Voice recognition method, electronic device, storage medium, and computer program product

Abstract

The present disclosure provides a method for voice recognition, comprising: performing voice recognition on an audio file to obtain text information same as the content of the audio file; converting the text information into a prompt format text according to an optimization instruction in the text information, wherein the prompt format text is composed of a main text and an optimization prompt, and the optimization prompt is used to represent the optimization instruction; and performing optimization processing on the main text in the prompt format text in response to the optimization prompt in the prompt format text to obtain a written language text conforming to a text language rule. The present disclosure also provides an electronic device, a storage medium and a computer program product.

Description

Technical Field

[0001] This disclosure relates to the field of intelligent voice technology, and in particular to a voice recognition method, electronic device, storage medium, and computer program product. Background Technology

[0002] Speech recognition technology is the technology of converting speech signals into text signals. The goal of automatic speech recognition is to enable computers to recognize the continuous speech spoken by different people, realize the conversion of sound into text, and thus enable computers to understand human needs and perform the correct actions accordingly.

[0003] In related technologies, ASR (Automatic Speech Recognition) is commonly used for audio file recognition and text conversion. ASR technology converts sound waves in an audio file into multiple recognizable digital audio segments, and uses a neural network acoustic model to map these audio segments to text fields, ultimately determining the corresponding text information based on the context. By iterating through all audio segments in this way, the semantic text mapped to the audio file can be obtained. However, colloquial expressions in real-world scenarios are characterized by redundant interjections, filler words, loose structures, and repetitive descriptions. Existing speech recognition methods can only support one-to-one audio-to-text conversion and cannot avoid the characteristics of colloquial expressions to generate concise semantic text suitable for written language forms. Summary of the Invention

[0004] To address at least one of the problems described above, this disclosure provides a speech recognition method, electronic device, storage medium, and computer program product.

[0005] According to one aspect of this disclosure, a speech recognition method is provided, comprising: performing speech recognition on an audio file to obtain text information with the same word order as the audio file; converting the text information into prompt format text according to an optimization instruction in the text information, wherein the prompt format text consists of a main text and an optimization prompt, the optimization prompt representing the optimization instruction; and optimizing the main text in the prompt format text in response to the optimization prompt to obtain written text conforming to text language rules.

[0006] In some embodiments, performing speech recognition on an audio file to obtain text information with the same word order as the audio file includes: extracting an acoustic feature sequence contained in the audio file based on the sound waveform of the audio file; matching corresponding characters for each acoustic feature in the acoustic feature sequence; and combining the order of the characters according to the feature order of the acoustic feature sequence to construct a character sequence, and using the character sequence as the text information corresponding to the audio file, wherein the text information has the same word order as the audio file.

[0007] In some implementations, converting the text information into prompt format text based on the optimization instructions in the text information includes: extracting optimization instructions and main text from the text information, wherein the optimization instructions are used to characterize the optimization direction of the main text; converting the optimization instructions into a preset prompt format to obtain an optimized prompt with the preset prompt format; and integrating the optimized prompt with the main text to construct the prompt format text containing the main text and the optimized prompt.

[0008] In some implementations, optimizing the main text in the prompt format text in response to the optimized prompt in the prompt format text to obtain written text that conforms to the text language rules includes: locating question fragments in the main text that contain question types indicated by the optimized prompt, wherein the question types include at least redundancy of interjections, repetition of adjacent sentences, and filler words; correcting the question fragments into text fragments that conform to the text language rules based on pre-configured text language rules; and replacing the question fragments with the text fragments to construct the written text containing the text fragments.

[0009] In some implementations, after optimizing the main text in the prompt format text in response to the optimization prompt in the prompt format text to obtain written text that conforms to the text language rules, the process includes: receiving a user's modification instruction and personalizing the written text so that the written text is edited into output text that conforms to the user's language habits.

[0010] In some implementations, this includes: calculating the confirmation rate of the written text as the output text.

