A marking method, device, equipment and storage medium

Abstract

The embodiments of the present application provide a marking method, apparatus, device and storage medium, the method comprising: obtaining a target phoneme in the audio to be marked and a wake-up word; determining the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme, and when the similarity meets a preset condition, marking the target audio with a feature tag, the target audio being the head-end audio of the audio to be marked, and the feature tag characterizing that the target audio is the residual audio of the wake-up word. The method provided in the embodiments of the present application can improve the efficiency of marking the residual audio of the wake-up word in the audio.

Description

Technical Field

[0001] The present application belongs to the field of speech recognition, and in particular, relates to a marking method, device, equipment and storage medium. Background Art

[0002] Voice interaction is increasingly used in various industries, and voice wake-up technology is the gateway to voice interaction with terminals. Currently, terminals can sometimes switch from standby mode to working mode without collecting a complete wake-up word. The residual wake-up word collected after switching to working mode will have an adverse effect on subsequent interactions between users and the terminal.

[0003] When technicians optimize speech recognition methods to reduce this adverse effect, they need to first manually mark the residual audio of the wake-up word in the audio.

[0004] Currently, manually labeling the residual audio of wake-up words in audio is inefficient. Summary of the invention

[0005] Embodiments of the present application provide a marking method, apparatus, device, and storage medium that can improve the efficiency of marking residual audio of a wake-up word in audio.

[0006] In a first aspect, an embodiment of the present application provides a marking method, the method comprising:

[0007] Obtain the target phonemes in the audio to be labeled and the wake-up word;

[0008] Determine the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme. When the similarity meets the preset conditions, use the feature label to mark the target audio. The target audio is the first-end audio of the audio to be marked, and the feature label represents that the target audio is the residual audio of the wake-up word.

[0009] In a possible implementation, the audio to be marked includes audio of instruction information, and the method further includes:

[0010] Receiving a user's marking operation on the audio of the instruction information and an instruction text corresponding to the instruction information input by the user;

[0011] In response to the marking operation, marking the audio of the instruction information in the audio to be marked with an instruction tag;

[0012] The audio to be marked with feature labels and command labels is recorded as marked audio. The command recognition model is trained based on the marked audio and the command text of the marked audio to obtain a target command recognition model. The target command recognition model is used to eliminate the residual audio of the wake-up word in the marked audio, recognize the audio of the command information, and obtain the command text corresponding to the command information.

[0013] In a possible implementation, before determining the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme, the method further includes:

[0014] The target audio is determined according to the amplitude variation rule of the head-end audio of the audio to be marked.

[0015] In a possible implementation, determining the target audio according to the amplitude variation rule of the head-end audio of the audio to be marked includes:

[0016] The audio whose amplitude variation rule at the beginning of the audio to be marked is from high to zero is determined as the target audio.

[0017] In a possible implementation, the target phoneme includes a phoneme of a preset length at the end of the wake-up word.

[0018] In a possible implementation, determining the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme, and marking the target audio with a feature tag when the similarity meets a preset condition, includes:

[0019] The labeling model is used to determine the similarity between the first phoneme of the target audio in the audio to be labeled and the target phoneme. When the similarity meets the preset conditions, the target audio is labeled with a feature tag.

[0020] In a possible implementation, a labeling model is used to determine the similarity between a first phoneme of a target audio in the audio to be labeled and a target phoneme. When the similarity satisfies a preset condition, before labeling the target audio with a feature tag, the method further includes:

[0021] Obtain positive samples and negative samples, where the positive sample includes the residual audio of the wake-up word at the beginning, and the negative sample does not include the residual audio of the wake-up word at the beginning;

[0022] According to the positive samples and negative samples, the model to be trained is trained to obtain a labeled model.

[0023] In a second aspect, an embodiment of the present application provides a marking device, the device comprising:

[0024] An acquisition module is used to acquire target phonemes in the audio to be labeled and the wake-up word;

[0025] The marking module is used to determine the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme. When the similarity meets the preset conditions, the target audio is marked with a feature label. The target audio is the first-end audio of the audio to be marked, and the feature label represents that the target audio is the residual audio of the wake-up word.

