Intention recognition method, electronic device, vehicle and medium

Abstract

The present invention discloses an intention recognition method, an electronic device, a vehicle and a storage medium, which are used in the field of vehicle technology. The method includes: when a voice command is received, the voice command is recognized to obtain a voice recognition result; when the voice recognition result meets a preset guidance condition, a guidance message is issued, and the guidance message is generated based on the voice recognition result; when a feedback voice for the guidance message is received, the feedback voice is recognized to obtain an intention recognition result. Even if a voice command is received that is difficult for the vehicle to clearly identify the intention or the vehicle cannot guarantee the accuracy of recognition, the intention recognition result can be obtained by interacting with the user by issuing a guidance message, rather than refusing to recognize, replying to the user that the intention cannot be recognized or executed incorrectly, etc., thereby improving the accuracy of the executed intention and improving the user's experience.

Description

Technical Field

[0001] The present application relates to the field of vehicle technology, and more specifically, to an intention recognition method, an electronic device, a vehicle, and a non-volatile computer-readable storage medium. Background Art

[0002] After the vehicle performs speech recognition and intent understanding on the sound signal, it can accurately understand and execute the corresponding intention for sentences with clear intentions, bringing a good experience to users. Users no longer need to repeatedly explain or clarify their needs. The vehicle can make a satisfactory response based on speech recognition and intent understanding.

[0003] However, in daily life and applications, not all voices received by vehicles contain clear intentions. When encountering voices with irregular expressions or unclear intentions, the vehicle will usually choose to refuse to recognize them, or issue an "unrecognizable" prompt to the user, or even directly execute the intention obtained by generalizing the voice with unclear intentions (this intention is usually obtained by generalizing the unclear intention when the vehicle cannot correctly recognize it, and the accuracy is usually low), resulting in a poor user experience. Summary of the invention

[0004] In view of this, the embodiments of the present application provide an intent recognition method, an electronic device, a vehicle and a non-volatile computer-readable storage medium, which can improve the user experience.

[0005] The intention recognition method of the present application includes: when a voice command is received, the voice command is recognized to obtain a voice recognition result; when the voice recognition result meets a preset guidance condition, guidance information is issued, and the guidance information is generated based on the voice recognition result; when feedback voice is received for the guidance information, the feedback voice is recognized to obtain an intention recognition result.

[0006] In some embodiments, the speech recognition result includes a classification result and an extracted word, the classification result includes whether there is an intention, the extracted word includes at least one of an entity word, a verb and a directional word, and the speech recognition result satisfies a preset guidance condition, including at least one of the following: the classification result is that there is an intention, there are only entity words in the extracted words, and the confidence of the entity words is greater than a first confidence threshold; the classification result is that there is an intention, there are only entity words and verbs in the extracted words, and the confidence of the entity words is between a second confidence threshold and the first confidence threshold, and the second confidence threshold is less than the first confidence threshold; the classification result is that there is an intention, there are only entity words in the extracted words, and the confidence of the entity words is between the second confidence threshold and the first confidence threshold, and the second confidence threshold is less than the first confidence threshold; the classification result is that there is an intention, there are only entity words in the extracted words, and the confidence of the entity words is between the second confidence threshold and the first confidence threshold, and the second confidence threshold is less than the first confidence threshold; the classification result is that there is an intention, there are only directional words in the extracted words; the classification result is that there is an intention, there are only verbs in the extracted words; the classification result is that there is an intention, and there are only directional words and verbs in the extracted words.

[0007] In some embodiments, the speech recognition result includes extracted words, and the extracted words include at least one of entity words, verbs, and directional words. The method also includes: generating the guidance information based on at least one of the target words missing from the extracted words and the confidence of the entity words, and the target words include at least one of verbs and directional words.

[0008] In certain embodiments, the generating of the guidance information based on at least one of a target word missing from the extracted word and a confidence of the entity word comprises: when the extracted word contains an entity word, generating the guidance information based on the target word missing from the extracted word and the confidence of the entity word; when the extracted word does not contain an entity word, generating the guidance information based on the target word missing from the extracted word.