[0011] In some implementations, after optimizing the main text in the prompt format text in response to the optimized prompt in the prompt format text to obtain written text that conforms to the text language rules, the process includes: randomly selecting multiple written texts for verification to obtain verification results; and marking the corresponding written texts with questions based on the questions raised by the verification results, and determining the expected output text of the audio file.

[0012] According to another aspect of this disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the speech recognition method as described in any of the above embodiments.

[0013] According to another aspect of this disclosure, a readable storage medium is provided that stores a computer program adapted for loading by a processor to perform the speech recognition method as described in any of the above embodiments.

[0014] According to another aspect of this disclosure, a computer program product is provided, including a computer program / instructions that, when executed by a processor, implement the speech recognition method as described in any of the above embodiments. Attached Figure Description

[0015] The accompanying drawings illustrate exemplary embodiments of the present disclosure and, together with the description thereof, serve to explain the principles of the present disclosure. These drawings are included to provide a further understanding of the present disclosure and are incorporated in and constitute a part of this specification.

[0016] Figure 1 This is a flowchart of a speech recognition method according to one embodiment of the present disclosure.

[0017] Figure 2 This is a diagram illustrating the architecture of a speech recognition method according to one embodiment of the present disclosure.

[0018] Figure 3 This is a configuration diagram of a speech recognition system according to one embodiment of the present disclosure.

[0019] Figure 4 This is a block diagram of a speech recognition device according to one embodiment of the present disclosure. Detailed Implementation

[0020] The present disclosure will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the disclosure. Furthermore, it should be noted that, for ease of description, only the parts relevant to the present disclosure are shown in the accompanying drawings.

[0021] It should be noted that, where there is no conflict, the embodiments and features described in this disclosure can be combined with each other. The technical solutions of this disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0022] Unless otherwise stated, the exemplary implementations / embodiments shown are to be understood as providing exemplary features of various details that provide ways in which the technical concepts of this disclosure can be implemented in practice. Therefore, unless otherwise stated, the features of various implementations / embodiments may be additionally combined, separated, interchanged and / or rearranged without departing from the technical concepts of this disclosure.

[0023] The terminology used herein is for the purpose of describing particular embodiments and is not restrictive. As used herein, unless the context clearly indicates otherwise, the singular forms “a” and “the” are intended to include the plural forms as well. Furthermore, when the terms “comprising” and / or “including” and variations thereof are used in this specification, it indicates the presence of the stated features, integrals, steps, operations, parts, components, and / or groups thereof, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, parts, components, and / or groups thereof. It should also be noted that, as used herein, the terms “substantially,” “about,” and other similar terms are used as approximate terms rather than as terms of degree, thus explaining the inherent biases in measurements, calculated values, and / or provided values ​​that would be recognized by one of ordinary skill in the art.

[0024] Figure 1 Here is a flowchart of a speech recognition method according to one embodiment of the present disclosure; Figure 2 This is a diagram illustrating the architecture of a speech recognition method according to one embodiment of the present disclosure. The following will be combined with… Figure 1 and Figure 2 The steps of the speech recognition method S100 are described in detail.

[0025] Step S102: Perform speech recognition on the audio file to obtain text information with the same word order as the audio file.

[0026] An audio file is the sound signal input by the user through a voice input module, containing the information expressed by the user. Essentially, an audio file is a recording of the user's speech; different content will result in different sound signals in the audio file. The audio file faithfully records the user's interjections, repetitions, filler words, and / or inverted sentences, among other colloquial expressions. Of course, along with the user's voice commands, optimization instructions for the main content are also provided. In other words, the audio file contains both the main audio content and the optimization instruction audio content.

[0027] Speech recognition is the process of using a speech recognition model to convert the sound signal of an audio file into text information. The speech recognition model can be an ASR (Automatic Speech Recognition) model to achieve this conversion. During speech recognition, the conversion is performed one-to-one based on the sound waveform in the audio file; no deletions or optimizations are made to the audio file's content.

