Voice interaction method and related device

Abstract

The invention provides a voice interaction method and a related device, and relates to the technical field of voice recognition. The voice interaction method comprises the following steps: receiving a first voice input by a user, and converting the first voice into a first text; according to a historical interaction text corresponding to historical voice interaction and the first text, determining whether the first voice has an error correction intention; under the condition that the first voice has an error correction intention, determining a real intention of a user according to the error correction intention; and responding to the real intention, and executing a corresponding processing operation. According to the technical scheme provided by the invention, the problem of response errors caused by the fact that the real intention of the user cannot be recognized in the voice interaction process in the prior art can be solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of speech recognition, in particular to a speech interaction method and related device. BACKGROUND

[0002] In the speech interaction process, the speech recognition system may be affected by the user's pronunciation, the recognition accuracy of the speech recognition model, the application environment and other factors, resulting in some errors in the speech recognition result, thereby causing the system to respond incorrectly, such as outputting incorrect reply information or performing incorrect control operations. In addition, sometimes the user changes the intention in the middle and the system does not recognize it, which also causes the system to respond incorrectly. SUMMARY

[0003] Based on the defects and deficiencies of the prior art, the present application provides a speech interaction method and related device, which can solve the problem of incorrect response due to failure to recognize the user's true intention in the existing speech interaction process.

[0004] According to a first aspect of an embodiment of the present application, a speech interaction method is provided, the method comprising: receiving a first speech input by a user and converting the first speech into a first text; determining whether the first speech has a correction intention according to a historical interaction text corresponding to a historical speech interaction and the first text; in a case where the first speech has a correction intention, determining a true intention of the user according to the correction intention; responding to the true intention and performing a corresponding processing operation.

[0005] In the present application, the text corresponding to the user's current speech and the text corresponding to the historical interaction speech can be combined to determine whether the user's current speech has a correction intention. If there is a correction intention, the true intention of the user can be determined based on the correction intention, so as to perform a processing operation corresponding to the true intention of the user, such as outputting reply information corresponding to the true intention of the user, performing a control operation corresponding to the true intention of the user, etc., thereby improving the accuracy of the system response.

[0006] In some optional embodiments, the determining of the true intention of the user according to the correction intention comprises: generating a target text for representing the true intention of the user according to the correction intention and the historical interaction text; the responding to the true intention and performing a corresponding processing operation comprises: responding to the target text and performing a corresponding processing operation.

[0007] In the present application, the user's correction intention and historical interaction text can be combined with the context modeling capability of the voice interaction model to understand the user's correction intention, and based on the correction intention, context, and context, the user's real intention can be determined, and the target text representing the user's real intention can be generated to obtain information that can be processed by the system, so that the system can respond to the user's real intention and perform corresponding processing operations.

[0008] In some optional embodiments, the determining whether the first voice has a correction intention according to the historical interaction text corresponding to the historical voice interaction and the first text comprises: According to the historical interaction text corresponding to the historical voice interaction of the preset round and the first text, it is determined whether the first voice has a correction intention.

[0009] Considering that user correction is generally based on the principle of proximity, that is, the most recent voice or text is corrected, therefore, the present application can set a suitable fixed round based on this, so that the object to be corrected can be included as much as possible, and the interference of too much historical information can be reduced.

[0010] In some optional embodiments, when the preset round is greater than or equal to 1, the responding to the real intention and performing corresponding processing operations comprises: For the current voice interaction round or the historical voice interaction round including the user's original intention, the real intention is responded to and corresponding processing operations are performed; wherein the original intention is the intention before correction.

[0011] In the present application, for the current voice interaction round, the user's real intention is responded to and corresponding processing operations are performed, which can make the user feel that the system has recognized the user's correction intention and responded, thereby improving the effectiveness of the interaction. For the historical voice interaction round including the user's original intention, the user's real intention is responded to and corresponding processing operations are performed, which can reduce the occupation of the system output error information on the information display interface in the scenario where the system needs to output text form reply information, reduce the interference of invalid information to the user when viewing effective historical information, and achieve the purpose of making the display interface more concise.

[0012] In some optional embodiments, the method further comprises: saving the corresponding original text in each round of voice interaction.

[0013] In the present application, the corresponding original text in each round of voice interaction can be saved to construct user-level historical interaction text. The saved historical interaction text can be used to determine whether the user's current voice has a correction intention.

[0014] In some optional embodiments, the method further comprises: input sample data into a target model, and identify a real intention of a user corresponding to the sample data through the target model; wherein the target model is a language model used to implement voice interaction; and the sample data comprises multi-round interactive texts; adjust the target model according to the real intention of the user predicted by the target model and a pre-set real intention of the user.