[0009] In some embodiments, the guidance information includes direction guidance information, action guidance information, action and direction guidance information, entity guidance information, action and entity guidance information, and direction and entity guidance information; when the extracted word contains an entity word, the guidance information is generated based on the target word missing from the extracted word and the confidence of the entity word, including: when only entity words exist in the extracted word and the confidence of the entity word is greater than a first confidence threshold, the action guidance information or the action and direction guidance information is generated; when only entity words and verbs exist in the extracted word and the confidence of the entity word is between the second confidence threshold and the first confidence threshold values, generating the entity guidance information or the orientation and entity guidance information; generating the action and entity guidance information when only entity words exist in the extracted words and the confidence of the entity words is between the second confidence threshold and the first confidence threshold; generating the guidance information based on the target words missing in the extracted words when the extracted words do not contain entity words, including: generating the action and entity guidance information when only orientation words exist in the extracted words; generating the orientation and entity guidance information when only verbs exist in the extracted words; generating entity guidance information when only orientation words and verbs exist in the extracted words.

[0010] In some embodiments, the speech recognition result includes a classification result and extracted words, the classification result includes whether there is an intention, and the extracted words include at least one of entity words, verbs, and directional words. The method also includes: when the recognition result meets a preset intention generation condition, generating the intention recognition result based on the speech recognition result; wherein the intention generation condition includes that the classification result is that there is an intention, the extracted words include verbs and entity words, and the confidence of the entity words is greater than a first confidence threshold, or the classification result is that there is an intention, the extracted words include verbs, directional words and entity words, and the confidence of the entity words is greater than a first confidence threshold.

[0011] In some embodiments, the recognition result includes a classification result and extracted words, the classification result includes whether there is an intention, the extracted words include at least one of entity words, verbs and directional words, and the method also includes: when the recognition result satisfies a preset stop condition, stopping the intention recognition; wherein, the recognition result satisfies the preset stop condition, including at least one of the following: the classification result is that there is no intention; the extracted words do not have directional words; the confidence of the entity words in the extracted words is less than a third confidence threshold.

[0012] In some embodiments, the confidence of the entity word is determined based on a first confidence and a second confidence, wherein the first confidence is determined based on the maximum similarity between each matching entity word that matches the entity word and the entity word in a preset intent configuration table, and the second confidence is determined based on the similarity between the pinyin of the entity word and the pinyin of the matching entity word corresponding to the maximum similarity.

[0013] In some embodiments, the recognizing the voice instruction to obtain the voice recognition result includes: recognizing the voice instruction based on a preset deep learning model to obtain the voice recognition result.

[0014] The electronic device of an embodiment of the present application includes a processor, which is connected to a memory; a computer program is stored in the memory, and the processor executes the computer program to implement instructions of the intent recognition method described in any of the above embodiments.

[0015] The vehicle according to the embodiment of the present application includes the electronic device described in any of the above embodiments.

[0016] The non-volatile computer-readable storage medium of an embodiment of the present application includes a computer program. When the computer program is executed by a processor, the processor executes the intention recognition method described in any of the above embodiments.

[0017] The intention recognition method, electronic device, vehicle and non-volatile computer-readable storage medium of the embodiments of the present application, when receiving a voice command, recognize the voice command to obtain a voice recognition result; when the voice recognition result meets the preset guidance condition, issue guidance information, the guidance information is generated based on the voice recognition result; when receiving feedback voice for the guidance information, recognize the feedback voice to obtain an intention recognition result, that is, based on the voice recognition result obtained after recognizing the voice command, issue guidance information to the user, and then obtain the intention recognition result by receiving and recognizing the user's feedback voice. Even if the vehicle receives a voice command whose intention is difficult to clarify or the vehicle cannot guarantee the accuracy of recognition, it is possible to interact with the user by issuing guidance information to obtain the intention recognition result, rather than refusing to recognize, replying to the user's intention that cannot be recognized or executed incorrectly, etc., thereby improving the accuracy of the executed intention and improving the user's usage experience.

[0018] Additional aspects and advantages of the embodiments of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the embodiments of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0020] Figure 1 It is a schematic diagram of an application scenario of the intention recognition method of certain embodiments of the present application;

[0021] Figure 2 is a flowchart of an intention recognition method in certain embodiments of the present application;

[0022] Figure 3 is a flowchart of an intention recognition method in certain embodiments of the present application;

[0023] Figure 4 is a flowchart of an intention recognition method in certain embodiments of the present application;

[0024] Figure 5 is a flowchart of an intention recognition method in certain embodiments of the present application;

[0025] Figure 6 is a schematic diagram of a scenario of an intention recognition method in certain embodiments of the present application;

[0026] Figure 7 is a schematic diagram of a scenario of an intention recognition method in certain embodiments of the present application;

[0027] Figure 8 is a schematic diagram of a scenario of an intention recognition method in certain embodiments of the present application;

[0028] Fig. 9 is a schematic diagram of a scenario of an intention recognition method in certain embodiments of the present application;

[0029] Fig.10 is a flowchart of an intention recognition method in certain embodiments of the present application;

[0030] Fig.11 is a flowchart of an intention recognition method in certain embodiments of the present application;

[0031] Fig.12 is a schematic diagram of a module of an intention recognition device in certain embodiments of the present application;

[0032] Fig.13 is a schematic diagram of the structure of a vehicle according to some embodiments of the present application;

[0033] Fig.14 It is a schematic diagram of the connection status of a non-volatile computer-readable storage medium and a processor in certain embodiments of the present application. DETAILED DESCRIPTION

[0034] The embodiments of the present application 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 from beginning to end. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the embodiments of the present application, and cannot be understood as limiting the embodiments of the present application.