[0026] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, it implements the method in the first aspect or any possible implementation manner of the first aspect.

[0027] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method in the first aspect or any possible implementation of the first aspect is implemented.

[0028] In a fifth aspect, an embodiment of the present application provides a computer program product. When instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes a method as in the first aspect or any possible implementation of the first aspect.

[0029] The embodiments of the present application provide a marking method, device, equipment and storage medium, which first obtains the target phoneme in the audio to be marked and the wake-up word; then, determines the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme, and when the similarity meets the preset conditions, marks the target audio with a feature tag, where the target audio is the head-end audio of the audio to be marked, and the feature tag characterizes the target audio as the residual audio of the wake-up word. The audio in the head-end audio of the audio to be marked that meets the preset conditions of similarity with the residual phoneme of the wake-up word is automatically marked with a feature tag as the residual audio of the wake-up word, thereby improving the efficiency of marking the residual audio of the wake-up word in the audio. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the technical solution of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0031] Figure 1 It is a schematic diagram of a flow chart of a marking method provided in an embodiment of the present application;

[0032] Figure 2 It is a structural schematic diagram of a marking device provided in an embodiment of the present application;

[0033] Figure 3 It is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0034] The features and exemplary embodiments of various aspects of the present application will be described in detail below. In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only configured to explain the present application and are not configured to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is only to provide a better understanding of the present application by illustrating the examples of the present application.

[0035] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "include..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0036] Voice interaction is increasingly used in various industries, and voice wake-up technology is the gateway to voice interaction with the terminal. At present, the terminal can sometimes switch from standby to working state when the complete wake-up word is not collected. After switching to working state, it continues to collect the remnants of the wake-up word and collects the user's voice command. The remnants of the wake-up word and the voice command are input into the command recognition model to recognize the user's command. However, due to the presence of the wake-up word residue, it will be difficult for the command recognition model to accurately recognize the user's command. Faced with the problem that the command recognition model is difficult to accurately recognize the user's command due to the presence of the wake-up word residue, the technicians need to optimize the command recognition model to improve the accuracy of recognizing commands. First, it is necessary to manually filter out the audio including the residual audio of the wake-up word in a large amount of audio, and mark the residual audio of the wake-up word to obtain samples for training the command recognition model, but the efficiency of manually screening and marking the residual audio of the wake-up word is low.

[0037] In response to the above problems, the embodiments of the present application provide a marking method, apparatus, device and storage medium, which first obtains the target phoneme in the audio to be marked and the wake-up word; then, determines the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme, and when the similarity meets the preset conditions, marks the target audio with a feature tag, where the target audio is the head-end audio of the audio to be marked, and the feature tag characterizes the target audio as the residual audio of the wake-up word. Automatically marking the audio in the head-end audio of the audio to be marked with a feature tag, where the audio whose similarity with the residual phoneme of the wake-up word meets the preset conditions, as the residual audio of the wake-up word, improves the efficiency of marking the residual audio of the wake-up word in the audio.

[0038] The method provided in the embodiment of the present application is performed by a terminal having audio processing capabilities, such as a computer.

[0039] The following will be combined Figure 1 A marking method provided in an embodiment of the present application is described in detail.

[0040] like Figure 1 As shown, the method may include the following steps:

[0041] S110, obtaining target phonemes in the audio to be marked and the wake-up word.

[0042] The terminal obtains the target phonemes in the audio to be labeled and the wake-up word.

[0043] The audio to be marked is the audio received after the terminal switches to the working state. When the user wakes up the terminal with voice, if the terminal is awakened before receiving the complete wake-up word, the terminal continues to receive the residual voice of the wake-up word after switching to the working state, and accepts the user's voice command. Therefore, the audio to be marked includes the command information and may also include the residual audio of the wake-up word. The command information represents the user's command to the terminal, and the terminal responds accordingly based on the command information. The target phonemes include the residual phonemes of the wake-up word.

[0044] In one example, the wake-up word is "baby". If the terminal is awakened when receiving "baby", the audio received after the terminal switches to the working state includes the residual of the wake-up word: "bei". The target phoneme in the wake-up word includes the residual phoneme of the wake-up word, that is, the phoneme "bei" of "bei".