[0028] Sentence order content refers to the order in which sentences are expressed and the meaning of the content when a user speaks. An audio file is a recording of the user's speaking process, so its sentence order content is the same as the sentence order content when the user speaks.

[0029] Text information is the result of converting audio files using a speech recognition model. It contains all the content mentioned in the audio file and is a complete representation of the recorded content. In other words, because the speech recognition model does not make any adjustments to the audio file, the text information includes colloquial expressions used by the user, such as interjections, repetitions, inversions, and filler words. That is to say, the word order of the text information is consistent with that of the audio file. Based on the user's colloquial expressions, the content presented in the text information is difficult to directly represent as a written form of language.

[0030] Based on the content structure of the audio file, the text information consists of the main text and optimization instructions. The main text corresponds to the main audio content in the audio file, while the optimization instructions correspond to the audio optimization instructions in the audio file.

[0031] Step S104: Convert the text information into prompt format text according to the optimization instructions in the text information.

[0032] Optimization commands are used to indicate the direction of optimization for the main body of text information. They are generated by the user based on their own optimization needs for the main body text. Optimization commands can be, for example, to remove unimportant information from the main body text. They can also be expressed in more detail as: removing redundant interjections from the main body text, removing repetitive phrases from the main body text, etc., which will not be listed here.

[0033] The cue format text is the result of formatting text information. It has the advantage of easily identifying the optimization direction, clearly marking the optimization instructions in the main text, which is beneficial for subsequent language optimization model recognition. The cue format text consists of optimization hints and the main text; the optimization hints are the formatted result of the optimization instructions.

[0034] The format of the prompt text is pre-configured by operations personnel based on the most easily recognizable format according to the needs of low-cost and high-efficiency conversion or language optimization model. For example: Please remove unimportant information in the "Main Text"; or delete useless information in the "Main Text", etc., not listed here.

[0035] The language optimization model can be an LLM (Large Language Model). It is used to optimize the main text based on optimization prompts to obtain the prompt formatted text.

[0036] Step 106: In response to the optimization prompt in the prompt format text, optimize the main text in the prompt format text to obtain written text that conforms to the text language rules.

[0037] The optimization process involves calling a language optimization model to respond to optimization suggestions in the main text of the text information. After optimization, the text information overcomes the problems indicated in the optimization suggestions, making it conform to the normative requirements of the text language rules.

[0038] Textual language rules are pre-defined normative requirements for written text, conforming to established writing conventions. Specifically, textual language rules might include, for example, the absence of redundant modifiers or repeated phrases. Of course, rules can also be set based on the actual needs of the written text in the application scenario; therefore, the specific content of the textual language rules is not limited here.

[0039] Written text is an optimized version of the main body of textual information. For example, it removes interjections that indicate pauses during the user's speech or repetitive descriptions of their speech. Written text conforms to the rules of written language, possesses a formal, written style, and clearly and concisely conveys the correct meaning, making it easy to read and understand.

[0040] In some implementations, the execution of step S102 includes: extracting the acoustic feature sequence contained in the audio file based on the sound waveform of the audio file; matching corresponding characters for each acoustic feature in the acoustic feature sequence; and combining the order of characters according to the feature order of the acoustic feature sequence to construct a character sequence, and using the character sequence as text information corresponding to the audio file, wherein the text information has the same word order as the audio file.

[0041] The aforementioned process in step S102 is implemented by calling a speech recognition model, which can be an ASR model. The ASR model is mainly used to convert the sound signals in the audio file into text information one by one. During the conversion process, it does not consider the colloquial expressions in the user's audio file and performs text conversion indiscriminately.

[0042] For example, when a user inputs an audio file containing the phrase, "The only thing it affects right now is the loan term because it expires in 2043, so right now, it's probably around twenty years if it's a second loan—this is unimportant information," the ASR model will faithfully transcribe each word in the audio, forming the text version, "The only thing it affects right now is the loan term because it expires in 2043, so right now, it's probably around twenty years if it's a second loan—this is unimportant information." Clearly, after being recognized by the ASR model, the resulting text accurately records the user's non-important information, such as interjections like "uh."

