifier; The power business information text is converted into a corresponding target power business prompt voice through a preset conversion method, and the voice is broadcast to the consulting object.

2. The method according to claim 1, characterized in that Generating the corresponding original voice text according to the original voice data includes inputting the original voice data into a preset voice recognition model, and identifying the voice text unit corresponding to the original voice data and the text probability distribution corresponding to the voice text unit through the voice recognition model; The words corresponding to each speech text unit are determined by the text probability distribution corresponding to the speech text unit and a preset vocabulary, and all the words are combined into the corresponding original speech text.

3. The method according to claim 2, characterized in that The speech recognition model is obtained by the following steps: inputting sample speech data into a preset basic speech recognition model, and obtaining a sample speech vector corresponding to the sample speech data, a corresponding sample text unit, and a basic text probability distribution corresponding to each sample text unit through an acoustic network in the basic speech recognition model; Calculate a corresponding text recognition loss value based on each of the sample text units, the basic text probability distribution corresponding to each of the sample text units, and the text unit label; The parameters of the basic speech recognition model are updated based on the text recognition loss value until the accuracy of the basic speech recognition model reaches a preset threshold, and the basic speech recognition model at this time is output as a speech recognition model.

4. The method according to claim 3, characterized in that The corresponding power business identification obtained through the original voice text includes inputting the original voice text into a preset business discrimination model to obtain a basic business score for each power business; The basic business score of each power business is calculated according to the preset business weight coefficient to obtain the comprehensive business score corresponding to each power business; The preset identifier of the electric power business corresponding to the maximum comprehensive business score is output as the electric power business identifier.

5. The method according to claim 4, characterized in that The business discrimination model is obtained by the following steps: inputting the sample speech text into the preset basic business discrimination model to determine the business discrimination probability corresponding to each of the sample speech texts; Calculate the business discrimination loss value according to the business discrimination probability corresponding to each of the sample speech texts and the preset business information label; The parameters of the basic business discrimination model are updated based on the business discrimination loss value until the accuracy of the basic business discrimination model reaches a preset threshold, and the basic business discrimination model at this time is output as a business discrimination model.

6. The method according to claim 5, characterized in that The generating of the electric power business information text corresponding to the original voice text includes obtaining the consulting object identifier corresponding to the original voice data, and determining the consulting object number information according to the resource object identifier; According to the original voice text and the consulting object number information, the corresponding power business information text is generated according to the power business generation rule.

7. The method according to claim 6, characterized in that Determining the consulting object number information according to the resource object identifier includes, if the number information associated with the resource object identifier exists in the information database, outputting the associated number information as the consulting object number information; If there is no numbering information associated with the resource object identifier in the information database, temporary numbering information is set for the resource object identifier according to a preset temporary information template, and the temporary numbering information is output as consulting object numbering information.

8. A speech recognition processing system for an electric power safety system, used to implement the method according to any one of claims 1 to 7, characterized in that: include: A data acquisition module is used to acquire the original voice data of the consulting object and generate the corresponding original voice text according to the original voice data; A text generation module, used to generate an electric power business information text corresponding to the original voice text according to a corresponding electric power business generation rule; wherein the electric power business generation rule is a corresponding electric power business identifier obtained through the original voice text, and a target electric power business generation rule determined according to the electric power business identifier; The information prompt module is used to convert the power business information text into the corresponding target power business prompt voice through a preset conversion method, and broadcast it to the consulting object.

9. The system according to claim 8, characterized in that The data acquisition module is also used to input the original voice data into a preset voice recognition model, and identify the voice text unit corresponding to the original voice data and the text probability distribution corresponding to the voice text unit through the voice recognition model; The words corresponding to each speech text unit are determined by the text probability distribution corresponding to the speech text unit and a preset vocabulary, and all the words are combined into the corresponding original speech text.

10. The system according to claim 9, characterized in that The text generation module is also used to input the original voice text into a preset business discrimination model to obtain a basic business score for each power business; The basic business score of each power business is calculated according to the preset business weight coefficient to obtain the comprehensive business score corresponding to each power business; Outputting the preset identifier of the electric power business corresponding to the maximum comprehensive business score as the electric power business identifier; Obtain the consultation object identifier corresponding to the original voice data, and determine the consultation object number information based on the resource object identifier; wherein, if the number information associated with the resource object identifier exists in the information database, the associated number information is output as the consultation object number information; if the number information associated with the resource object identifier does not exist in the information database, temporary number information is set for the resource object identifier according to a preset temporary information template, and the temporary number information is output as the consultation object number information; According to the original voice text and the consulting object number information, the corresponding power business information text is generated according to the power business generation rule.

Images