[0045] S120, determining the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme, and when the similarity meets a preset condition, marking the target audio with a feature tag.

[0046] The target audio is the head-end audio of the audio to be labeled, and the feature label indicates that the target audio is the residual audio of the wake-up word.

[0047] Since the residual audio of the wake word is the first audio received after the terminal switches to the working state, the terminal first recognizes the first phoneme of the target audio at the start of the audio to be marked, determines the similarity between the first phoneme and the target phoneme. When the similarity meets the preset condition, it indicates that the target audio is the residual audio of the wake word, and the target audio is marked with a feature label.

[0048] In one example, the wake word is "baby", the target phoneme is the phoneme "bei" of "贝", and the preset condition is that the similarity is greater than 60%. The terminal recognizes that the first phoneme of the target audio is "bei", so the similarity between the first phoneme and the target phoneme is 100%, meeting the preset condition. Therefore, the terminal determines that the target audio is the residual audio of the wake word and marks the target audio with a feature label.

[0049] In one example, the wake word is "baby", the target phoneme is the phoneme "bei" of "贝", and the preset condition is that the similarity is greater than 60%. The terminal recognizes that the first phoneme of the target audio is "ei" or "pei", so the similarity between the first phoneme and the target phoneme is 66.7%, meeting the preset condition. Therefore, the terminal determines that the target audio is the residual audio of the wake word and marks the target audio with a feature label.

[0050] In one example, the wake word is "baby", the target phoneme is the phoneme "bei" of "贝", and the preset condition is that the similarity is greater than 60%. The terminal recognizes that the first phoneme of the target audio is "i", so the similarity between the first phoneme and the target phoneme is 33.3%, not meeting the preset condition. Therefore, the terminal determines that the target audio is not the residual audio of the wake word and does not mark the target audio with a feature label.

[0051] The method provided by the embodiments of the present application first obtains the audio to be marked and the target phoneme in the wake word; then, determines the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme. When the similarity meets the preset condition, the target audio is marked with a feature label. The target audio is the start audio of the audio to be marked, and the feature label indicates that the target audio is the residual audio of the wake word. Automatically marking the audio in the start audio of the audio to be marked whose similarity with the residual phoneme of the wake word meets the preset condition as the residual audio of the wake word improves the efficiency of marking the residual audio of the wake word in the audio.

[0052] In some embodiments, the audio to be marked includes the audio of the instruction information, and the method may further include the following steps:

[0053] First, receive the marking operation of the user on the audio of the instruction information and the instruction text corresponding to the instruction information input by the user.

[0054] The instruction information represents the instruction of the user to the terminal, and the terminal makes corresponding responses according to the instruction information.

[0055] In order to train the instruction recognition model and enable it to more accurately recognize instruction information in the audio, the technicians mark the audio of the instruction information in the audio to be marked and obtain the instruction text corresponding to the instruction information.

[0056] The command text represents the text of the command information. For example, the command text of a command information is "turn on the light".

[0057] The user can operate the terminal through the human-computer interaction interface to mark the audio of the instruction information to be marked and enter the instruction text corresponding to the instruction information. The terminal receives the user's marking operation on the audio of the instruction information and the instruction text corresponding to the instruction information entered by the user.

[0058] Secondly, in response to the marking operation, the audio of the instruction information in the audio to be marked is marked with an instruction tag.

[0059] In response to the marking operation of the user, the terminal marks the audio of the instruction information in the audio to be marked with an instruction tag.

[0060] The instruction tag is a pre-set tag, indicating that the tagged audio is audio of instruction information.

[0061] In response to the marking operation, the terminal marks the audio of the instruction information in the audio to be marked with the instruction tag, thereby obtaining the audio of the manually marked instruction information.

[0062] Again, the audio to be marked with feature labels and instruction labels is recorded as marked audio, and the instruction recognition model is trained according to the marked audio and the instruction text of the marked audio to obtain the target instruction recognition model.

