Voice interaction method, device, server and readable storage medium

Abstract

The present invention discloses a voice interaction method, device, server, and readable storage medium. The voice interaction method comprises: receiving a voice request forwarded by a vehicle for adjusting a preset function of the vehicle, wherein the preset function refers to a function for simulating the scale adjustment of the operation of a vehicle component; reading a voice request for adjusting the preset function of the vehicle from a previous round; rewriting the voice request for the current round using the previous round of voice requests; performing precision recognition on the rewritten voice request for the current round; and completing the voice interaction based on the result of the precision recognition. The present invention identifies the scale adjustment precision of the voice request by combining instruction rewriting and precision recognition, thereby achieving accurate recognition of the scale adjustment range under multiple rounds of voice requests.

Description

Technical Field

[0001] The present invention relates to the field of voice technology, and in particular to a voice interaction method and device, a server, and a readable storage medium. Background Art

[0002] Currently, voice interaction can be used in smart car scenarios to enable users to control vehicle hardware devices, such as "open the window" and "turn up the volume." However, scenarios where users want to make continuous adjustments require multiple rounds of interaction. After the previous round of voice interaction, users naturally omit parts of the subsequent rounds of conversation. For example, the following conversation between a user and the voice assistant Xiao P:

[0003] User: What’s the weather like today?

[0004] Xiao P: It’s sunny in Guangzhou today, 26-30°.

[0005] User: What about the weather in Shanghai?

[0006] In a multi-round conversation, as in the first example above, the user literally wants to ask about Shanghai, but actually wants to ask about the weather in Shanghai. Omitting some content is in line with human conversation habits, but this may cause the vehicle's on-board system to be unable to accurately recognize voice requests in certain rounds, or prompt that it cannot understand.

[0007] Furthermore, if the user needs to adjust the volume, they can operate the mechanical knob for adjusting the car volume and rotate it to the desired volume. However, if the user uses voice to adjust the volume, they can only turn it up or down. In the second example below:

[0008] User: Volume up

[0009] Xiao P: The volume is turned up

[0010] User: Big Big

[0011] As can be seen from the second example, the current vehicle's onboard system cannot accurately identify the size of the current wheel, or the prompts are incomprehensible. This situation cannot meet the user's demand for continuous adjustment of the scale as precise as a mechanical knob. Summary of the Invention

[0012] The embodiments of the present invention provide a voice interaction method, a device thereof, a server and a readable storage medium.

[0013] An embodiment of the present invention provides a voice interaction method. The voice interaction method includes: receiving a voice request forwarded by a vehicle for adjusting a current round of preset vehicle functions, wherein the preset function is a function that simulates the scale adjustment of vehicle component operations; reading a previous round of voice request for adjusting the preset vehicle functions; rewriting the current round of voice request using the previous round of voice request; performing precision recognition on the rewritten current round of voice request; and completing the voice interaction based on the precision recognition result.

[0014] In this way, the voice interaction method of the present invention can, after receiving a user's voice request for a preset vehicle function, read the previous voice request and use it to rewrite the current voice request so that the rewritten current voice request can be recognized by the vehicle's onboard system as corresponding to the scale adjustment precision. Based on the precision recognition result, the vehicle's components can be adjusted using voice interaction. By combining command rewriting and precision recognition to identify the scale adjustment precision of the voice request, accurate recognition of the scale adjustment range is achieved under multiple voice requests.

[0015] The voice interaction method includes: performing reverse mining on two adjacent rounds of voice requests whose occurrence frequency is greater than a preset frequency to build a regularization engine.

[0016] In this way, the present invention realizes the construction of a regular engine for high-frequency voice requests by reversely mining two adjacent rounds of voice requests whose occurrence frequency is greater than a preset frequency.

[0017] After reading the previous round of voice requests for adjusting the vehicle's preset functions, the voice interaction method includes: using a regularization engine to identify the scale adjustment accuracy corresponding to the current round of voice requests and the previous round of voice requests; when the recognition result of the regularization engine is that the corresponding scale adjustment accuracy is identified, completing the voice interaction according to the identified scale adjustment accuracy.

[0018] In this way, the present invention combines the current round of voice requests with the previous round of voice requests to perform regular engine recognition, and can directly determine the corresponding scale adjustment accuracy for high-frequency rules, so that voice interaction can be completed according to the determined scale adjustment accuracy.

[0019] The rewriting of the current round of voice requests by using the previous round of voice requests includes: when the recognition result of the regular engine is that the corresponding scale adjustment accuracy cannot be recognized, rewriting the current round of voice requests by using the previous round of voice requests.

[0020] In this way, when the regular engine of the present invention cannot identify the scale adjustment accuracy of the current round of voice requests based on the previous round of voice requests and the current round of voice requests, the previous round of voice requests can be used to rewrite the current round of voice requests, so that the rewritten current round of voice requests can be recognized by the vehicle's on-board system with the corresponding scale adjustment accuracy.

[0021] The rewriting of the current round of voice requests using the previous round of voice requests includes: obtaining a rewriting model by training rewriting training data, wherein the rewriting training data includes two adjacent rounds of voice requests; and rewriting the current round of voice requests using the previous round of voice requests and the rewriting model.

[0022] In this way, the present invention obtains a rewriting model through machine learning training of two adjacent rounds of voice requests, so that the voice request of the current round can be rewritten according to the voice request of the previous round and the rewriting model, so that the rewritten voice request can be recognized by the vehicle's on-board system with the corresponding scale adjustment accuracy.

[0023] The precision recognition of the rewritten voice request of the current round includes: obtaining a precision recognition model through training with precision training data, wherein the precision training data is related to vehicle parts that can be scale-adjusted, the scale adjustment range of the vehicle parts, and the scale adjustment precision range of the parts; and using the precision recognition model to precision recognize the rewritten voice request of the current round.