[0015] In this application, the voice interaction model can be trained so that the voice interaction model has the functions of identifying a correction intention and determining a real intention based on the correction intention.

[0016] According to a second aspect of the embodiments of the present application, a voice interaction device is provided, and the device comprises: a voice recognition module, configured to receive a first voice input by a user and convert the first voice into a first text; a first intention determination module, configured to determine whether the first voice has a correction intention according to a historical interactive text corresponding to historical voice interaction and the first text; a second intention determination module, configured to determine a real intention of the user according to the correction intention in a case where the first voice has the correction intention; a processing module, configured to perform a corresponding processing operation in response to the real intention.

[0017] In some optional embodiments, the second intention determination module comprises: a generation unit, configured to generate a target text used to represent the real intention of the user according to the correction intention and the historical interactive text.

[0018] The processing module is specifically configured to perform a corresponding processing operation in response to the target text.

[0019] In some optional embodiments, the first intention determination module comprises: an intention determination unit, configured to determine whether the first voice has a correction intention according to a historical interactive text corresponding to a preset round of historical voice interaction and the first text.

[0020] In some optional embodiments, the processing module 304 is specifically configured to: perform a corresponding processing operation in response to the real intention for a current voice interaction round or a historical voice interaction round including an original intention of the user; wherein the original intention is an intention before correction.

[0021] In some optional embodiments, the device further comprises: a saving module, configured to save a corresponding original text in each round of voice interaction.

[0022] In some optional embodiments, the apparatus further includes: a training module configured to input sample data into a target model, and identify a real intention of a user corresponding to the sample data through the target model.

[0023] The target model is a language model used to implement voice interaction, and the sample data includes interaction text corresponding to multi-round voice interaction.

[0024] an adjusting module configured to adjust the target model according to the real intention of the user predicted by the target model and a pre-set real intention of the user.

[0025] According to a third aspect of the embodiments of the present application, an electronic device is provided, including a memory and a processor; The memory is connected with the processor, and is configured to store a program; The processor is configured to realize the voice interaction method according to the first aspect by running the program in the memory.

[0026] According to a fourth aspect of the embodiments of the present application, a storage medium is provided, and the storage medium stores a computer program.

[0027] According to a fifth aspect of the embodiments of the present application, a computer program product or a computer program is provided. BRIEF DESCRIPTION OF DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of the provided drawings.

[0029] FIG. 1 A flowchart of a voice interaction method provided by the embodiments of the present application.

[0030] FIG. 2 An example flowchart of a voice interaction method provided by the embodiments of the present application.

[0031] FIG. 3 A block diagram of a voice interaction apparatus provided by the embodiments of the present application.

[0032] FIG. 4A structural schematic diagram of an electronic device is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0033] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work are within the scope of protection of the present application.

[0034] SUMMARY With the rapid development of artificial intelligence technology, voice interaction technology has been widely used in intelligent customer service, smart home, vehicle-mounted system and mobile terminal assistant due to its convenience and naturalness. As the core link of the voice interaction model, the task of the speech recognition technology is to convert the received user voice signal into corresponding text information, and its accuracy directly determines the correctness of the subsequent system reply and user satisfaction.

[0035] However, the speech recognition system may be affected by factors such as user pronunciation, recognition accuracy of the speech recognition model, application environment, etc., resulting in some errors in the speech recognition result, thereby causing system response errors. For example, when a user wants to turn on a fan and speaks in a heavy local accent, the speech recognition system may incorrectly convert the voice into the text "turn on gray shirt", and the system cannot respond correctly based on the "turn on gray shirt" text, i.e., cannot turn on the fan, and may output a reply such as "I'm sorry, I couldn't find a device called 'gray shirt'", or directly ignore the instruction, resulting in operation execution failure and affecting the user's interaction experience. For another example, in a smart home scenario, the user instruction "turn the air conditioner in the bedroom to dehumidification mode" may be incorrectly recognized as "turn the air conditioner in the bedroom to chef mode" due to the relatively low frequency of the word "dehumidification" in the training data or confusion with environmental noise, thereby causing the system to perform incorrect operations or give confusing replies.

[0036] In addition, sometimes the user changes the intention halfway but the system does not recognize it, which also causes system response errors. For example, the user first says "navigate to Donghu Village" and then says "forget it, go to Nangang instead". The system will respond to the "navigate to Donghu Village" instruction, but the "forget it, go to Nangang instead" may be ignored due to the lack of explicit navigation keywords, and cannot affect the subsequent voice instruction.