[0063] The terminal records the audio to be marked with feature tags and instruction tags as marked audio, inputs the marked audio into the instruction recognition model to obtain a recognition result, compares the recognition result with the instruction text to obtain the recognition accuracy, and if the accuracy meets the preset accuracy, the target instruction recognition model is obtained; if the accuracy does not meet the preset accuracy, the terminal automatically adjusts or the technician adjusts the instruction recognition model, and inputs the marked audio into the adjusted instruction recognition model, and if the accuracy of the adjusted instruction recognition model meets the preset accuracy, the target instruction recognition model is obtained.

[0064] The target command recognition model is used to remove the residual audio of the wake-up word in the marked audio, and recognize the audio of the command information to obtain the command text corresponding to the command information.

[0065] In one embodiment, if the audio to be marked does not include residual audio of the wake-up word, the target instruction recognition model is used to recognize the audio of the instruction information to obtain the instruction text corresponding to the instruction information.

[0066] The command recognition model of the terminal is updated to a target command recognition model. The terminal adopts the target command recognition model. When the audio received after the terminal switches to a working state includes residual audio of a wake-up word, the terminal can also accurately recognize the audio of the command information.

[0067] The method provided in the embodiment of the present application is that the terminal receives the user's marking operation on the audio of the instruction information and the instruction text corresponding to the instruction information input by the user; in response to the marking operation, the audio of the instruction information in the audio to be marked is marked with an instruction tag; the audio to be marked marked with the feature tag and the instruction tag is recorded as the marked audio, and the instruction recognition model is trained according to the marked audio and the instruction text of the marked audio to obtain a target instruction recognition model. The trained target instruction recognition model can eliminate the residual audio of the wake-up word in the marked audio, and recognize the audio of the instruction information to obtain the instruction text corresponding to the instruction information. The terminal uses the target instruction recognition model, and when the audio received after the terminal switches to the working state includes the residual audio of the wake-up word, it can also accurately recognize the audio of the instruction information.

[0068] In some embodiments, before determining the similarity between the first phoneme of the target audio in the audio to be labeled and the target phoneme, that is, before S120, the method may further include:

[0069] The target audio is determined according to the amplitude variation rule of the head-end audio of the audio to be marked.

[0070] The terminal identifies the amplitude variation rule of the head-end audio of the audio to be marked, and determines the target audio in the head-end audio according to the amplitude variation rule.

[0071] The method provided in the embodiment of the present application determines the target audio according to the amplitude change rule of the head-end audio of the audio to be marked, and obtains the residual audio that may include the wake-up word.

[0072] In some embodiments, determining the target position audio according to the amplitude variation rule of the head-end audio of the audio to be marked may include:

[0073] The audio whose amplitude variation rule at the beginning of the audio to be marked is from high to zero is determined as the target audio.

[0074] The user's speaking habit is generally to say the complete wake-up word and then say the voice command after a small interval. Therefore, the amplitude change pattern of the received audio from the beginning of receiving the residual audio of the wake-up word to the beginning of receiving the audio of the command information is from high to zero. The terminal determines that the audio whose amplitude change pattern at the beginning of the audio to be marked is from high to zero is the target audio.

[0075] In the method provided in the embodiment of the present application, the terminal determines that the audio whose amplitude change pattern at the beginning of the audio to be marked is from high to zero is the target audio, thereby obtaining more accurate residual audio that may include the wake-up word.

[0076] In some embodiments, determining the target audio according to the amplitude variation rule of the head-end audio of the audio to be marked may include:

[0077] The marking model is used to determine the target audio according to the amplitude variation rule of the head-end audio of the audio to be marked.

[0078] The marking model is used to identify the amplitude variation law of the head-end audio of the audio to be marked, and the target audio is determined according to the amplitude variation law.

[0079] In the method provided in the embodiment of the present application, a marking model is provided, which identifies the amplitude variation law of the head-end audio of the audio to be marked, and determines the target audio according to the amplitude variation law, thereby improving the efficiency of identifying the target audio.

[0080] In some embodiments, determining that the audio to be marked has a first-end amplitude variation rule of decreasing from high to zero as the target audio may include:

[0081] The marking model is used to determine that the audio whose amplitude variation pattern at the beginning of the audio to be marked is from high to zero is the target audio.