[0035] To facilitate understanding of this application, the terms appearing in this application are explained below:

[0036] 1. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines so that the machines have the functions of perception, reasoning and decision-making. Artificial intelligence technology is a comprehensive discipline that involves a wide range of fields, including both hardware-level technology and software-level technology. Basic artificial intelligence technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, and mechatronics. Artificial intelligence software technology mainly includes several major directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning. The technical solutions provided in the embodiments of this application mainly involve natural language processing technology and machine learning / deep learning in artificial intelligence.

[0037] 2. Machine Learning (ML) is a multi-disciplinary subject involving probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and other disciplines. It specializes in studying how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications are spread across all areas of artificial intelligence. Machine learning and deep learning usually include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and self-learning.

[0038] 3. Deep Learning (DL): It is a branch of machine learning and an algorithm that attempts to use multiple processing layers containing complex structures or composed of multiple nonlinear transformations to perform high-level abstraction of data. Deep learning is to learn the inherent laws and representation levels of training sample data. The information obtained in the learning process is of great help in the interpretation of data such as text, images and sounds. The ultimate goal of deep learning is to enable machines to have analytical learning capabilities like humans and to be able to recognize data such as text, images and sounds. Deep learning is a complex machine learning algorithm that has achieved results in speech and image recognition that far exceed previous related technologies.

[0039] 4. Natural Language Understanding (NLU): It is a general term for all methods, models or tasks that support machines to understand text content. NLU plays a very important role in text information processing systems and is an essential module for recommendation, question-answering, search and other systems. In the implementation of this application, NLU is used to learn and understand user intentions in order to better respond to instructions and needs and improve interaction efficiency and accuracy.

[0040] 5. Automatic Speech Recognition (ASR): The goal is to convert the vocabulary content in human speech into computer-readable input, such as keystrokes, binary codes, or character sequences.

[0041] Vehicle voice recognition technology refers to the automatic recognition and understanding of vehicle status, operation or environment by analyzing sound signals inside or around the vehicle. It can be used in the fields of vehicle safety, driver assistance systems, in-vehicle entertainment, etc. In-vehicle voice data processing refers to the technology that uses voice as the main input and output method for operation and feedback of the interactive interface inside the vehicle.

[0042] In vehicle voice data processing, the clear intention contained in the sound signal is the key for the system to make correct decisions and actions. After the vehicle performs voice recognition and intent understanding on the sound signal, it can accurately understand and execute the corresponding intention for sentences with clear intentions, bringing a good experience to users. Users no longer need to repeatedly explain or clarify their needs. The vehicle can make a satisfactory response based on voice recognition and intent understanding.

[0043] However, in daily life and applications, not all voices received by vehicles contain clear intentions. When encountering voices with irregular expressions or unclear intentions, the vehicle will usually choose to refuse to recognize them, or issue a "cannot recognize" prompt to the user, or even directly execute the intention obtained by generalizing the voice with unclear intentions (the intention is obtained by generalizing the unclear intention when the vehicle cannot correctly recognize it, and the accuracy is usually low), resulting in a poor user experience.

[0044] In order to solve the above technical problems, an embodiment of the present application provides an intent recognition method.

[0045] The following first introduces an application scenario of the technical solution of this application. Figure 1 As shown, the intention recognition method provided by this application can be applied to Figure 1 In the application scenario shown, the intention recognition method can be applied to the vehicle 100 .

[0046] The vehicle 100 is any vehicle 100 capable of processing voice data, such as a car, a truck, etc.

[0047] The vehicle 100 includes a vehicle body 50 .

[0048] In one embodiment, the vehicle 100 further includes a microphone 20 , which is disposed inside the vehicle body 50 . The microphone 20 is used to collect audio information inside or around the vehicle. A user can issue voice commands to the vehicle through the microphone 20 .

[0049] The intent recognition method of this application is described in detail below:

[0050] See also Figure 1 and Figure 2 , the embodiment of the present application provides an intention recognition method, the intention recognition method comprising:

[0051] Step 011: When a voice command is received, the voice command is recognized to obtain a voice recognition result.