[0024] In this way, the present invention uses machine learning to train a precision recognition model using training data corresponding to vehicle parts that can be scale-adjusted, the scale adjustment range of the vehicle parts, and the scale adjustment accuracy range of the parts, and then performs precision recognition on the rewritten voice request to achieve accurate recognition of the scale adjustment accuracy corresponding to the user's voice request.

[0025] The voice interaction is completed according to the result of the precision identification, including: obtaining the accuracy discrimination probability of each preset scale adjustment accuracy corresponding to the result of the precision identification; determining one of the preset scale adjustment accuracy whose accuracy discrimination probability is greater than a probability threshold as the target scale adjustment accuracy corresponding to the voice request to complete the voice interaction.

[0026] In this way, the voice interaction method of the present invention can obtain the accuracy judgment probability of each preset scale adjustment accuracy corresponding to the accuracy recognition result, and determine a preset scale adjustment accuracy with an accuracy judgment probability greater than the probability threshold as the target scale adjustment accuracy corresponding to the voice request, thereby identifying the scale adjustment accuracy of the vehicle parts precisely adjusted by the user.

[0027] The voice interaction method includes: determining that the accuracy recognition error of the voice request of the current round is in error when the accuracy discrimination probability of each of the preset scale adjustment accuracies is not greater than a probability threshold.

[0028] In this way, when the accuracy judgment probability of each preset scale adjustment accuracy is not greater than the probability threshold, it is determined that the voice request accuracy recognition error is made, and voice requests not related to the scale adjustment accuracy can be excluded.

[0029] The present invention also provides a voice interaction device. The voice interaction device includes: a command receiving module, a command reading module, a rewriting module, a precision identification module, and an interaction module. The command receiving module is used to receive a voice request forwarded by a vehicle for adjusting a preset function of the vehicle, wherein the preset function is a function that simulates the scale adjustment of the operation of a vehicle component; the command reading module is used to read the previous voice request for adjusting the preset function of the vehicle; the rewriting module is used to rewrite the current voice request using the previous voice request; the precision identification module is used to perform precision identification on the rewritten current voice request; and the interaction module is used to complete the voice interaction based on the result of the precision identification.

[0030] In this way, after receiving a user's voice request for a preset vehicle function, the voice interaction device of the present invention can read the previous voice request and use it to rewrite the current voice request so that the rewritten voice request can be recognized by the vehicle's onboard system as corresponding to the scale adjustment precision. Based on the precision recognition result, the vehicle component can be adjusted through voice interaction. By combining command rewriting and precision recognition to identify the scale adjustment precision of the voice request, accurate recognition of the scale adjustment range can be achieved under multiple voice requests.

[0031] The present invention provides a server comprising a processor and a memory, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the voice interaction method described in any one of the above embodiments is implemented.

[0032] In this manner, the server of the present invention, through a processor executing a computer program, can, upon receiving a user's voice request for a preset vehicle function, read the previous voice request and use it to rewrite the current voice request so that the rewritten voice request can be recognized by the vehicle's onboard system as corresponding to the scale adjustment precision. Based on the precision recognition result, the scale adjustment of the vehicle components can then be implemented through voice interaction. By combining command rewriting and precision recognition to identify the scale adjustment precision of the voice request, accurate recognition of the user's scale adjustment range can be achieved under multiple voice requests.

[0033] The present invention also provides a non-volatile computer-readable storage medium containing a computer program. When the computer program is executed by one or more processors, the voice interaction method described in any one of the above embodiments is implemented.

[0034] Thus, when a processor executes the computer program stored in a readable storage medium of the present invention, upon receiving a user's voice request for a preset vehicle function, the computer program can read the previous voice request and use it to rewrite the current voice request so that the rewritten voice request can be recognized by the vehicle's onboard system as corresponding to the scale adjustment precision. The system can then implement scale adjustment of vehicle components through voice interaction based on the precision recognition result. By combining command rewriting and precision recognition to identify the scale adjustment precision of the voice request, accurate recognition of the scale adjustment range can be achieved under multiple voice requests.

[0035] Additional aspects and advantages of the embodiments of the present invention will be given in part in the following description and in part will be obvious from the following description or learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments with reference to the following drawings, in which:

[0037] Figure 1 It is a flowchart of the voice interaction method of the present invention;

[0038] Figure 2 It is a structural diagram of the voice interaction device of the present invention;

[0039] Figure 3 It is a flowchart of the voice interaction method of the present invention;

[0040] Figure 4 It is a structural diagram of the voice interaction device of the present invention;

[0041] Figure 5 It is a flowchart of the voice interaction method of the present invention;

[0042] Figure 6 It is a flowchart of the voice interaction method of the present invention;

[0043] Figure 7 is a schematic structural diagram of a mapping module in the voice interaction device of the present invention;

[0044] Figure 8 It is a flowchart of the voice interaction method of the present invention;

[0045] Figure 9It is a structural diagram of the voice interaction device of the present invention;

[0046] Figure 10 It is a flowchart of the voice interaction method of the present invention;

[0047] Figure 11 It is a flowchart of the voice interaction method of the present invention;

[0048] Figure 12 It is a structural diagram of the voice interaction device of the present invention;

[0049] Figure 13 It is a flowchart of the voice interaction method of the present invention;

[0050] Figure 14 It is a structural diagram of the voice interaction device of the present invention;

[0051] Figure 15 It is a flowchart of the voice interaction method of the present invention;

[0052] Figure 16 It is a structural diagram of the interaction module in the voice interaction device of the present invention;

[0053] Figure 17 It is a structural diagram of the server of the present invention;

[0054] Figure 18 It is a schematic structural diagram of the computer-readable storage medium of the present invention. DETAILED DESCRIPTION

[0055] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the embodiments of the present invention, and should not be understood as limiting the embodiments of the present invention.