[0082] The marking model is used to identify the amplitude variation rule of the head-end audio of the audio to be marked. The marking model determines the audio whose amplitude variation rule is from high to zero as the target audio.

[0083] From the time the terminal starts to receive the residual audio of the wake-up word to before receiving the audio of the command information, the amplitude of the received audio changes from high to zero. The marking model determines the audio with an amplitude change pattern from high to zero as the target audio, that is, the marking model determines the residual audio that may include the wake-up word as the target audio.

[0084] In the method provided in the embodiment of the present application, a marking model is provided, which identifies the amplitude change pattern of the head-end audio of the audio to be marked, and determines that the audio with an amplitude change pattern from high to zero is the target audio based on the amplitude change pattern. While improving the efficiency of identifying the target audio, the residual audio that may include the wake-up word is determined as the target audio.

[0085] In some embodiments, before using the marking model to determine that the audio to be marked has a first-end amplitude variation rule of decreasing from high to zero as the target audio, the method may further include:

[0086] First obtain positive samples and negative samples.

[0087] The terminal obtains a large number of positive samples and negative samples for training the model to be trained.

[0088] Both the positive sample and the negative sample are audio to be marked. The positive sample includes the audio whose amplitude changes from high to zero at the beginning, and the negative sample does not include the audio whose amplitude changes from high to zero at the beginning.

[0089] Then, based on the positive samples and negative samples, the model to be trained is trained to obtain a labeled model.

[0090] The terminal inputs the positive samples and negative samples into the model to be trained. The model to be trained identifies the amplitude change pattern of the head-end audio of the positive samples and the negative samples, determines the samples with a change pattern of high to zero as the target audio, and counts the accuracy rate. The model to be trained is adjusted according to the accuracy rate until the accuracy rate meets the preset accuracy rate, stops training, and obtains a labeled model.

[0091] The method provided in the embodiment of the present application obtains a labeling model through training, which can automatically identify the amplitude pattern of the head-end audio in the audio to be labeled, and automatically determine the samples whose amplitude change pattern is from high to zero, and determine them as target audio, thereby improving the efficiency of determining the target audio in the audio to be labeled.

[0092] In some embodiments, the target phoneme includes a phoneme of a preset length at the end of the wake-up word.

[0093] The residue of the wake-up word is the word at the end of the wake-up word, and the target phoneme in the wake-up word includes the phoneme of a preset length at the end of the wake-up word, that is, the target phoneme in the wake-up word includes the residual phoneme of the wake-up word.

[0094] The preset length is set according to the number of phonemes in the wake-up word.

[0095] In an example, the wake-up word is "baby", the residue of the wake-up word is "bei", the phonemes in the wake-up word include "baobei", the preset length is equal to the number of residual phonemes of the wake-up word 3, and the terminal obtains the target phoneme "bei" in the wake-up word.

[0096] In an example, the wake-up word is "baby", the residue of the wake-up word is "bei", the phonemes in the wake-up word include "baobei", the preset length is equal to the number of residual phonemes of the wake-up word 2, and the terminal obtains the target phoneme "ei" in the wake-up word.

[0097] The method provided in the embodiment of the present application uses the phoneme of a preset length at the end of the wake-up word as the target phoneme, which facilitates the identification of the residual audio of the wake-up word in the audio to be marked.

[0098] In some embodiments, the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme is determined, and when the similarity meets a preset condition, the target audio is marked with a feature tag, that is, S120 may include:

[0099] The labeling model is used to determine the similarity between the first phoneme of the target audio in the audio to be labeled and the target phoneme. When the similarity meets the preset conditions, the target audio is labeled with a feature tag.

[0100] A labeling model is used to identify the first phoneme of the target audio in the audio to be labeled, and the similarity between the first phoneme of the target audio in the audio to be labeled and the target phoneme is calculated. When the similarity meets the preset conditions, a feature tag is used to label the target audio.

[0101] In the method provided in the embodiment of the present application, a labeling model is provided, and the labeling model is used to determine the similarity between the first phoneme of the target audio in the audio to be labeled and the target phoneme. When the similarity meets the preset conditions, the target audio is labeled with a feature tag to achieve automatic labeling of the wake-up word residue.