[0052] Among them, voice instructions may include requests or queries sent by users to the system by voice, representing the functions or actions that users want the vehicle to perform. In the field of voice recognition and voice control, terms such as "voice request", "voice instruction", "voice command", and "voice control instruction" are often used to describe the voice phrases used by users to trigger specific operations or tasks. These voice instructions can be single words, phrases, or complete sentences, which are called "queries". In other words, voice instructions can be queries used by users in voice data processing. For example, voice instructions can be "please open the window", "raise the driver's seat", and "turn off the vehicle air conditioning".

[0053] Specifically, the user can issue a voice command through the microphone on the vehicle. The voice command can be a voice command issued by the user collected by the microphone of the vehicle; or the vehicle is associated with the terminal (such as through a network connection, etc.), the user issues a voice command through the terminal, and after receiving the voice command, the terminal sends the received voice command to the vehicle. When the vehicle receives the voice command, it can first pre-process the voice command, for example, by pre-processing methods such as denoising, audio enhancement, and audio segmentation to improve the audio quality; then based on ASR speech recognition technology, such as voice feature extraction, the voice command is recognized and the corresponding speech recognition result is generated.

[0054] See also Figure 3 Optionally, step 011: upon receiving a voice command, recognizing the voice command to obtain a voice recognition result, including:

[0055] Step 0111: Based on a preset deep learning model, the voice command is recognized to obtain a voice recognition result.

[0056] The preset deep learning model may be a neural network model based on deep learning. The deep learning model may recognize voice commands to output corresponding voice recognition results (for example, voice recognition results in text form).

[0057] Among them, the preset deep learning model can be a named entity recognition (Named Entity Recognition) model (Named Entity NER model), which can include binary classification output (i.e., whether the voice command contains intent, etc.). The model structure can include Bert (an open source pre-trained model), CRF (a conditional random field pattern, commonly used in named entity recognition tasks) and Dense (a fully connected layer in deep learning). Each character is digitally serialized through tokens (the abbreviation of digital serialization mapping of characters in deep learning), and then input into the Bert pre-trained encoding layer (input layer) to obtain the feature vector representation of each character, and then pass the feature vector through the model to output the results of the CRF layer and the Dense layer, that is, to obtain the classification results of the speech recognition results and the words (such as verbs, entity words, etc.) included in the speech recognition results.

[0058] Specifically, the deep learning model can be obtained through training, and the training process is as follows: by obtaining samples of voice commands and label information corresponding to the samples of voice commands, the label information includes the voice recognition results of the corresponding voice commands; the voice commands are input into the deep learning model for training, and the training voice recognition results are output. Based on the loss value between the training voice recognition results and the voice recognition results included in the label information, the deep learning model is adjusted until the deep learning model converges.

[0059] When a voice command is received, the corresponding voice recognition result can be output by inputting the voice command into the deep learning model.

[0060] Step 012: When the speech recognition result meets the preset guidance condition, a guidance message is issued, and the guidance message is generated based on the speech recognition result.

[0061] The preset guidance condition may be a condition for distinguishing whether to issue guidance information and what kind of guidance information to issue. For example, the preset guidance condition may be a keyword. When the voice recognition result includes the keyword, it is considered that the voice recognition result meets the preset guidance condition, and the vehicle is controlled to issue guidance information. Guidance information may be information for guiding the user to confirm the intention included in the voice instruction, and the guidance information is generated based on the voice recognition result.

[0062] Specifically, when the voice recognition result meets the preset guidance conditions (i.e., when the vehicle receives a voice command with unclear intention), guidance information can be sent to the user through the vehicle's speakers, display screen, etc. to guide the user to confirm the intention included in the voice command. For the received voice command with unclear intention, the voice recognition result generated by the voice command can be judged, and corresponding guidance information can be generated based on the voice recognition result to guide the user to confirm.

[0063] Optionally, the speech recognition results include classification results and extracted words. The classification results include whether there is an intention. The extracted words include at least one of entity words, verbs and directional words. The entity words are used to represent objects operated by voice commands. For example, in a vehicle, the entity words can be the vehicle's windows, seats, etc.

[0064] The classification result includes whether there is an intention, and not all voices received by the vehicle include intentions. When the vehicle receives voices, it can classify the voices to determine whether the voices received at this time contain intentions.