[0037] In summary, the system response errors caused by the poor recognition of speech recognition errors and dynamic changes in intention in the prior art are not conducive to improving the voice interaction experience, and therefore, there is an urgent need in the field for a technical solution that can effectively capture the real intention of the user.

[0038] To this end, the embodiments of the present application provide a voice interaction method technology that can combine the text corresponding to the current voice of the user and the text corresponding to the historical interactive voice to determine whether the current voice of the user has a correction intention, and if so, determine the real intention of the user based on the correction intention, so as to perform a processing operation corresponding to the real intention of the user, such as outputting reply information corresponding to the real intention of the user, performing a control operation corresponding to the real intention of the user, etc., thereby improving the accuracy of the system response.

[0039] Exemplary methods The embodiments of the present application also provide a voice interaction method applied to an electronic device, wherein a voice interaction model is arranged in the electronic device. The electronic device can be a terminal device such as a mobile phone, a computer, a vehicle-mounted computer, etc., or a server such as a cloud server.

[0040] The method will be described in detail through some embodiments below. The following embodiments can be combined with each other, and the same or similar concepts or processes can not be described in detail in some embodiments.

[0041] As shown in FIG. 1 , the voice interaction method can include steps 101 to 104, as described below.

[0042] Step 101: receiving a first voice input by a user and converting the first voice into a first text.

[0043] After receiving the voice of the user, the voice interaction model can convert the voice of the user into corresponding text information through voice recognition technology. For example, as shown in FIG. 2 , the voice interaction model includes a voice recognition module, which can recognize the first voice input by the user through automatic speech recognition (ASR) technology after receiving the voice of the user, to obtain the first text t1.

[0044] Step 102: determining whether the first voice has a correction intention according to the historical interactive text corresponding to the historical voice interaction and the first text.

[0045] In the embodiments of the present application, the original text corresponding to each round of voice interaction can be saved to construct a user-level historical interactive text. For example, FIG. 2As shown, the first text t1 converted from the first voice can be saved to the historical interaction text database. The saved historical interaction text can be used to determine whether the current voice of the user has a correction intention.

[0046] It should be noted that in each round of voice interaction in the embodiments of the present application, the user and the system can both have voice output, or only the user can have voice output, and the system can only output reply information in the form of text or only perform corresponding control operations (such as a control operation of turning on an air conditioner). Therefore, the interaction text corresponding to each round of voice interaction can include text converted from voice output by the user and the system, or can only include text converted from voice output by the user, or can include text converted from voice output by the user and text directly output by the system, and is not limited to text converted from voice.

[0047] In each case of starting a new round of voice interaction, that is, in each case of receiving a newly input voice of the user (corresponding to the first voice in step 101), the newly input voice of the user can be converted into text, and then whether the newly input voice has a correction intention can be determined according to the historical interaction text and the text converted from the newly input voice. For example, FIG. 2 As shown, the first text t1 converted from the first voice can be saved to the historical interaction text database. The saved historical interaction text can be used to determine whether the current voice of the user has a correction intention.

[0048] For example, the second text converted from the second voice (as a historical voice) input by the user before by the voice recognition system is "navigate to Donghu Village", and is displayed through the display device of the electronic device. The user finds that "Donghu Village" is not the destination "Donghu Village" that the user wants to go to, and then continues to output the first voice "hu is the hu of an erhu". The system can then determine whether the first voice has a correction intention by collecting the historical interaction text "navigate to Donghu Village" and the first text "hu is the hu of an erhu" corresponding to the first voice.

[0049] For example, the second text converted from the second voice (as a historical voice) input by the user before by the voice recognition system is "navigate to Donghu Village", and is displayed through the display device of the electronic device. Then the user suddenly changes his mind and outputs a new first voice "forget it, go to Nangang instead". The system can then determine whether the first voice has a correction intention by collecting the historical interaction text "navigate to Donghu Village" and the first text "forget it, go to Nangang instead" corresponding to the first voice.

[0050] For another example, the voice recognition system converts the second text obtained based on the second voice input by the user previously as "introduce the ancient road", and displays it through the display device of the electronic device. The user finds that the "ancient road" recognized by the system is not the "Guo rice" that he wants to understand, and continues to output the first voice "gu is the Guo of Shangguo and dao is the rice of rice field", and the system can collect the historical interaction text "introduce the ancient road" and the first text "gu is the Guo of Shangguo and dao is the rice of rice field" corresponding to the first voice to determine whether the first voice has a correction intention.