[0043] In some implementations, step S104 is specifically implemented by: extracting optimization instructions and main text from the text information, wherein the optimization instructions are used to characterize the optimization direction of the main text; converting the optimization instructions into a preset prompt format to obtain an optimization prompt with the preset prompt format; and integrating the optimization prompt and the main text to construct a prompt format text containing the main text and the optimization prompt.

[0044] The text message contains optimization instructions and main text. The optimization instructions indicate the direction of optimization for the main text and represent the user's optimization requirements for the main text. The main text is the core information in the text message, which is the content actually expressed by the user and is also the main object of optimization.

[0045] The preset prompt format is set according to the needs of users and operations personnel. It is a relatively time-saving and efficient format during the construction process, and / or a format that is easy for language optimization models to recognize. For example, if the text information is "What it affects now is the loan term because it expires in 2043, so now it is probably around 20 years if it is a second loan. This sentence is not important information," the optimized prompt with the preset prompt format can be "Please remove the unimportant information from the sentence 'Main Text'."

[0046] Referring to the previous example, after integrating and optimizing the prompt and main text, the constructed prompt format text can be: "Please remove the unimportant information from the sentence 'What it currently only affects is the loan term, because it expires in 2043, so currently, if it's a second loan, it's approximately twenty years.'" Clearly, after the format conversion,

[0047] In some implementations, step S106 is performed as follows: based on the problem type indicated by the optimization prompt, the problem fragments in the main text that contain the problem type are located, wherein the problem type includes at least redundancy of interjections, repetition of adjacent sentences, and verbal tics; based on pre-configured text language rules, the problem fragments are corrected into text fragments that conform to the text language rules; and the problem fragments are replaced with text fragments to construct a written text containing the text fragments.

[0048] For example, if the optimization prompt indicates that the problem type is "unimportant information exists", then it is necessary to locate the area of ​​"unimportant information" in the main text, and then optimize the problem fragment in that area through the language optimization model, that is, "delete unimportant information".

[0049] Referring to the previous example, if the prompt text is "Please remove the unimportant information in the sentence 'What it currently only affects is the loan term, because it expires in 2043, so currently, if it's a second loan, it's about 20 years,'" then the language optimization model will determine that the problem type in the main text is "existence of unimportant information," and then identify the location of the unimportant information in the main text and perform the deletion operation. Ultimately, the optimized result is "What it currently only affects is the loan term, because it expires in 2043, so currently, if it's a second loan, it's about 20 years."

[0050] In some implementations, after step S106, the method further includes: receiving a user's modification instruction and personalizing the written text so that the written text is edited into an output text that conforms to the user's language habits.

[0051] Editing instructions are the user's suggestions for modifying the written text output by the language optimization model. Because each user has different language habits or optimization needs for each subject text, and because the language optimization model may also have some instability during the optimization process, users may perform some editing operations on some written texts, thereby generating output text that meets the user's needs.

[0052] Then, the user can click send to send the output text to the destination.

[0053] In order to determine the availability of the written text generated through the aforementioned steps S102 to S106, this disclosure also includes the following step: calculating the confirmation rate of the written text as the output text.

[0054] The confirmation rate of output text using written language is essentially the rate at which users output written language text without editing. As one of the standards for evaluating the performance of the aforementioned module parameters, the confirmation rate can intuitively present the strengths and weaknesses of these modules. It provides direction for model optimization or process improvement, ensuring the highest possible alignment between written language text and user needs.

[0055] In some implementations, after step S106, the method further includes: after obtaining the written text, saving the audio file and the corresponding ASR result (i.e., text information), the prompt format text output by the prompt model, and the written text.

[0056] In some implementations, after step S106, the method further includes: randomly selecting multiple written texts for verification to obtain verification results; and based on the questions raised in the verification results, marking the corresponding written texts with questions and determining the expected output text of the audio file.