[0065] Entity words can be used to refer to things or devices in the vehicle. Entity words can include objects that users expect to be acted upon in the intent. For example, entity words can be car windows, sunroofs, etc. Verbs can be actions that users expect to be acted upon in the intent. For example, open, close, etc. Direction words can be used to represent the direction of the entity words, for example, driver's seat, co-driver's seat, etc.

[0066] It can be understood that the vehicle understands the intention of the user's voice command through the entity words, verbs and directional words in the intent, and understands what kind of device (entity word) the user wants to perform what action (verb) on where (directional word).

[0067] Optionally, the speech recognition result satisfies a preset guidance condition, including at least one of the following:

[0068] (1) The classification result is that there is an intent, only entity words exist in the extracted words, and the confidence of the entity words is greater than the first confidence threshold.

[0069] Optionally, the confidence of the entity word is determined based on a first confidence and a second confidence, the first confidence being determined based on the maximum similarity between each matching entity word and the entity word in a preset intent configuration table, and the second confidence being determined based on the similarity between the pinyin of the entity word and the pinyin of the matching entity word corresponding to the maximum similarity.

[0070] Among them, the preset intention configuration table can be generated through experiments, table lookups, etc. The preset intention configuration table includes entity words existing in the vehicle (such as windows, etc.) and pinyin information corresponding to each entity word.

[0071] By obtaining a preset intention configuration table, the intention configuration table includes entity words existing in the vehicle and the pinyin corresponding to the entity words. According to the intention configuration table, the matching entity word with the greatest similarity to the entity word is determined, and the matching entity word is obtained, and then the first confidence is determined based on the entity word and the matching entity word; then the pinyin information of the entity word is obtained (such as obtained by a pinyin function, etc.), and the second confidence is determined based on the pinyin information of the entity word and the pinyin information of the matching entity word, which can avoid the problem of misinterpretation of homophones during speech recognition due to different tones of the same word in different regions. For example, through a preset similarity comparison model, the similarity comparison model can be used to determine the matching entity word with the greatest similarity to the entity word, and the similarity between the entity word and the matching entity word (i.e., the first confidence). The entity word is input into the similarity comparison model to output the first confidence; then based on the pinyin information of the entity word and the pinyin information of the matching entity word, for example, the second confidence is calculated by a cosine function (Cosine).

[0072] The confidence of the entity word can be determined based on the first confidence and the second confidence. For example, confidence S = 0.5 * first confidence S1 + 0.5 * second confidence S2; or, confidence S = 0.4 * first confidence S1 + 0.6 * second confidence S2, etc. By combining the first confidence and the second confidence to determine the confidence, the misrecognition of homophones and synonyms can be reduced, and the recognition accuracy can be improved.

[0073] The first confidence threshold may be determined based on an empirical value, etc. For example, the first confidence threshold may be 0.5, 0.6, 0.7, 1, etc.

[0074] In the case where the classification result is that there is an intention and only entity words exist in the extracted words, since there are only entity words in the extracted words of the voice command, it is impossible to clearly know what intention the user wants the vehicle to execute based on the entity words. In the case where the confidence of the entity words is greater than the first confidence threshold, it can be considered that the voice command has a greater probability of including matching entity words. Therefore, guidance information is sent to the user, and the user's intention is confirmed through interaction with the user to improve the user's experience. That is, in the case where the classification result is that there is an intention, only entity words exist in the extracted words, and the confidence of the entity words is greater than the first confidence threshold, it is considered that the preset guidance conditions are met, and the user is guided to confirm the intention by sending guidance information.

[0075] (2) The classification result is that there is an intention, only entity words and verbs exist in the extracted words, and the confidence of the entity words is between the second confidence threshold and the first confidence threshold, and the second confidence threshold is less than the first confidence threshold.

[0076] The second confidence threshold is smaller than the first confidence threshold, for example, the first confidence threshold may be 1 and the second confidence threshold may be 0.6; or the first confidence threshold may be 0.8 and the second confidence threshold may be 0.4, etc. When the confidence of the entity word is between the second confidence threshold and the first confidence threshold, it can be considered that the probability of including the matching entity word in the voice instruction is medium.

[0077] When the classification result shows that there is an intent and only entity words and verbs exist in the extracted words, it is impossible to clearly know the entity words in what direction the user expects to perform an action, even for voice commands such as "open the sunroof" that do not require directional words to clarify the intent. Since the confidence of the entity words is between the second confidence threshold and the first confidence threshold, the accuracy of the entity words is average. Therefore, the user's intent can be confirmed by sending guidance information to the user (such as confirming the entity words, confirming the directional words and entity words, etc.) to improve the user experience.