[0051] In a case where it is determined that the first voice does not have the correction intention, the voice interaction model can then respond to the first text and perform a corresponding processing operation. For example, as shown in FIG. 13A, in a case where the first voice does not have the correction intention, the first text is sent to a processing module in the voice interaction model, and the processing module responds to the first text. FIG. 2

[0052] For example, in a vehicle use scenario, the second voice is "open the sunroof", and the first voice is "open the radio". Obviously, the first voice does not have a correction intention for the previous voice or text, and the first text is input to the processing module in the voice interaction model to respond to the first voice. For example, an operation of opening the radio can be performed based on the first text corresponding to the first voice, and a reply content "the radio has been opened for you" is generated and output.

[0053] In a case where it is determined that the first voice has the correction intention, the following step 103 is performed.

[0054] Step 103: In a case where the first voice has the correction intention, the real intention of the user is determined according to the correction intention.

[0055] In the embodiments of the present application, in a case where the first voice has the correction intention, the real intention of the user can be determined according to the recognized correction intention.

[0056] For example, the second text obtained by converting the second voice is "navigate to Donghucun", and the first text corresponding to the first voice is "hu is the hu of erhu". The first text in combination with the second text can determine that the first voice has a correction intention for the second text, and based on the correction intention, the real intention of the user can be determined as wanting to navigate to Donghucun.

[0057] For another example, the second text obtained by converting the second voice is "navigate to Donghucun", and the first text corresponding to the first voice is "I give up, let's go to Nangang". The first text in combination with the second text can determine that the first voice has a correction intention for the second voice, and based on the correction intention, the real intention of the user can be determined as wanting to navigate to Nangang.

[0058] ​For example, the second text obtained from the second speech conversion is "Introduce the ancient road", and the first text corresponding to the first speech is "gu is the valley of Shanggu, dao is the rice of Daotian". The combination of the first text and the second text can determine that the first speech has the intention to correct errors in the second text. Based on this error correction intention, it can also be determined that the user's true intention is to obtain information about rice.

[0059] As can be seen from the above, the error correction intent in the embodiments of this application can be to correct the speech recognition results or to correct the user's previous instructions.

[0060] Step 104: Respond to the true intent and execute the corresponding processing operation.

[0061] In this embodiment of the application, after obtaining the user's true intent, the system can respond to that intent and execute corresponding processing operations. For example... FIG. 2 As shown, the user's true intent can be input into the processing module of the voice interaction model, and the processing module will respond to the true intent and perform the corresponding processing operation.

[0062] The execution of the corresponding processing operation described herein may include, but is not limited to, at least one of the following: outputting corresponding response information, executing corresponding control operations, etc. The output response information may be in the form of voice and / or text.

[0063] For example, for the first example in step 103, the processing operation corresponding to the user's true intention can be: control the navigation function to navigate to Donghu Village as the destination, and output the reply message: "Navigation has been restarted for you, the destination is Donghu Village".

[0064] For example, in the second example in step 103, the processing operation corresponding to the user's true intention can be: control the navigation function to navigate to Nangang as the destination, and output the reply message: "Navigation has been restarted for you, destination is Nangang".

[0065] For example, in the second example in step 103, the processing operation corresponding to the user's true intent could be: generating and outputting introductory information related to rice.

[0066] In this embodiment, the system can combine the text corresponding to the user's current voice and the text corresponding to the historical interactive voice to determine whether the user's current voice has an error correction intention. If there is an error correction intention, the system can determine the user's true intention based on the error correction intention, and then perform processing operations corresponding to the user's true intention, such as outputting reply information corresponding to the user's true intention, or performing control operations corresponding to the user's true intention, thereby improving the accuracy of the system's response.

[0067] In some optional embodiments, step 103, "determining the user's true intent based on the error correction intent," may include: generating target text representing the user's true intent based on the error correction intent and historical interaction text. Correspondingly, step 104, "responding to the true intent and performing the corresponding processing operation," may include: responding to the target text and performing the corresponding processing operation.

[0068] In this embodiment, the system can understand the user's error-correction intention based on the user's error-correction intent and historical interaction text, combined with the context modeling capability of the voice interaction model. Furthermore, based on the error-correction intent, context, and linguistic context, the system can determine the user's true intent, thereby generating target text representing the user's true intent, such as... FIG. 2 The target text t2 is used to obtain information that the system can process, so that the system can respond to the user's true intentions and perform corresponding processing operations.

[0069] For example, the second text obtained from the second speech conversion is "Navigate to Donghu Village", and the first text corresponding to the first speech is "hu is the hu in erhu". The first text combined with the second text can determine that the first speech has the intention to correct errors in the second text. Based on this error correction intention, it can also be determined that the user's true intention is to want to navigate to Donghu Village. Then, the target text used to represent the user's true intention can be generated: "Navigate to Donghu Village".