[0057] To ensure verification efficiency, it is not necessary to verify every single written text output; random sampling is sufficient. For example, within a certain verification cycle, multiple written texts and their corresponding audio files are selected to verify the consistency between the written text and the audio files, including whether the main text clearly expresses its meaning and whether optimization instructions have been executed effectively.

[0058] Furthermore, the question annotations and expected output text are fed back to the various models involved in generating written text, such as the ASR model, prompt model, and LLM model, to adjust the parameters of these models. Through continuous iteration of model parameters, the output written text is made closer to the expected output text, ensuring the formality, conciseness, and readability of the text obtained during the text conversion process.

[0059] Finally, based on the verification results, indicators such as the accuracy of intelligent recognition are evaluated.

[0060] Figure 3 This is a configuration diagram of a speech recognition system according to one embodiment of the present disclosure. The following is in conjunction with… Figure 3 This disclosure provides a detailed description of the speech recognition system involved.

[0061] A speech recognition system is designed to implement the various steps of speech recognition methods. It provides a voice input module for users, an operation configuration module for operators, a data module, and an annotation module.

[0062] The core of the voice input module is user-centric, providing users with functions such as recording, text transfer, text editing, and sending. Specifically, the user recording function obtains the user's audio file; the text transfer function, provided by the ASR model, converts the audio file into text to obtain text information; and the text editing function provides users with the space to personalize written text.

[0063] The core of the operations configuration module is geared towards operations personnel, providing them with functions such as configuring prompts (i.e., preset prompt formats for optimization tips), implementing error correction configurations, and configuring proper nouns. The preset prompt formats for optimization tips are implemented using the prompt model.

[0064] The data module is used for audio storage and the storage of processing results, as well as for statistics on related data, such as the non-editing transmission rate (i.e., the confirmation rate of written text).

[0065] The annotation module is used for data extraction, providing annotation tools, managing annotation tasks, and generating annotation data. It obtains written text to be verified through manual or machine-generated random extraction, annotates issues within the text, and finally uploads the results to the statistical indicator platform. The extracted data originates from the data module.

[0066] More specifically, the main application scenarios for speech recognition methods are intelligent voice input functions within IM (Instant Messaging) or message management assistant functions within IM.

[0067] When the speech recognition method disclosed herein is implemented in a specific product, the prompt arrangement and prompt response processing of each application should include the following capabilities: process arrangement capability, basic control capability, special scenario circuit breaker capability, and cross-system interaction capability.

[0068] The process orchestration capabilities include: multi-template execution process, supporting logical connection between multiple sets of templates, such as intent recognition and scene processing using the same preset prompt format template; multi-template switching, building a complete template switching capability, such as being able to easily handle customers switching topics freely in multiple rounds of dialogue, reducing the risk of being interrupted by circuit breakers.

[0069] Basic control capabilities include: circuit breaking for time delays or information overload, circuit breaking for excessively long interface call times, and circuit breaking for response information exceeding preset formats.

[0070] Special scenario circuit breaker capabilities include: circuit breaker for sensitive or abnormal responses, and monitoring of sensitive information and abnormal callbacks generated by LLM.

[0071] Cross-system interaction capabilities include: prompting responses to external interface calls, response information for LLM callbacks, identification of valid system actions using the same formatting protocol, such as recording the validity of a single output on the data sending side, and supporting data indexing.

[0072] The prompt engineering includes a prompt engine and prompt operations. The prompt engine includes prompt template invocation and execution. Prompt operations include prompt template content production, prompt template quality management, and prompt template application management.

[0073] The prompt template call execution is used to perform specific prompt actions based on the template ID requested from the upstream request. The prompt encapsulates the LLM interaction by establishing an interface with the LLM and receiving a formatted response after inputting the prompt.

[0074] The prompt template content generation is based on a knowledge base, producing prompt templates for each scenario or orchestration unit. Prompt template quality management ensures the accuracy of the prompt template execution results, guaranteeing their safety and stability. Prompt template application management supports the entire process of connecting upstream product libraries to the production configuration limit.