[0078] (3) The classification result is that there is an intention, only entity words exist in the extracted words, and the confidence of the entity words is between the second confidence threshold and the first confidence threshold, and the second confidence threshold is less than the first confidence threshold.

[0079] When the classification result is that there is an intent, only entity words exist in the extracted words, and the confidence of the entity words is between the second confidence threshold and the first confidence threshold, the vehicle determines that the user's voice command contains an intent, but cannot clearly identify the direction of the intent. Therefore, the vehicle can send guidance information to the user to confirm the user's intent (such as confirming entity words, confirming verbs and entity words, etc.) to improve the user experience.

[0080] (4) The classification result shows that there is an intention and only directional words exist in the extracted words.

[0081] When the classification result shows that there is an intention and only directional words exist in the extracted words, the vehicle can only determine the direction but cannot clearly identify the specific device based on the direction. The user's intention can be confirmed by sending guidance information to the user (such as confirming entity words, confirming verbs and entity words, etc.).

[0082] (5) The classification result is that there is an intention and the extracted words only contain verbs.

[0083] When the classification result shows that there is an intention and only directional words exist in the extracted words, the vehicle can only confirm the action that the user wants to perform, but cannot clearly identify the entity of the action to be performed included in the intention. Therefore, the user's intention can be confirmed by sending guidance information to the user (such as confirming entity words, confirming directional words, etc.).

[0084] (6) The classification result shows that there is an intention and the extracted words only contain directional words and verbs.

[0085] When the classification result shows that there is an intention and the extracted words only contain directional words and verbs, the vehicle cannot clearly identify the entities of the actions being performed included in the intention. Therefore, the user's intention can be confirmed by sending guidance information to the user (such as confirming entity words, etc.).

[0086] Step 013: When feedback voice for guidance information is received, the feedback voice is recognized to obtain an intention recognition result.

[0087] The feedback voice may be a response or feedback made by the user to the guidance information.

[0088] Specifically, the vehicle sends out guidance information to guide the user to confirm the intention; when receiving feedback voice for the guidance information, the feedback voice is recognized to obtain the intention recognition result, and through interaction with the user, the accuracy of the recognition of the intention is improved, thereby improving the user experience.

[0089] In this way, when a voice command is received, the voice command is recognized to obtain a voice recognition result; when the voice recognition result meets the preset guidance condition, guidance information is issued, and the guidance information is generated based on the voice recognition result; when feedback voice is received for the guidance information, the feedback voice is recognized to obtain an intention recognition result, that is, based on the voice recognition result obtained after recognizing the voice command, guidance information is issued to the user, and then the intention recognition result is obtained by receiving and recognizing the user's feedback voice. Even if a voice command is received whose intention is difficult to clarify for the vehicle, or the vehicle cannot guarantee the accuracy of recognition, the intention recognition result can be obtained by interacting with the user by issuing guidance information, rather than refusing to recognize, replying to the user's unrecognizable or erroneous intention, etc., thereby improving the accuracy of the executed intention and improving the user's experience.

[0090] See also Figure 4 In some embodiments, the speech recognition result includes an extracted word, the extracted word includes at least one of an entity word, a verb, and a directional word, and the intention recognition method further includes:

[0091] Step 014: Generate guidance information based on at least one of the confidences of target words and entity words missing from the extracted words, where the target words include at least one of verbs and directional words.

[0092] Among them, the guidance information can be generated based on the target word missing in the extracted words; or, the guidance information can be generated based on the confidence of the entity word; or, the guidance information can be generated based on the confidence of the target word and entity word missing in the extracted words, and this application does not impose any restrictions on this.

[0093] Among them, the target word may include a verb; or, the target word may include a directional word; or, the target word may include a verb and a directional word, and this application does not impose any restrictions on this.

[0094] Specifically, guidance information is generated based on at least one of the confidences of the target word and entity word missing from the extracted words. The vehicle determines the user's intention by understanding the extracted words, and therefore, can generate corresponding guidance information based on the target word missing from the extracted words. In the case where the confidence of the entity word is high, the vehicle can determine the entity included in the intention. In the case where the confidence of the entity word is low, the vehicle cannot determine the entity included in the intention. In order to ensure the user's experience, corresponding guidance information can also be generated to guide the user to confirm.

[0095] See also Figure 5 Optionally, step 014: generating guidance information based on at least one of the confidences of the target word and entity word missing from the extracted word, the target word including at least one of a verb and a directional word, including:

[0096] Step 0141: when the extracted words contain entity words, generating guidance information based on the confidence of the target words and entity words missing from the extracted words;

[0097] Step 0142: When the extracted word does not contain an entity word, guide information is generated based on the target word missing from the extracted word.