[0056] At present, when a user makes multiple rounds of voice requests, for example, the first round of voice request is "brighten the screen", and the second round of voice request uses a simplified voice request "brighten bright bright", the vehicle's voice interaction system cannot accurately identify the user's second round of voice request as requiring the screen brightness to be increased by 3 scales, and cannot correctly issue vehicle-side instructions, causing the screen brightness to be accurately increased by the three brightness levels required by the user, resulting in a poor user experience.

[0057] To solve the above problem, please refer to Figure 1 The present invention provides a voice interaction method. The voice interaction method includes:

[0058] 01, receiving a voice request forwarded by the vehicle for adjusting a preset function of the vehicle, wherein the preset function refers to a function of simulating the scale adjustment of the operation of a vehicle component;

[0059] 02, read the last round of voice request for adjusting the vehicle's preset functions;

[0060] 03. Use the previous round of voice requests to rewrite the current round of voice requests;

[0061] 04. Accurately recognize the rewritten voice request of the current round;

[0062] 05. Complete voice interaction based on the results of precision recognition.

[0063] See also Figure 2 The present invention further provides a voice interaction device 10. The voice interaction device 10 includes: a command receiving module 11, a command reading module 12, a rewriting module 13, a precision identification module 14 and an interaction module 15.

[0064] Step 01 can be implemented by the instruction receiving module 11, step 02 can be implemented by the instruction reading module 12, step 03 can be implemented by the rewriting module 13, step 04 can be implemented by the precision identification module 14, and step 05 can be implemented by the interaction module 15. That is, the instruction receiving module 11 is used to receive a voice request forwarded by the vehicle for adjusting a preset function of the vehicle, where the preset function refers to a function that simulates the scale adjustment of the operation of a vehicle component; the instruction reading module 12 is used to read the previous voice request for adjusting the preset function of the vehicle; the rewriting module 13 is used to rewrite the current voice request using the previous voice request; the precision identification module 14 is used to perform precision identification on the rewritten current voice request; and the interaction module 15 is used to complete the voice interaction based on the precision identification result.

[0065] The user uses voice interaction to simulate the process of adjusting the scale of vehicle components. The corresponding voice requests may include but are not limited to "screen bright bright bright", "volume up high", and "seat back, back, back". Among them, the preset function refers to the function of completing the scale adjustment through the vehicle components. The vehicle components may refer to physical components such as mechanical knobs or buttons, which are components that can adjust the scale. Currently, in smart cars, for scenarios where users want to make continuous adjustments, they are reflected in multiple rounds of interaction in the voice scenario. For example, the user's previous voice request is "volume up high", and after the system turns up the volume, the user issues the current voice request "lower the volume a little bit". At this time, the system will not be able to accurately identify the specific scale as reducing the volume by 2 scales for the second round of voice request. It will only prompt the user that it cannot recognize the current command or simply adjusts the wrong hardware by mistake, etc., which cannot meet the user's demand for continuous adjustment of the scale as precise as a mechanical knob.

[0066] The voice interaction method of the present invention can, after receiving a user's voice request for a preset vehicle function, read the previous round of voice requests and use them to rewrite the current round of voice requests so that the rewritten voice requests can be recognized by the vehicle's onboard system as corresponding scale adjustment precision. Control instructions are then issued based on the precision recognition results to control the corresponding vehicle components, completing the voice interaction. By combining instruction rewriting and precision recognition to identify the scale adjustment precision of the voice request, it is possible to accurately identify the scale adjustment precision for simulating the operation of vehicle components through voice interaction under multiple rounds of voice requests.

[0067] It should be noted that after receiving the user's voice request for the current wheel of the vehicle's preset function, the received voice request for the current wheel is voice recognized to obtain the current wheel voice recognition text for subsequent processing. For example, the user's voice request for the current wheel with the vehicle's preset function adjustment "the screen is bright, bright, bright" is voice recognized, and the obtained current wheel recognition text is "the screen is bright, bright, bright".

[0068] In actual situations, the text instructions recognized by ASR may not be clear and accurate due to vehicle hardware limitations, network instability, or the user's colloquial or dialectal expression. In this case, the voice request received in the current round can be preprocessed. The preprocessing includes correcting some common text errors, such as correcting "volume deep deep deep" to "volume increase increase increase increase increase", and removing some meaningless words, such as "ah" and "please".

[0069] Please combine Figure 3 Before step 01, the voice interaction method includes:

[0070] 011. Determine the control range and non-control range of vehicle parts.

[0071] Please combine Figure 4 , the voice interaction device 10 also includes a first determination module 111.

[0072] Step 011 can be implemented by the first determining module 111. That is, the first determining module 111 can be used to determine the control range and the non-control range of the vehicle component.

[0073] Understandably, not all vehicle functions can, are capable of, or require precise, graduated adjustment. For example, seat movement can be adjusted in all directions using vehicle components. However, vehicle doors lack components like knobs or buttons for graduated adjustment and are typically opened and closed solely via the door handle. Therefore, seat adjustment falls within the controllable range of vehicle components, while door adjustment falls within the uncontrollable range of vehicle components.

[0074] The information of vehicle parts is obtained, and based on the information of the vehicle parts, the hardware that can be calibrated by the parts is determined as the control range of the vehicle parts, and the hardware that cannot be adjusted by the vehicle parts is determined as the non-control range.