[0070] For example, the second text obtained from the second speech conversion is "Navigate to Donghu Village", and the first text corresponding to the first speech is "Forget it, let's go to Nangang". The first text combined with the second text can determine that the first speech has the intention to correct the second speech. Based on this correction intention, it can also be determined that the user's true intention is to navigate to Nangang. Then, the target text used to represent the user's true intention can be generated: "Navigate to Nangang".

[0071] For example, the second text obtained from the second speech conversion is "Introduce the ancient road", and the first text corresponding to the first speech is "gu is the valley of Shanggu, dao is the rice of Daotian". The first text combined with the second text can determine that the first speech has the intention to correct errors in the second text. Based on this error correction intention, it can also be determined that the user's true intention is to obtain information about rice. Then, the target text used to represent the user's true intention can be generated: "Introduce rice".

[0072] Optionally, in order to enable the voice interaction model to recognize error correction intent and determine the true intent based on the error correction intent, the voice interaction model can be trained. The specific training method is as follows.

[0073] Step A1: Input the sample data into the target model and use the target model to identify the user's true intent corresponding to the sample data.

[0074] The target model described here is a language model used to implement voice interaction, also known as a voice interaction model. The sample data described here includes voice interaction text, which includes error correction intent.

[0075] In this embodiment of the application, voice interaction text including error correction intent can be collected or compiled as sample data, and the user's true intent corresponding to each sample data can be labeled.

[0076] The voice interaction text mentioned here can be multi-turn voice interaction text, such as "Navigate to Donghu Village" or "Never mind, let's go to Nangang instead"; or it can be single-turn voice interaction text, such as "Navigate to Donghu Village, never mind, let's go to Nangang instead".

[0077] Step A2: Adjust the target model based on the user's true intent predicted by the target model and the pre-set user true intent.

[0078] The embodiments of this application can determine the gap between the predicted value and the true value of the target model based on the user's true intent predicted by the target model and the user's true intent set in advance, thereby adjusting the parameters of the target model to achieve the purpose of optimizing the model.

[0079] Optionally, a loss value can be calculated based on a preset loss function, the user's true intent predicted by the target model, and a pre-set user's true intent, thereby determining the gap between the target model's predicted value and the true value. The required loss function can be set according to actual needs.

[0080] Optionally, the target model can be optimized through supervised fine-tuning training. Supervised fine-tuning training (SFT) refers to the process of using labeled, high-quality supervised data to perform additional, targeted training on a pre-trained large model (base model) to finely adjust the model to adapt it to a specific task or follow specific instructions. In short, supervised fine-tuning training is a key technology for transforming powerful, general-purpose pre-trained models into AI assistants that can reliably, safely, and effectively serve specific scenarios. It involves using high-quality data to finely adjust the model.

[0081] In some optional embodiments, step 102: determining whether the first voice has an error correction intent based on the historical interaction text corresponding to the historical voice interaction and the first text may include: Based on the historical interactive text corresponding to the historical interactive voice in the preset round and the first text, determine whether the first voice has the intention to correct errors.

[0082] In this embodiment, the historical interactive text corresponding to a fixed number of historical interactive voices can be used, and combined with the first text corresponding to the current voice, to determine whether the first voice has an error correction intention.

[0083] The preset number of rounds mentioned here can be a value greater than or equal to 1, and can be set according to actual needs, such as 3 times, 4 times, etc.

[0084] Considering that users generally follow the principle of proximity in error correction, that is, they correct the most recent voice or text, the embodiments of this application can set an appropriate fixed number of rounds based on this principle. This can maximize the inclusion of the object to be corrected and reduce the interference of too much historical information.

[0085] In some alternative embodiments, step 104: responding to the true intent and performing the corresponding processing operation may include: For the current voice interaction round or historical voice interaction rounds including the user's original intent, respond to the true intent and execute the corresponding processing operation.

[0086] Here, the original intent refers to the intent before error correction. It can be the user's original intent or the user's intent initially determined by the system based on the user's voice. For example, the original intent could be "navigate to Donghu Village," and the corrected intent could be "navigate to Donghu Village"; another example is that the original intent could be "navigate to Donghu Village," and the corrected intent could be "navigate to Nangang"; yet another example is that the original intent could be "introduce the ancient road," and the corrected intent could be "introduce rice."