[0098] Specifically, the corresponding guidance information can be generated by determining whether the extracted words contain entity words. In the case where the extracted words contain entity words, the guidance information is generated based on the target words missing from the extracted words and the confidence of the entity words. Based on the missing target words, the operation that the user expects the vehicle to perform in the voice command is clarified. Based on the confidence of the entity words, it is determined whether the entity words need to be confirmed, thereby generating the corresponding guidance information. In the case where the extracted words do not contain entity words, the vehicle cannot clarify the intention, so the guidance information can be generated based on the target words missing from the extracted words.

[0099] For example, when a user's voice command "open the main driving bed for me" is received, the voice command is recognized to obtain a voice recognition result, which includes the verb "open", the directional word "main driving" and the entity word "lathe", the extracted words include the entity words, and there is no lack of target words, and the corresponding guidance information can be generated based on the confidence corresponding to the entity words: according to the intention configuration table, it is determined that the matching entity word with the greatest similarity to "lathe" is "car window", based on "lathe" and "car window", a first confidence is determined, and then based on the pinyin "chechuang" of "lathe" and the pinyin "chechuang" of "car window", a second confidence is determined, and the confidence is determined according to the first confidence and the second confidence, and the corresponding guidance information is generated according to the confidence of the entity word, such as "Do you want to open the main driving window? You can try to say to me to open the main driving window", and when feedback voice for the guidance information is received, such as "yes" or "open the main driving window for me" answered by the user, the feedback voice is recognized, and the intention recognition result is obtained to execute the corresponding intention.

[0100] For another example, when the user's voice command "Tianchuang" is received, the voice command is recognized to obtain a voice recognition result. The voice recognition result only includes the entity word "Tianchuang" and lacks target words (verbs and directional words). Then, by calculating the confidence of "Tianchuang" and "Skylight", based on the missing target words (verbs and directional words) and the confidence of "Tianchuang" and "Skylight", corresponding guidance information is generated, such as "Do you want to open or close the skylight? You can try to say open the skylight to me", and when feedback voice is received for the guidance information, such as the user's answer of "yes" or "open the skylight for me", the feedback voice is recognized and the intention recognition result is obtained to execute the corresponding intention.

[0101] For another example, when the user's voice command is "Close the front passenger window for me", the voice command is recognized to obtain a voice recognition result, which includes the verb "close" and the directional word "front passenger window", but does not include entity words. Therefore, the guidance information generated based on the target word (i.e., entity word) missing in the extracted word can be: "What do you want to close on the front passenger window? You can try to say to me to close the front passenger window". When feedback voice is received for the guidance information, such as the user's answer of "yes" or "Close the front passenger window XX", the feedback voice is recognized and the intention recognition result is obtained to execute the corresponding intention.

[0102] Optionally, the guidance information includes orientation guidance information, action guidance information, action and orientation guidance information, entity guidance information, action and entity guidance information, orientation and entity guidance information;

[0103] The guidance information includes direction guidance information, action guidance information, action and direction guidance information, entity guidance information, action and entity guidance information, and direction and entity guidance information. The direction guidance information is used to guide the user to confirm the direction in the voice command; the action guidance information is used to guide the user to confirm the action that the vehicle is expected to perform in the voice command; the action and direction guidance information is used to guide the user to confirm the action and direction in the voice command; the entity guidance information is used to guide the user to confirm the entity on which the action is performed in the voice command; the action and entity guidance information is used to guide the user to confirm the entity on which the action is performed and the specific action to be performed; the direction and entity guidance information is used to guide the user to confirm the entity on which the action is performed and the specific location of the entity.

[0104] Step 0141: When the extracted words contain entity words, guide information is generated based on the confidence of the target words and entity words missing from the extracted words, including:

[0105] (1) If only entity words exist in the extracted words and the confidence of the entity words is greater than a first confidence threshold, action guidance information or action and orientation guidance information is generated;

[0106] If only entity words exist in the extracted words and the confidence of the entity words is greater than a first confidence threshold, action guidance information or action and direction guidance information is generated to guide the user to confirm the action to be performed or the action to be performed and the direction. For example, refer to Figure 6 , when the received voice command is "Please open the car window", that is, the extracted words only include the entity word "car window" and the confidence of the entity word (0.9) is greater than the first confidence threshold (0.8), since there are no verbs and directional words in the voice command, action guidance information can be generated, such as "Do you want to open or close the car window? You can try to say to me to open the car window." Or, action and directional guidance information can be generated, such as "Do you want to open or close the main driver's window? You can try to say to me to open the main driver's window" and so on.