[0102] In some embodiments, before using the marking model to determine the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme, when the similarity meets a preset condition, before using the feature tag to mark the target audio, the method may further include the following steps:

[0103] First obtain positive samples and negative samples.

[0104] The terminal obtains a large number of positive samples and negative samples for training the model to be trained.

[0105] Both positive samples and negative samples are audio. The positive sample includes the residual audio of the wake-up word at the beginning, and the residual audio of the wake-up word has been marked with a feature label. The negative sample does not include the residual audio of the wake-up word at the beginning.

[0106] Then, based on the positive samples and negative samples, the model to be trained is trained to obtain a labeled model.

[0107] The terminal inputs the positive samples and negative samples into the model to be trained, and counts the accuracy of the labeling, adjusts the model to be trained according to the accuracy, and stops training when the accuracy meets the preset accuracy to obtain the labeled model.

[0108] The method provided in the embodiment of the present application obtains a marking model through training, which can automatically identify and mark the residual audio of the wake-up word in the audio to be marked, thereby improving the efficiency of marking the residual audio of the wake-up word in the audio to be marked.

[0109] The present application also provides a marking device, such as Figure 2 As shown, the device 200 may include an acquisition module 210 and a marking module 220 .

[0110] The acquisition module 210 is used to acquire the target phonemes in the audio to be marked and the wake-up word.

[0111] The marking module 220 is used to determine the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme, and when the similarity meets a preset condition, mark the target audio with a feature tag.

[0112] The target audio is the head-end audio of the audio to be labeled, and the feature label indicates that the target audio is the residual audio of the wake-up word.

[0113] The device provided in the embodiment of the present application first obtains the target phoneme in the audio to be marked and the wake-up word; then, determines the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme, and when the similarity meets the preset conditions, marks the target audio with a feature tag, where the target audio is the head-end audio of the audio to be marked, and the feature tag characterizes the target audio as the residual audio of the wake-up word. The audio in the head-end audio of the audio to be marked that meets the preset conditions of similarity with the residual phoneme of the wake-up word is automatically marked with a feature tag as the residual audio of the wake-up word, thereby improving the efficiency of marking the residual audio of the wake-up word in the audio.

[0114] In some embodiments, the audio to be marked includes audio of instruction information, and the apparatus 200 may further include a receiving module 230 , a responding module 240 , and a training module 250 .

[0115] The receiving module 230 is used to receive the user's marking operation on the audio of the instruction information and the instruction text corresponding to the instruction information input by the user.

[0116] The response module 240 is used to respond to the marking operation and mark the audio of the instruction information in the audio to be marked with an instruction tag.

[0117] The training module 250 records the audio to be marked with the feature label and the instruction label as the marked audio, trains the instruction recognition model according to the marked audio and the instruction text of the marked audio, and obtains the target instruction recognition model.

[0118] The target command recognition model is used to remove the residual audio of the wake-up word in the marked audio, and recognize the audio of the command information to obtain the command text corresponding to the command information.

[0119] In one embodiment, if the audio to be marked does not include residual audio of the wake-up word, the target instruction recognition model recognizes the audio of the instruction information to obtain the instruction text corresponding to the instruction information.

[0120] The device provided in the embodiment of the present application receives a user's marking operation on the audio of the instruction information and the instruction text corresponding to the instruction information input by the user; in response to the marking operation, the audio of the instruction information in the audio to be marked is marked with an instruction tag; the audio to be marked marked with the feature tag and the instruction tag is recorded as the marked audio, and the instruction recognition model is trained according to the marked audio and the instruction text of the marked audio to obtain a target instruction recognition model. The trained target instruction recognition model can eliminate the residual audio of the wake-up word in the marked audio, and recognize the audio of the instruction information to obtain the instruction text corresponding to the instruction information. The terminal uses the target instruction recognition model, and when the audio received after the terminal switches to the working state includes the residual audio of the wake-up word, it can also accurately recognize the audio of the instruction information.

[0121] In some embodiments, the apparatus 200 may further include a determination module 260 .

[0122] The determination module 260 is used to determine the target audio according to the amplitude variation rule of the head-end audio of the audio to be marked.