[0087] In this embodiment, after determining the user's true intent based on the user's error-correction intention, the response operation can take effect for the current voice interaction round when responding to the true intent. For example, based on the first round of voice interaction, if the user's intent is determined to be "introduce the ancient road," in the second round of voice interaction, the user corrects the error by inputting the voice "gu is the valley of Shanggu, dao is the rice of Daotian." If the response is in accordance with the user's true intent, it can take effect in the second round of voice interaction, that is, respond to the corrected voice and output the introduction information of Gu and Dao, so as to prominently display the latest information response to the user.

[0088] In this embodiment, in response to the user's true intent during the current voice interaction round, the corresponding processing operation is performed, which allows the user to clearly feel that the system has recognized their error correction intent and responded accordingly, thus improving the effectiveness of the interaction.

[0089] Optionally, after determining the user's true intention based on the user's error correction intention, when responding to the true intention, the response operation can take effect in the historical voice interaction rounds including the user's original intention, that is, change the response operation in the historical interaction. Especially when the system needs to output a reply message in text form, the response operation taking effect in the historical voice interaction rounds including the user's original intention can be manifested as replacing the incorrect reply message with a new reply message. For example, based on the first round of voice interaction, it is determined that the user's intention is to "introduce the ancient road". In the second round of voice interaction, the user corrects the error by inputting the voice "gu is the gu of Shanggu, and dao is the dao of paddy field". When responding to the user's true intention, it can take effect in the first round of voice interaction. At this time, the introduction information of gu and dao can be used to replace the introduction information of the ancient road, which can reduce the occupation of the information display interface by invalid information (that is, the previous incorrect information), reduce the interference of invalid information on the user's viewing of valid historical information, and at the same time achieve the purpose of making the display interface more concise.

[0090] In summary, the technical solution provided by the embodiment of the present application can combine the text corresponding to the user's current voice and the text corresponding to the historical interaction voice to determine whether there is an error correction intention in the user's current voice. If there is an error correction intention, the true intention of the user can be determined based on the error correction intention, so as to execute the processing operation corresponding to the user's true intention, thereby improving the accuracy of the system response.

[0091] Exemplary apparatus Correspondingly, the embodiment of the present application also provides a voice interaction device, which is applied to an electronic device, and a voice interaction model is set in the electronic device. The electronic device can be a terminal device, such as a mobile phone, a computer, a vehicle-mounted computer, etc., or a server, such as a cloud server.

[0092] As FIG. 3 shown, the device may include: A voice recognition module 301, configured to receive the first voice input by the user and convert the first voice into a first text.

[0093] A first intention determination module 302, configured to determine whether the first voice has an error correction intention according to the historical interaction text corresponding to the historical voice interaction and the first text.

[0094] A second intention determination module 303, configured to determine the true intention of the user according to the error correction intention when the first voice has an error correction intention.

[0095] A processing module 304, configured to respond to the true intention and execute a corresponding processing operation.

[0096] In the voice interaction device provided in this application embodiment, after the voice recognition module 301 receives the first voice input by the user and converts the first voice into first text, the first intent determination module 302 can combine the first text and historical interaction text to determine whether the first voice has an error correction intent. If it is determined that the first voice has an error correction intent, the second intent determination module 303 can determine the user's true intent based on the error correction intent determined by the first intent determination module 302. Finally, the processing module 304 responds to the true intent and outputs corresponding reply information, thereby improving the accuracy of the reply information.

[0097] In some alternative embodiments, the second intent determination module 303 includes: The generation unit is used to generate target text representing the user's true intent based on the error correction intent and the historical interaction text.

[0098] The processing module 304 is specifically used to: respond to the target text and perform corresponding processing operations.

[0099] In some alternative embodiments, the first intent determination module 302 includes: The intent determination unit is used to determine whether the first voice has an error correction intent based on the historical interaction text corresponding to the historical voice interaction in a preset round and the first text.

[0100] In some alternative embodiments, the processing module 304 is specifically used for: For the current voice interaction round or historical voice interaction rounds including the user's original intent, respond to the true intent and execute the corresponding processing operation; wherein, the original intent is the intent before the error correction.

[0101] In some alternative embodiments, the apparatus further includes: The save module is used to save the original text corresponding to each round of voice interaction.

[0102] In some alternative embodiments, the apparatus further includes: The training module is used to input sample data into the target model and use the target model to identify the user's true intent corresponding to the sample data.

[0103] The target model is a language model used to implement voice interaction; the sample data includes interactive text corresponding to multiple rounds of voice interaction.

[0104] The adjustment module is used to adjust the target model based on the user's true intent predicted by the target model and the user's true intent preset.