[0075] When using the speech recognition method disclosed herein, if the input main text is "Previously, this was all furnished by the developer, but he thought it was too ugly. Since he was planning to live there himself, he probably demolished the developer's entire interior. After demolishing half of it, his wife didn't want to come to Suzhou anymore, so he put the house up for sale, and the interior decorations weren't installed," the output written text is "Previously, this was all furnished by the developer, but he thought it was too ugly. Since he was planning to live there himself, he probably demolished the developer's entire interior. After demolishing half of it, his wife didn't want to come to Suzhou anymore, so he put the house up for sale, and the interior decorations weren't installed." When the input text is "Ah, it's already being filmed," the output written text is "It's already being filmed." When the input text is "Your deposit agreement is useless without a signature; please edit a written version." When the input text is "For China Construction Jinan No. 1, no, for China Construction Jinan No. 2, it's valid only for the day of filing. If you don't show it today, you'll have to re-file tomorrow.", the output written text is "For China Construction Jinan No. 2, it's valid only for the day of filing. If you don't show it today, you'll have to re-file tomorrow."

[0076] The speech recognition method proposed in this disclosure converts the text information of an audio file into a prompt format text through a pre-set prompt format. This prompt format text guides a language optimization model to optimize the main text based on the prompt format text, thereby removing colloquial and useless content from the main text and obtaining a concise and readable written text, thus improving the user's reading experience.

[0077] Figure 4 This is a block diagram of a speech recognition device according to one embodiment of the present disclosure. The following is in conjunction with… Figure 4 The speech recognition device 1000 is described.

[0078] This disclosure proposes a speech recognition device 1000, comprising: an audio conversion module 1002 for performing speech recognition on an audio file to obtain text information with the same word order as the audio file; an optimization prompt matching module 1004 for converting the text information into prompt format text according to optimization instructions in the text information, wherein the prompt format text consists of main text and optimization prompts, and the optimization prompts are used to represent optimization instructions; and an optimization processing module 1006 for optimizing the main text in the prompt format text in response to optimization prompts in the prompt format text to obtain written text that conforms to the rules of textual language.

[0079] The various modules in the speech recognition device 1000 are proposed to implement the various steps of the speech recognition method. The execution steps and implementation principles can be referred to the foregoing and will not be repeated here.

[0080] The device 1000 may include corresponding modules that perform one or more steps in the flowchart described above. Therefore, each or more steps in the flowchart can be performed by a corresponding module, and the device may include one or more of these modules. A module may be one or more hardware modules specifically configured to perform a corresponding step, or implemented by a processor configured to perform a corresponding step, or stored in a computer-readable medium for implementation by a processor, or implemented through some combination thereof.

[0081] This hardware architecture can be implemented using a bus architecture. The bus architecture can include any number of interconnect buses and bridges, depending on the specific application and overall design constraints of the hardware. Bus 1100 connects various circuits, including one or more processors 1200, memory 1300, and / or hardware modules. Bus 1100 can also connect various other circuits 1400, such as peripherals, voltage regulators, power management circuits, external antennas, etc.

[0082] Bus 1100 can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Component (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, only one connection line is used in this diagram, but this does not imply that there is only one bus or only one type of bus.

[0083] Any process or method description in the flowcharts or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of this disclosure includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of this disclosure pertain. The processor performs the various methods and processes described above. For example, the method embodiments of this disclosure may be implemented as software programs tangibly contained in a machine-readable medium, such as memory. In some embodiments, part or all of the software program may be loaded and / or installed via memory and / or a communication interface. When the software program is loaded into memory and executed by the processor, one or more steps of the methods described above may be performed. Alternatively, in other embodiments, the processor may be configured to perform one of the methods described above by any other suitable means (e.g., by means of firmware).

[0084] The logic and / or steps represented in the flowchart or otherwise described herein may be specifically implemented in any readable storage medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a processor-included system or other system that can fetch and execute instructions from, an instruction execution system, apparatus or device).