[0075] First, identify the vehicle components that can be adjusted, such as the volume knob, screen brightness button, air conditioning volume knob / button, and seat adjustment knob / button. Furthermore, determine the control range of vehicle components, which may include the vehicle audio system, the screen, the air conditioning, the seats, the interior ambient lighting, the exterior lights, or the windows. The non-controllable range of vehicle components may include the doors, rearview mirrors, and the trunk.

[0076] During the subsequent voice interaction process, voice prompts can be given when the voice request is for a non-controllable range of vehicle parts.

[0077] In this way, by collecting vehicle component information, the functions that can be adjusted through the components are confirmed, thereby determining the control range of the vehicle components, that is, the control range that can be adjusted through voice interaction.

[0078] Voice interaction methods include:

[0079] 012. Determine the adjustable range of vehicle components.

[0080] The voice interaction device 10 further includes a second determining module 112 .

[0081] Step 012 may be implemented by the second determining module 112. That is, the second determining module 112 may be used to determine the adjustable range of the vehicle component.

[0082] It can be understood that after determining the control range and non-control range of vehicle components, it is necessary to determine the adjustable range for each vehicle component in the control range. The adjustable range of a vehicle component corresponds to the scale range that can be adjusted by operating the vehicle component. Corresponding to different vehicle components, the adjustable range can be a gear or a range. For example, if the screen brightness button is pressed 5 times in a row, the screen brightness is adjusted from 1 to 5 gears to the maximum brightness, then the adjustable range of the screen brightness button is 1 to 5 gears. For another example, the total scale value of the knob for adjusting the seat forward and backward is 90, then the adjustable range of the seat adjustment knob is a scale value of 1 to 90.

[0083] See also Figure 5 , step 012 includes:

[0084] 0121, determine the adjustable range of simplified terms corresponding to vehicle parts.

[0085] Step 0121 can be implemented by the second determining module 112. That is, the second determining module 112 can be used to determine the adjustable range of the vehicle parts corresponding to the simplified words.

[0086] Specifically, simplified words refer to simplified words used by users that accurately represent the degree of adjustment. For example, reduplicated words can be used as simplified words. In this way, when inputting voice commands, users only need to enter the simplified command. For example, adjusting the brightness of the car display screen can be simplified as "screen bright bright", "screen bright bright bright", "screen dim", and "screen dim dim dim"..., adjusting the volume of the car audio can be simplified as "volume up", "volume up high", "volume down low", and "volume down low"..., and adjusting the air volume of the air conditioner can be simplified as "air volume up", "air volume up high", "air volume down low", and "air volume down low"... Of course, simplified words can also be repeated words that users are accustomed to using, such as "brighter", "dimmer", "bigger", and "smaller". Accordingly, the user's voice request can be simplified as "screen brighter", "screen dimmer", "volume up", and "volume down". There is no specific limitation here.

[0087] The adjustable range for simplified words can be determined based on the adjustable range of vehicle components. For example, when adjusting the vehicle screen, the screen brightness can be adjusted from 1 to 5 levels. During voice recognition, each brightness-related voice request can recognize up to 5 simplified words, so the adjustable range for simplified words can be 1 to 5. If a voice request includes multiple simplified words, each simplified word can adjust the screen brightness by one level.

[0088] For example, when adjusting the car audio system, the volume can be adjusted using the simplified words "large," "larger," "smaller," or "small." The total volume adjustment range is 30 scales. However, voice recognition can recognize up to 10 simplified words for volume-related voice requests. In this case, the adjustable range of simplified words is 1 to 10, and each simplified word can adjust the car audio volume by 3 scales. If voice recognition recognizes a voice request with more than 10 simplified words, the volume can be adjusted directly to maximum or minimum.

[0089] Voice interaction methods include:

[0090] 013, map the control range and adjustable range to the corresponding preset scale adjustment accuracy.

[0091] Step 013 may be implemented by the mapping module 113. That is, the mapping module 113 may be used to map the control range and the adjustable range to corresponding preset scale adjustment precisions.

[0092] For example, if the volume is adjusted by 3 notches each time during voice interaction simulation of vehicle component operation, with a total volume of 30, the preset scale adjustment precision range can be 1 to 10. For another example, if the seat is adjusted forward and backward by 18 notches each time during voice interaction simulation of vehicle component operation, with a total volume of 90, the preset scale adjustment precision range can be 1 to 5.

[0093] See also Figure 6 , step 013 includes:

[0094] 0131, set the simplified words as slots, and extract the preset recognition texts corresponding to vehicle parts to obtain repeated fields;

[0095] 0132, repeat the slot values ​​of the repeated fields to get the number of repetitions;

[0096] 0133, maps the number of repetitions to the preset scale adjustment accuracy based on the adjustable range of the simplified word.

[0097] Please combine Figure 7 The mapping module 113 includes an extraction unit 1131 , a statistics unit 1132 and a mapping unit 1133 .

[0098] Step 0131 can be implemented by extraction unit 1131, step 0132 can be implemented by counting unit 1132, and step 0133 can be implemented by mapping unit 1133. Specifically, extraction unit 1131 can be used to set simplified words as slots and perform slot extraction on the preset recognition text corresponding to the vehicle parts to obtain repeated fields; counting unit 1132 can be used to repeatedly count the slot values ​​of the repeated fields to obtain the number of repetitions; and mapping unit 1133 can be used to map the number of repetitions to a preset scale adjustment precision based on the adjustable range of the simplified words.