[0123] The device provided in the embodiment of the present application determines the target audio according to the amplitude change rule of the head-end audio of the audio to be marked, and obtains the residual audio that may include the wake-up word.

[0124] In some embodiments, the determination module 260 may be specifically configured to:

[0125] The audio whose amplitude variation rule at the beginning of the audio to be marked is from high to zero is determined as the target audio.

[0126] The device provided in the embodiment of the present application determines that the audio to be marked has a first-end amplitude change pattern that decreases from high to zero as the target audio, thereby obtaining more accurate residual audio that may include the wake-up word.

[0127] In some embodiments, the target phoneme includes a phoneme of a preset length at the end of the wake-up word.

[0128] The device provided in the embodiment of the present application uses the phoneme of a preset length at the end of the wake-up word as the target phoneme, which facilitates the identification of the residual audio of the wake-up word in the audio to be marked.

[0129] In some embodiments, the marking module 120 may be specifically used to:

[0130] The labeling model is used to determine the similarity between the first phoneme of the target audio in the audio to be labeled and the target phoneme. When the similarity meets the preset conditions, the target audio is labeled with a feature tag.

[0131] The device provided in the embodiment of the present application provides a labeling model, and uses the labeling model to determine the similarity between the first phoneme of the target audio in the audio to be labeled and the target phoneme. When the similarity meets the preset conditions, the target audio is labeled with a feature tag to achieve automatic labeling of the wake-up word residue.

[0132] In some embodiments, the acquisition module 210 may also be used to acquire positive samples and negative samples.

[0133] Both positive samples and negative samples are audio to be labeled. The positive sample includes the residual audio of the wake-up word at the beginning, while the negative sample does not include the residual audio of the wake-up word at the beginning.

[0134] The training module 270 can also be used to train the model to be trained based on the positive samples and the negative samples to obtain a labeled model.

[0135] The device provided in the embodiment of the present application obtains a marking model through training, and can automatically identify and mark the residual audio of the wake-up word in the audio to be marked, thereby improving the efficiency of marking the residual audio of the wake-up word in the audio to be marked.

[0136] The marking device provided in the embodiment of the present application performs Figure 1 The various steps in the method shown can achieve the technical effect of improving the efficiency of marking the residual audio of the wake-up word in the audio. For the sake of brevity, they will not be described in detail here.

[0137] Figure 3 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application is shown.

[0138] The electronic device may include a processor 301 and a memory 302 storing computer program instructions.

[0139] Specifically, the processor 301 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0140] The memory 302 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 302 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, the memory 302 may include a removable or non-removable (or fixed) medium. In appropriate cases, the memory 302 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 302 is a non-volatile solid-state memory. In a specific embodiment, the memory 302 includes a read-only memory (ROM). In appropriate cases, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM) or a flash memory or a combination of two or more of these.

[0141] The processor 301 reads and executes the computer program instructions stored in the memory 302 to implement Figure 1 Any one of the marking methods in the embodiments shown.

[0142] In one example, the electronic device may further include a communication interface 303 and a bus 310. Figure 3 As shown, the processor 301, the memory 302, and the communication interface 303 are connected via a bus 310 and communicate with each other.

[0143] The communication interface 303 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiments of the present application.

[0144] Bus 310 includes hardware, software or both, and the parts of electronic equipment are coupled to each other. For example, but not limitation, bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front side bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 310 may include one or more buses. Although the present application embodiment describes and shows a specific bus, the application considers any suitable bus or interconnection.

[0145] The electronic device can execute the marking method in the embodiment of the present application, thereby realizing the combination Figure 1 Describe the marking method.

[0146] In addition, in combination with the marking method in the above embodiment, the embodiment of the present application can provide a computer-readable storage medium for implementation. The computer-readable storage medium stores computer program instructions; when the computer program instructions are executed by a processor, any one of the marking methods in the above embodiment is implemented.

[0147] In combination with the marking method in the above embodiment, the present application embodiment can provide a computer program product to implement. When the instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes any one of the marking methods in the above embodiment.

[0148] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.

[0149] The functional blocks shown in the above-described block diagram can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present application are programs or code segments that are used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link by a data signal carried in a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, etc. The code segment can be downloaded via a computer network such as the Internet, an intranet, etc.