[0105] The voice interaction device provided in this embodiment belongs to the same concept as the voice interaction method provided in the above embodiments of this application. It can execute the voice interaction method provided in any of the above embodiments of this application and has the corresponding functional modules and beneficial effects of the execution method. Technical details not described in detail in this embodiment can be found in the specific processing content of the voice interaction method provided in the above embodiments of this application, and will not be repeated here.

[0106] It should be understood that the modules in the above map-using device can be implemented by a processor calling software. For example, the device includes a processor connected to a memory containing instructions. The processor calls the instructions stored in the memory to implement any of the above methods or to implement the functions of each unit of the device. The processor can be a general-purpose processor, such as a CPU or microprocessor, and the memory can be internal or external to the device. Alternatively, the units in the device can be implemented as hardware circuits. By designing the hardware circuits, some or all of the unit functions can be implemented. The hardware circuit can be understood as one or more processors. For example, in one implementation, the hardware circuit is an ASIC, and the functions of some or all of the above units are implemented by designing the logical relationships between the components within the circuit. In another implementation, the hardware circuit can be implemented by a PLD, such as an FPGA, which can include a large number of logic gates. The connection relationships between the logic gates are configured through configuration files to implement the functions of some or all of the above units. All units of the above device can be implemented entirely by a processor calling software, entirely by hardware circuits, or partially by a processor calling software with the remaining parts implemented by hardware circuits.

[0107] In this application embodiment, a processor is a circuit with signal processing capabilities. In one implementation, the processor can be a circuit with instruction reading and execution capabilities, such as a CPU, microprocessor, GPU, or DSP. In another implementation, the processor can implement certain functions through the logical relationships of hardware circuits. These logical relationships are fixed or reconfigurable. For example, the processor may be a hardware circuit implemented as an ASIC or PLD, such as an FPGA. In a reconfigurable hardware circuit, the process of the processor loading a configuration document and configuring the hardware circuit can be understood as the processor loading instructions to implement the functions of some or all of the above units. Furthermore, it can also be a hardware circuit designed for artificial intelligence, which can be understood as an ASIC, such as an NPU, TPU, or DPU.

[0108] As can be seen, each unit in the above device can be one or more processors (or processing circuits) configured to implement the above methods, such as: CPU, GPU, NPU, TPU, DPU, microprocessor, DSP, ASIC, FPGA, or a combination of at least two of these processor forms.

[0109] Furthermore, the units in the above devices can be integrated in whole or in part, or they can be implemented independently. In one implementation, these units are integrated together and implemented in the form of a System-on-Chip (SoC). The SoC may include at least one processor for implementing any of the above methods or implementing the functions of the units in the device. The at least one processor may be of different types, such as CPU and FPGA, CPU and artificial intelligence processor, CPU and GPU, etc.

[0110] Exemplary electronic device This application also provides an electronic device, such as... FIG. 4 As shown, the electronic device includes a memory 400 and a processor 410.

[0111] The memory 400 is connected to the processor 410 and is used to store programs.

[0112] The processor 410 is used to implement the voice interaction method in the above embodiments by running the program stored in the memory 400.

[0113] Specifically, the aforementioned electronic device may also include: a communication interface 420, an input device 430, an output device 440, and a bus 450.

[0114] The processor 410, memory 400, communication interface 420, input device 430, and output device 440 are interconnected via a bus. Among them: Bus 450 may include a pathway for transmitting information between various components of a computer system.

[0115] The processor 410 can be a general-purpose processor, such as a general-purpose central processing unit (CPU), a microprocessor, etc., or an application-specific integrated circuit (ASIC), or one or more integrated circuits used to control the execution of the program of the present invention. It can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0116] Processor 410 may include a main processor, as well as a baseband chip, modem, etc.

[0117] The memory 400 stores a program that executes the technical solution of this invention, and may also store an operating system and other key business functions. Specifically, the program may include program code, which includes computer operation instructions. More specifically, the memory 400 may include read-only memory (ROM), other types of static storage devices capable of storing static information and instructions, random access memory (RAM), other types of dynamic storage devices capable of storing information and instructions, disk storage, flash memory, etc.

[0118] Input device 430 may include a device for receiving user input data and information, such as a keyboard, mouse, camera, scanner, light pen, voice input device, touch screen, pedometer, or gravity sensor.

[0119] Output device 440 may include devices that allow information to be output to a user, such as a display screen, printer, speaker, etc.

[0120] The communication interface 420 may include a device that uses any transceiver to communicate with other devices or communication networks, such as Ethernet, Radio Access Network (RAN), Wireless Local Area Network (WLAN), etc.