[0099] It can be understood that the number of repetitions of the simplified word can represent the number of times the vehicle parts are scaled. Therefore, the simplified word can be set as a slot. For example, the adjustable range of the simplified word of the volume knob is 1 to 10, and the preset scale adjustment precision corresponding to the volume knob is 1 to 10. If the preset recognition text corresponding to the voice request is "volume greatly greatly", then "great greatly greatly" can be extracted as a slot, and the slot can be set as a repeated field. Then, the slot value of the extracted repeated field is repeatedly counted, and its repetition number is mapped to the corresponding preset scale adjustment precision. For the extracted slot "great greatly greatly", the repetition number of "large" is 4, which can be mapped to the corresponding preset scale adjustment precision of 4.

[0100] In other embodiments of the present invention, with the user's permission, different user instructions corresponding to the same scale adjustment precision can be collected. For example, regarding the statement "volume very loud", the user can expand it with different degrees of freedom, such as "volume increase increase increase", "volume increase increase increase", and "volume high high high". The scale adjustment precision obtained by recognizing different expansion words is "volume adjustment 3 times".

[0101] Please combine Figure 8 , the voice interaction method includes:

[0102] 014. Perform reverse mining on two adjacent rounds of voice requests whose occurrence frequency is greater than the preset frequency to build a regularization engine.

[0103] Please combine Figure 9 , the voice interaction device 10 includes a regularization engine module 114 .

[0104] Step 014 can be implemented by the regularization engine module 114. That is, the regularization engine module 114 can be used to perform reverse mining on two adjacent rounds of voice requests whose occurrence frequency is greater than a preset frequency to build a regularization engine.

[0105] In this way, the present invention realizes the construction of a regular engine for high-frequency voice requests by reverse mining two adjacent rounds of voice requests with a frequency greater than a preset frequency.

[0106] First, with the user's permission, the server can collect the user's historical voice information over a period of time. The collected voice requests must include at least two rounds of voice requests. It is estimated that more than 10,000 historical voice messages will need to be collected.

[0107] Secondly, the server can perform a simple screening of the collected historical voice information to filter out voice information with obvious unclear semantics and some shorter voice information containing only modal particles, such as "ah" and "oh", leaving voice information with clear semantics and specific purposes, such as "navigate to Peking University", "help me turn on the air conditioner", "search for nearby hospitals", "play Jay Chou's songs", "how is the weather today", etc.; and remove voice requests with only one round in the screening.

[0108] The server can then perform high-frequency set statistics on the filtered voice requests. By using reverse mining to extract extractable templates, the server can collect high-frequency rules and build a regularization engine for two consecutive rounds of voice requests with a frequency greater than a preset frequency. The frequency of occurrence can refer to the number of times the corresponding two consecutive rounds of voice requests appear in the collected voice information. When the number of occurrences exceeds a certain number, the usage frequency of the corresponding two consecutive rounds of voice requests can be considered to be greater than the preset frequency.

[0109] For example, if the first round of voice requests is "volume up", and the second round of voice requests is "lower", they can be mapped into a rule "(volume)xxx[lower]*" through reverse mining, where the square brackets represent "lower can be matched once or multiple times", thereby establishing a regular engine through various high-frequency rules.

[0110] Voice interaction methods include:

[0111] 015, using the regular expression engine to identify the scale adjustment accuracy corresponding to the current round of voice requests and the previous round of voice requests;

[0112] 016. When the recognition result of the regular engine is that the corresponding scale adjustment accuracy is recognized, the voice interaction is completed according to the recognized scale adjustment accuracy.

[0113] Step 015 can be implemented by the regularization engine module 114, and step 016 can be implemented by the interaction module 15. That is, the regularization engine module 114 can be configured to use the regularization engine to identify the scale adjustment precision corresponding to the current round of voice requests and the previous round of voice requests. The interaction module 15 can be configured to complete the voice interaction based on the identified scale adjustment precision when the regularization engine identifies the corresponding scale adjustment precision.

[0114] In this way, the present invention establishes a regular engine, which can identify the current round of voice requests and the previous round of voice requests to determine whether they belong to the high-frequency set instructions. When the current round of voice requests and the previous round of voice requests belong to the high-frequency set instructions, the regular engine can identify the corresponding scale adjustment accuracy, thereby completing the voice interaction according to the identified scale adjustment accuracy.

[0115] For example, if the first round of voice requests is "volume up a little, up a little", and the second round of voice requests is "volume down a little", then it can be mapped to the rule "(volume)xxx[down a little]*2". If this rule corresponds to the high-frequency set instruction, the regular engine can recognize that "down a little" is repeated twice, so the corresponding scale adjustment range is "volume adjustment 2 times". By simulating the operation of vehicle parts through voice interaction, accurate identification of scale adjustment accuracy is achieved under multiple rounds of voice requests.

[0116] See also Figure 10 , step 03 includes:

[0117] 031. When the recognition result of the regular expression engine is that the corresponding scale adjustment accuracy cannot be recognized, the voice request of the previous round is used to rewrite the voice request of the current round.

[0118] Step 031 can be implemented by the rewriting module 13. That is, the rewriting module 13 can be used to rewrite the current round of voice requests using the previous round of voice requests when the recognition result of the regularization engine is that the corresponding scale adjustment accuracy cannot be recognized.

[0119] It can be understood that when the recognition result of the regular engine is that the corresponding scale adjustment accuracy cannot be recognized, the current round of voice requests and the previous round of voice requests are non-high-frequency set instructions or belong to compound instructions. It is not easy to identify the scale adjustment accuracy of the current round of voice requests through the regular engine. At this time, the current round of voice requests can be rewritten so that the rewritten voice requests can be recognized by the vehicle's on-board system with the corresponding scale adjustment accuracy.

[0120] Please combine Figure 11 , step 03 includes:

[0121] 032, obtaining a rewriting model by training rewriting training data, where the rewriting training data includes two adjacent rounds of voice requests;

[0122] 033, rewrite the current round of voice requests using the previous round of voice requests and the rewriting model.