[0150] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.

[0151] The above is only a specific implementation of the present application. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. It should be understood that the protection scope of the present application is not limited to this. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in this application, and these modifications or replacements should be included in the protection scope of this application.

Claims

1. A marking method, characterized in that: The method comprises: Obtaining target phonemes in the audio to be marked and the wake-up word; the audio to be marked is the audio received after the terminal switches to the working state; the target phonemes in the wake-up word include residual phonemes of the wake-up word; Determine the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme, and when the similarity meets a preset condition, mark the target audio with a feature tag, where the target audio is the head-end audio of the audio to be marked, and the feature tag indicates that the target audio is the residual audio of the wake-up word; The audio to be marked includes audio of instruction information, and the method further includes: receiving a marking operation of the user on the audio of the instruction information and an instruction text corresponding to the instruction information input by the user; In response to the marking operation, marking the audio of the instruction information in the audio to be marked with an instruction tag; The audio to be marked with the feature label and the instruction label is recorded as marked audio, and an instruction recognition model is trained according to the marked audio and the instruction text corresponding to the marked audio to obtain a target instruction recognition model. The target instruction recognition model is used to eliminate the residual audio of the wake-up word in the marked audio, and recognize the audio of the instruction information to obtain the instruction text corresponding to the instruction information.

2. The method according to claim 1, characterized in that Before determining the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme, the method further includes: The target audio is determined according to the amplitude variation rule of the head-end audio of the audio to be marked.

3. The method according to claim 2, characterized in that The step of determining the target audio according to the amplitude variation rule of the audio at the head end of the audio to be marked includes: The audio whose amplitude variation rule at the beginning of the audio to be marked decreases from high to zero is determined as the target audio.

4. The method according to claim 1, characterized in that: The target phoneme includes a phoneme with a preset length at the end of the wake-up word.

5. The method according to claim 1, characterized in that The determining the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme, and marking the target audio with a feature tag when the similarity meets a preset condition, comprises: The marking model is used to determine the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme, and when the similarity meets a preset condition, the target audio is marked with a feature tag.

6. The method according to claim 5, characterized in that In the step of adopting the marking model, determining the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme, and when the similarity meets a preset condition, before adopting the feature tag to mark the target audio, the method further includes: Acquire a positive sample and a negative sample, wherein the positive sample head end includes the residual audio of the wake-up word, and the negative sample head end does not include the residual audio of the wake-up word; The model to be trained is trained according to the positive samples and the negative samples to obtain the labeled model.

7. A marking device, characterized in that: The device comprises: An acquisition module, used to acquire target phonemes in the audio to be marked and the wake-up word; the audio to be marked is the audio received after the terminal switches to the working state; the target phonemes in the wake-up word include the residual phonemes of the wake-up word; the audio to be marked includes the audio of the instruction information; A marking module, used to determine the similarity between the first phoneme of the target audio in the audio to be marked and the target phoneme, and when the similarity meets the preset condition, mark the target audio with a feature tag, wherein the target audio is the head-end audio of the audio to be marked, and the feature tag represents that the target audio is the residual audio of the wake-up word, the target audio is the head-end audio of the audio to be marked, and the feature tag represents that the target audio is the residual audio of the wake-up word; A receiving module, used for receiving a marking operation of the user on the audio of the instruction information and an instruction text corresponding to the instruction information input by the user; A response module, configured to respond to the marking operation and mark the audio of the instruction information in the audio to be marked with an instruction tag; A training module is used to record the audio to be marked with the feature label and the instruction label as marked audio, and train an instruction recognition model based on the marked audio and the instruction text corresponding to the marked audio to obtain a target instruction recognition model. The target instruction recognition model is used to eliminate the residual audio of the wake-up word in the marked audio, and recognize the audio of the instruction information to obtain the instruction text corresponding to the instruction information.

8. An electronic device, characterized in that: The device comprises: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the marking method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed by a processor, the marking method according to any one of claims 1 to 6 is implemented.

10. A computer program product, characterized in that When the instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes the marking method as described in any one of claims 1 to 6.

Images