[0121] The processor 410 executes the program stored in the memory 400 and calls other devices, which can be used to implement the various steps of the voice interaction method provided in the above embodiments of this application.

[0122] Exemplary computer program product and storage medium In addition to the methods and devices described above, embodiments of this application may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform the steps in the voice interaction method described in the embodiments of this application.

[0123] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium; in another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0124] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0125] Furthermore, embodiments of this application may also be storage media storing a computer program, which is executed by a processor of the steps in the voice interaction method described in the embodiments of this application.

[0126] In addition, embodiments of this application may also be chips, which include processors and data interfaces. The processor reads instructions stored in the memory through the data interface to execute the steps in the voice interaction method described in the embodiments of this application.

[0127] For the foregoing method embodiments, in order to simplify the description, they are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to this application, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0128] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For apparatus embodiments, since they are basically similar to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0129] The steps in the methods of the various embodiments of this application can be adjusted, merged, or deleted in order according to actual needs, and the technical features described in each embodiment can be replaced or combined.

[0130] The modules and sub-modules in the devices and terminals in the various embodiments of this application can be merged, divided, and deleted according to actual needs.

[0131] It should be understood that the disclosed terminals, devices, and methods can be implemented in other ways, given the several embodiments provided in this application. For example, the terminal embodiments described above are merely illustrative. For instance, the division of modules or sub-modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple sub-modules or modules may be combined or integrated into another module, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0132] The modules or submodules described as separate components may or may not be physically separate. The components that constitute a module or submodule may or may not be physical modules or submodules; that is, they may be located in one place or distributed across multiple network modules or submodules. Some or all of the modules or submodules can be selected to achieve the purpose of this embodiment's solution, depending on actual needs.

[0133] Furthermore, the functional modules or sub-modules in the various embodiments of this application can be integrated into one processing module, or each module or sub-module can exist physically separately, or two or more modules or sub-modules can be integrated into one module. The integrated modules or sub-modules described above can be implemented in hardware or in the form of software functional modules or sub-modules.

[0134] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0135] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software unit executed by a processor, or a combination of both. The software unit can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0136] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A voice interaction method, characterized in that, The method includes: Receive first voice input from the user and convert the first voice into first text; Based on the historical interaction text corresponding to the historical voice interaction and the first text, determine whether the first voice has an error correction intention; If the first voice message has an error correction intent, the user's true intent is determined based on the error correction intent. In response to the stated true intent, perform the corresponding processing operation.

2. The voice interaction method according to claim 1, characterized in that, Determining the user's true intent based on the error correction intent includes: Based on the error correction intent and the historical interaction text, generate target text to represent the user's true intent; The response to the true intent, executing the corresponding processing operation, includes: In response to the target text, perform the corresponding processing operation.

3. The voice interaction method according to claim 1, characterized in that, The step of determining whether the first voice has an error correction intent based on the historical interaction text corresponding to the historical voice interaction and the first text includes: Based on the historical interaction text corresponding to the preset rounds of historical voice interaction and the first text, determine whether the first voice has an error correction intention.

4. The voice interaction method according to claim 1 or 3, characterized in that, The response to the true intent, executing the corresponding processing operation, includes: For the current voice interaction round or historical voice interaction rounds including the user's original intent, respond to the true intent and execute the corresponding processing operation; wherein, the original intent is the intent before the error correction.

5. The voice interaction method according to claim 1, characterized in that, The method also includes saving the original text corresponding to each round of voice interaction.

6. The voice interaction method according to claim 1, characterized in that, The method further includes: Sample data is input into a target model, which identifies the user's true intent corresponding to the sample data. The target model is a language model used to implement voice interaction. The sample data includes voice interaction text, which includes error correction intent. The target model is adjusted based on the user's true intent predicted by the target model and the pre-set user true intent.

7. A voice interaction device, characterized in that, The speech recognition module is used to receive the first speech input by the user and convert the first speech into first text; The first intent determination module is used to determine whether the first voice has an error correction intent based on the historical interaction text corresponding to the historical voice interaction and the first text. The second intent determination module is used to determine the user's true intent based on the error correction intent when the first voice has an error correction intent. The processing module is used to respond to the stated true intent and perform corresponding processing operations.

8. An electronic device, characterized in that, include: Memory and processor; The memory is connected to the processor and is used to store programs; The processor is used to implement the voice interaction method as described in any one of claims 1 to 6 by running a program in the memory.

9. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the voice interaction method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product stores a computer program, which, when executed by a processor, implements the voice interaction method as described in any one of claims 1 to 6.

Images