[0123] Please combine Figure 12 , the voice interaction device 10 includes a rewriting training module 115 .

[0124] Step 032 can be implemented by the rewriting training module 115, and step 033 can be implemented by the rewriting module 13. In other words, the rewriting training module 115 can be used to train a rewriting model using rewriting training data. The rewriting module 13 can be used to rewrite the current round of voice requests using the previous round of voice requests and the rewriting model.

[0125] In this way, the present invention uses machine learning to train a rewriting model based on two consecutive rounds of voice requests. This allows the current round of voice requests to be rewritten based on the previous round of voice requests and the rewriting model, allowing the rewritten voice requests to be recognized by the vehicle's onboard system with the corresponding scale adjustment accuracy. The rewriting model can be trained using BERT (Bidirectional Encoder Representation from Transformers) and sequence labeling to obtain a trained rewriting model.

[0126] The rewritten data can be obtained by annotating two adjacent rounds of voice requests from the aforementioned screened voice requests. Specifically, the second round of voice requests can be manually rewritten and annotated. For example, if the first round of voice requests is "Volume up, volume up," and the second round of voice requests is "Volume down," the second round of voice requests can be rewritten and annotated as "Volume down." In this way, the annotated two adjacent rounds of voice requests are fed into the established rewriting model. During the training process, the rewriting model can learn how to rewrite the second round of voice requests into the annotated second round of voice requests based on the two adjacent rounds of voice requests through feature extraction.

[0127] During the training process, two adjacent rounds of voice requests in the annotated voice information are divided into a rewriting training set and a rewriting verification set. The division ratio can be set according to needs and is not limited here. For example, the rewriting training set is 80% and the rewriting verification set is 20%. For the established rewriting model, at least part of the data in the rewriting training set is first used to train the rewriting model, and then the accuracy of the trained rewriting model is rewritten and verified using at least part of the data in the rewriting verification set. If the accuracy of the rewriting verification does not reach the rewriting accuracy threshold, the rewriting model is trained again using at least another part of the data in the rewriting training set, and the accuracy of the re-trained rewriting model is rewritten and verified using at least another part of the data in the rewriting verification set. The training and rewriting verification process is repeated until the accuracy of the rewriting verification reaches the rewriting accuracy threshold. It can be considered that the rewriting model has met the standard and the training of the rewriting model is completed.

[0128] It should be noted that each data in the rewriting training set and the rewriting verification set is used only once. If the rewriting model fails to meet the training standards after traversing all the data in the rewriting training set and the rewriting verification set, more voice information can be collected again with the user's permission, so as to screen and annotate more rewriting training data to train the rewriting model, thereby ensuring that the rewriting model can accurately rewrite the voice request.

[0129] Please combine Figure 13 , step 04 includes:

[0130] 041. Obtaining a precision recognition model by training with precision training data, the precision training data being related to vehicle parts that can be scale-adjusted by vehicle parts, the scale adjustment range of the vehicle parts, and the scale adjustment precision range of the parts;

[0131] 042. Use the precision recognition model to accurately recognize the rewritten voice request of the current round.

[0132] Please combine Figure 14 , the voice interaction device 10 includes an accuracy training module 116 .

[0133] Step 041 can be implemented by the precision training module 116, and step 042 can be implemented by the precision recognition module 14. In other words, the precision training module 116 can be used to train a precision recognition model using precision training data. The precision recognition module 14 can be used to perform precision recognition on the rewritten voice request of the current round using the precision recognition model.

[0134] In this way, the present invention uses machine learning to train a precision recognition model using training data corresponding to adjustable vehicle parts, their adjustment ranges, and their adjustment precision ranges. This model then performs precision recognition on the rewritten voice request for the current round, accurately identifying the user's adjustment precision. Model training can utilize models such as BERT, ALBERT, XLNet, and RoBERTa.

[0135] The precision training data is related to vehicle parts that can be adjusted through scale and the scale adjustment range of the parts. This means that the precision training data includes all vehicle parts that can be adjusted through scale, such as the "volume knob," "screen brightness button," "air conditioning air volume knob / button," and "seat adjustment knob / button." The adjustable range of a vehicle part corresponds to the scale range that can be adjusted by operating the vehicle part. For different vehicle parts, the scale adjustment range can be a gear or a range, and the scale adjustment precision range can be the scale value for each adjustment.

[0136] The precision training data can be obtained by annotating the voice requests in the above-mentioned filtered voice information. Specifically, the voice requests in two adjacent rounds of voice requests can be manually annotated. It can be understood that the voice request should include content related to the scale adjustment accuracy that the user needs to adjust. For example, the voice request is "volume up a little bit", and the user needs to adjust the volume up 2 times. At this time, the scale adjustment accuracy corresponding to the voice request can be manually annotated as "volume adjustment 2 times". In this way, the annotated voice request is given to the established precision recognition model. During the training process, the precision recognition model can learn how to identify the target scale adjustment accuracy that the user wants to achieve through the input voice request through feature extraction.

[0137] During the training process, the annotated voice requests can be divided into a precision training set and a precision verification set. The division ratio can be set according to needs and is not limited here. For example, the precision training set is 80% and the precision verification set is 20%. For the established precision recognition model, at least part of the data in the precision training set is first used to train the precision recognition model, and then the accuracy of the trained precision recognition model is verified using at least part of the data in the precision verification set. If the accuracy of the precision verification does not reach the precision accuracy threshold, the precision recognition model is trained again using at least another part of the data in the precision training set, and the accuracy of the re-trained precision recognition model is verified using another part of the data in the precision verification set. The training and precision verification process is repeated until the accuracy of the precision verification reaches the precision accuracy threshold. It can be considered that the precision recognition model has met the standard and the training of the precision recognition model is completed.