[0085] For the purposes of this specification, a "readable storage medium" can be any means capable of containing, storing, communicating, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and portable read-only memory (CDROM). Furthermore, a readable storage medium can even be paper or other suitable media on which a program can be printed, since a program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in memory.

[0086] It should be understood that various parts of this disclosure can be implemented in hardware, software, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0087] Those skilled in the art will understand that all or part of the steps of the methods described above can be implemented by a program instructing related hardware, and the program can be stored in a readable storage medium. When executed, the program includes one or a combination of the steps of the method implementation.

[0088] Furthermore, the functional units in the various embodiments of this disclosure can be integrated into a single processing module, or each unit can exist physically separately, or two or more units can be integrated into a single module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a readable storage medium. The storage medium can be a read-only memory, a disk, or an optical disk, etc.

[0089] Those skilled in the art should understand that the above embodiments are merely for illustrating the present disclosure and are not intended to limit the scope of the disclosure. Those skilled in the art can make other changes or modifications based on the above disclosure, and these changes or modifications still fall within the scope of the present disclosure.

Claims

1. A speech recognition method, characterized in that, include: Perform speech recognition on the audio file to obtain text information with the same word order as the audio file; Based on the optimization instructions in the text information, the text information is converted into prompt format text, wherein the prompt format text consists of main text and optimization prompts, and the optimization prompts are used to characterize the optimization instructions; In response to the optimized prompt in the prompt format text, the main text in the prompt format text is optimized to obtain written text that conforms to the rules of text language; The step of converting the text information into prompt format text based on the optimization instructions in the text information includes: extracting optimization instructions and main text from the text information, wherein the optimization instructions are used to characterize the optimization direction of the main text; converting the optimization instructions into a preset prompt format to obtain an optimized prompt with the preset prompt format; and integrating the optimized prompt with the main text to construct the prompt format text containing the main text and the optimized prompt. The method of optimizing the main text in the prompt format text in response to the optimized prompt text to obtain written text that conforms to the text language rules includes: locating question fragments in the main text that contain the question type indicated by the optimized prompt, wherein the question type includes at least redundancy of interjections, repetition of adjacent sentences, and filler words; correcting the question fragments into text fragments that conform to the text language rules based on the pre-configured text language rules; and replacing the question fragments with the text fragments to construct the written text containing the text fragments.

2. The speech recognition method according to claim 1, characterized in that, The step of performing speech recognition on the audio file to obtain text information with the same word order as the audio file includes: Based on the sound waveform of the audio file, extract the acoustic feature sequence contained in the audio file; Matching corresponding characters to each acoustic feature in the acoustic feature sequence; and The order of the characters is combined according to the characteristic order of the acoustic feature sequence to construct a character sequence, and the character sequence is used as the text information corresponding to the audio file, wherein the text information has the same word order as the audio file.

3. The speech recognition method according to claim 1, characterized in that, After optimizing the main text in the prompt format text in response to the optimized prompt in the prompt format text to obtain written text that conforms to the rules of textual language, the process includes: The system receives modification instructions from the user and personalizes the written text to make it output text that conforms to the user's language habits.

4. The speech recognition method according to claim 3, characterized in that, include: The accuracy rate of the output text using the written text is statistically analyzed.

5. The speech recognition method according to claim 1, characterized in that, After optimizing the main text in the prompt format text in response to the optimized prompt in the prompt format text to obtain written text that conforms to the rules of textual language, the process includes: Multiple written texts were randomly selected for verification to obtain verification results; and Based on the questions raised by the verification results, the corresponding written text is labeled with questions, and the expected output text of the audio file is determined.

6. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the speech recognition method as described in any one of claims 1 to 5.

7. A readable storage medium, characterized in that, The readable storage medium stores a computer program adapted for loading by a processor to perform the speech recognition method as described in any one of claims 1 to 5.

8. 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 speech recognition method according to any one of claims 1 to 5.

Images