[0138] It should be noted that each data in the precision training set and the precision verification set is used only once. If the precision recognition model fails to meet the training standards after traversing all the data in the precision training set and the precision verification set, more voice information can be collected again with the user's permission, so as to filter and annotate more precision training data to train the precision recognition model, thereby ensuring that the precision recognition model can accurately recognize the scale adjustment accuracy corresponding to the input voice request.

[0139] It is understood that the training of the above-mentioned rewriting model and precision recognition model can be performed offline. After the offline-trained rewriting module 13 and precision recognition model are deployed to a server or vehicle, the server or vehicle can, upon receiving the current round of voice requests, rewrite the current round of voice requests using the rewriting model of the previous round of voice requests, and perform precision recognition on the rewritten current round of voice requests using the precision recognition model. In particular, if the current round of voice requests is a simplified voice request input by the user after two rounds of voice requests, the rewritten previous round of voice requests can be obtained when the previous round of voice requests is obtained. For example, when the current round of voice request "Lower" is the third round of voice request, if the first round of voice request is "Volume up, up, up" and the second round of voice request is "Lower", and the second round of voice request is in the process of completing the previous voice interaction, the second round of voice request can be rewritten to "Lower" through the rewriting model. In this way, for the current round of voice request received, the previous round of voice request read in step 02 can be the rewritten second round of voice request "Lower", so that the current round of voice request can be rewritten according to the second round of voice request and the rewriting model.

[0140] Please combine Figure 15 , step 05 includes:

[0141] 051, obtaining the accuracy identification result corresponding to the accuracy of each preset scale adjustment accuracy;

[0142] 052, determining a preset scale adjustment accuracy whose accuracy judgment probability is greater than the probability threshold as the target scale adjustment accuracy corresponding to the current round of voice request to complete the voice interaction.

[0143] See also Figure 16 The interaction module 15 includes an acquisition unit 151 and an accuracy determination unit 152 .

[0144] Step 051 can be implemented by acquisition unit 151, and step 052 can be implemented by accuracy determination unit 152. That is, acquisition unit 151 can be used to obtain the accuracy discrimination probability corresponding to each preset scale adjustment precision based on the accuracy identification result. Accuracy determination unit 152 can be used to determine a preset scale adjustment precision with an accuracy discrimination probability greater than a probability threshold as the target scale adjustment precision corresponding to the voice request to complete the voice interaction.

[0145] Specifically, based on the identification results for each category of vehicle parts corresponding to multiple preset scale adjustment accuracies, the accuracy identification module 14 can generate an accuracy discrimination probability that each preset scale adjustment accuracy matches, thereby obtaining multiple accuracy discrimination probabilities. If the probability threshold is 0.9, and the accuracy identification result indicates that the accuracy discrimination probability of the preset scale adjustment accuracy for a certain category of vehicle parts exceeds 0.9, the server will determine that the preset scale adjustment accuracy for that category of vehicle parts is the target scale adjustment accuracy requested by the current user. The probability threshold can also be other values, and can be a default value or customized according to user needs, without limitation here.

[0146] In this way, the voice interaction method of the present invention can obtain the accuracy judgment probability of each preset scale adjustment accuracy corresponding to the result of accuracy recognition, and determine a preset scale adjustment accuracy with an accuracy judgment probability greater than the probability threshold as the target scale adjustment accuracy corresponding to the voice request, thereby realizing the recognition of the user's need to accurately adjust the scale adjustment accuracy of vehicle parts.

[0147] Voice interaction methods include:

[0148] 053. When the accuracy judgment probability of each preset scale adjustment accuracy is not greater than the probability threshold, it is determined that the current round of voice request accuracy recognition error.

[0149] Step 053 may be implemented by the accuracy determination unit 152. That is, the accuracy determination unit 152 may be configured to determine that the voice request accuracy recognition error occurs when the accuracy determination probability of each preset scale adjustment accuracy is not greater than the probability threshold.

[0150] For example, when the accuracy discrimination probabilities corresponding to the preset scale adjustment accuracies of each category are not greater than the probability threshold, that is, the probability that the user's accuracy recognition results obtained based on the voice request match the preset scale adjustment accuracy images of each category is relatively low, and are all lower than the probability threshold. For example, the probability threshold may be 0.9, then it is determined that the accuracy recognition of the voice request is incorrect. For example, the voice request input by the user is "car door open open open", because the car door is not adjusted through scale-adjusted vehicle parts. Therefore, the accuracy recognition of the voice request "car door open open open" is incorrect.

[0151] In this way, when the accuracy judgment probability of each preset scale adjustment accuracy is not greater than the probability threshold, it is determined that the voice request accuracy recognition error is made, and voice requests not related to the scale adjustment accuracy can be excluded.

[0152] See also Figure 17The present invention further provides a server 20. The server 20 includes a processor 21 and a memory 22. The memory 22 stores a computer program 221. When the computer program 221 is executed by the processor 21, the voice interaction method in any of the above embodiments is implemented.

[0153] The server 20 of the present invention, through the processor 21 executing 221, can, upon receiving a user's voice request for a preset vehicle function, read the previous round of voice requests and use the previous round of voice requests to rewrite the current round of voice requests so that the rewritten voice requests can be recognized by the system as corresponding to the scale adjustment precision. Then, based on the precision recognition result, control instructions are issued to control the corresponding vehicle components to complete the voice interaction. By combining instruction rewriting and precision recognition to identify the scale adjustment precision of the voice request, it is possible to accurately identify the scale adjustment precision for simulating the operation of vehicle components through voice interaction under multiple rounds of voice requests.

[0154] See also Figure 18 The present invention further provides a non-volatile computer-readable storage medium 30 containing a computer program 31. When the computer program 31 is executed by one or more processors 40, the voice interaction method of any of the above-mentioned implementation regulations is implemented.

[0155] For example, when the computer program 31 is executed by the processor 40, the steps of the following data processing method are implemented:

[0156] 01, receiving a voice request forwarded by the vehicle for adjusting a preset function of the vehicle, wherein the preset function refers to a function of simulating the scale adjustment of the operation of a vehicle component;

[0157] 02, read the last round of voice request for adjusting the vehicle's preset functions;

[0158] 03. Use the previous round of voice requests to rewrite the current round of voice requests;

[0159] 04. Accurately recognize the rewritten voice request of the current round;

[0160] 05. Complete voice interaction based on the results of precision recognition.

[0161] It is understood that a computer program includes computer program code. The computer program code may be in source code form, object code form, executable file, or some intermediate form. Computer-readable storage media may include any entity or device capable of carrying computer program code, recording media, USB flash drives, mobile hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), and software distribution media.

[0162] When the computer program 31 stored in the computer-readable storage medium 30 of the present invention is executed by the processor 40, upon receiving a user's voice request for a preset vehicle function, the system can read the previous voice request and use it to rewrite the current voice request so that the rewritten voice request can be recognized by the system as corresponding to the scale adjustment precision. The system then issues control instructions based on the precision recognition results to control the corresponding vehicle components, completing voice interaction. By combining instruction rewriting and precision recognition to identify the scale adjustment precision of the voice request, the system can accurately identify the scale adjustment precision for simulating the operation of vehicle components through voice interaction under multiple voice requests.

Claims

1. A voice interaction method, characterized in that: include: receiving a voice request forwarded by the vehicle for adjusting a preset function of the vehicle, wherein the preset function is a function of simulating a scale adjustment of an operation of a vehicle component; Read the last round of voice requests for adjusting the vehicle's preset functions; Rewriting the voice request of the current round by using the voice request of the previous round; performing precision recognition on the rewritten voice request of the current round; Complete voice interaction according to the result of the precision recognition; The voice interaction method comprises: Perform reverse mining on two adjacent rounds of voice requests that occur more frequently than a preset frequency to build a regularization engine; After reading the last round of voice request for adjusting the preset function of the vehicle, the voice interaction method includes: Using a regular expression engine to identify the scale adjustment precision corresponding to the current round of voice requests and the previous round of voice requests; When the recognition result of the regularization engine is that the corresponding scale adjustment precision is recognized, the voice interaction is completed according to the recognized scale adjustment precision.

2. The voice interaction method according to claim 1, characterized in that: The rewriting of the current round of voice requests by using the previous round of voice requests includes: When the recognition result of the regularization engine is that the corresponding scale adjustment accuracy cannot be recognized, the voice request of the current round is rewritten using the voice request of the previous round.

3. The voice interaction method according to claim 1, wherein: The rewriting of the current round of voice requests by using the previous round of voice requests includes: A rewriting model is obtained by training rewriting training data, wherein the rewriting training data includes two adjacent rounds of voice requests; The voice request of the current round is rewritten using the voice request of the previous round and the rewriting model.

4. The voice interaction method according to claim 1, wherein: The performing accurate recognition on the rewritten voice request of the current round includes: Obtaining a precision recognition model through training with precision training data, wherein the precision training data is related to a vehicle component capable of scale adjustment, a scale adjustment range of the vehicle component, and a scale adjustment precision range of the component; The precision recognition model is used to perform precision recognition on the rewritten voice request of the current round.

5. The voice interaction method according to claim 4, characterized in that: The completing the voice interaction according to the result of the precision recognition includes: Obtaining the accuracy discrimination probability of each preset scale adjustment accuracy corresponding to the accuracy identification result; The preset scale adjustment accuracy whose accuracy judgment probability is greater than a probability threshold is determined as the target scale adjustment accuracy corresponding to the voice request to complete the voice interaction.

6. The voice interaction method according to claim 5, characterized in that: The voice interaction method includes: When the accuracy discrimination probabilities of the preset scale adjustment accuracies are not greater than a probability threshold, it is determined that the accuracy recognition error of the voice request of the current round is a error.

7. A voice interaction device, characterized in that: The voice interaction device includes: a command receiving module, the command receiving module being configured to receive a voice request forwarded by the vehicle for adjusting a current wheel of a preset function of the vehicle, the preset function being a function of simulating a scale adjustment of an operation of a vehicle component; A command reading module, the command reading module being used to read a previous round of voice requests for adjusting a preset function of the vehicle; a rewriting module, configured to rewrite the voice request of the current round using the voice request of the previous round; an accuracy recognition module, the accuracy recognition module being used to perform accuracy recognition on the rewritten voice request of the current round; An interaction module, configured to perform voice interaction based on the result of the precision recognition; A regularization engine module, which is used to perform reverse mining on two adjacent rounds of voice requests whose occurrence frequency is greater than a preset frequency to build a regularization engine; After reading the previous round of voice requests for adjusting the vehicle's preset functions, the regularization engine module is used to use the regularization engine to identify the scale adjustment accuracy corresponding to the current round of voice requests and the previous round of voice requests; when the recognition result of the regularization engine is that the corresponding scale adjustment accuracy is identified, the voice interaction is completed according to the identified scale adjustment accuracy.

8. A server, characterized in that: The server includes a processor and a memory, and a computer program is stored in the memory. When the computer program is executed by the processor, the voice interaction method according to any one of claims 1 to 6 is implemented.

9. A non-volatile computer-readable storage medium containing a computer program, characterized in that: When the computer program is executed by one or more processors, the voice interaction method according to any one of claims 1 to 6 is implemented.

Images