[0107] (2) when only entity words and verbs exist in the extracted words and the confidence of the entity words is between the second confidence threshold and the first confidence threshold, generating entity guidance information or orientation and entity guidance information;

[0108] When only entity words and verbs exist in the extracted words, and the confidence of the entity words is between the second confidence threshold and the first confidence threshold, entity guidance information or direction and entity guidance information is generated to guide the user to confirm the entity on which the action is performed or the entity on which the action is performed and the direction of the entity. For example, when the received voice instruction is "Please open Tianchuang for me", that is, only entity words ("Tianchuang") and verbs ("open") exist in the extracted words, and the confidence of the entity words (0.7) is between the second confidence threshold (0.6) and the first confidence threshold (0.8), since the confidence of the entity word "Tianchuang" and "sunroof" in the intent configuration table is 0.7, entity guidance information such as "Do you want to open the sunroof?" can be generated to confirm the entity word "sunroof"; or direction and entity guidance information such as "Do you want to open the sunroof at the front of the vehicle? You can try to say to me to open the sunroof at the front of the vehicle" can be generated to guide the user to confirm.

[0109] (3) when only entity words exist in the extracted words and the confidence of the entity words is between the second confidence threshold and the first confidence threshold, generating action and entity guidance information;

[0110] When only entity words exist in the extracted words and the confidence of the entity words is between the second confidence threshold and the first confidence threshold, action and entity guidance information is generated to guide the user to confirm the action to be performed and the entity on which the action is performed. For example, when the received voice instruction is "Please open the car window", that is, only the entity word ("car window") exists in the extracted words and the confidence of the entity word (0.7) is between the second confidence threshold (0.6) and the first confidence threshold (1), since the confidence of the entity word "car window" and the "window" in the intent configuration table is 0.7, the user can be guided to confirm the entity ("window") and the action performed on the "window" ("open" or "close") by generating action and entity guidance information, such as "Do you want to open or close the window? You can try to say to me to open the window".

[0111] Optionally, step 0142: when the extracted words do not contain entity words, generating guidance information based on target words missing from the extracted words includes:

[0112] (1) When only directional words exist in the extracted words, action and entity guidance information is generated;

[0113] When only directional words exist in the extracted words, action and entity guidance information is generated to guide the user to confirm the action to be performed and the entity on which the action is performed. For example, see Figure 7When the received voice command is "main driver", that is, when only the directional word ("main driver") exists in the extracted word, action and entity guidance information is generated, such as "What do you want to open on the main driver's side? You can try to say to me to open the main driver's side window" to guide the user to confirm the command.

[0114] (2) When the extracted word only contains verbs, generate direction and entity guidance information;

[0115] When the extracted word only contains verbs, the location and entity guidance information is generated to guide the user to confirm the entity on which the action is performed and the location of the entity. Figure 8 , when the received voice command is "close", that is, when there is only a verb ("close") in the extracted word, direction and entity guidance information is generated, such as "What do you want to close? You can try to tell me to close the main driving window" to guide the user to confirm the command.

[0116] (3) When only directional words and verbs exist in the extracted words, entity guidance information is generated.

[0117] When only directional words and verbs exist in the extracted words, entity guidance information is generated to guide the user to confirm the entity on which the action is to be performed. For example, see Fig. 9 When the received voice command is "Open the main driver's window for me", that is, the extracted words only contain directional words ("main driver's window") and verbs ("open"), entity guidance information is generated, such as "What do you want to open on the main driver's window? You can try to say to me to open the main driver's window", to guide the user to confirm the command.

[0118] See also Fig.10 In some embodiments, the speech recognition result includes a classification result and an extracted word, the classification result includes whether there is an intention, and the extracted word includes at least one of an entity word, a verb, and a directional word. The method further includes:

[0119] Step 015: When the recognition result satisfies the preset intention generation condition, generate an intention recognition result based on the speech recognition result;

[0120] Among them, the preset intention generation conditions include the classification result being the existence of intention, the extracted words containing verbs and entity words, and the confidence of the entity words being greater than the first confidence threshold, or the classification result being the existence of intention, the extracted words containing verbs, directional words and entity words, and the confidence of the entity words being greater than the first confidence threshold.

[0121] Specifically, when the classification result is that there is an intention, the extracted words include verbs and entity words, and the confidence of the entity words is greater than the first confidence threshold (for example, the extracted words include "open the sunroof", and the confidence of "sunroof" is greater than the first confidence threshold (0.9)); or, when the classification result is that there is an intention, the extracted words include verbs, directional words and entity words, and the confidence of the entity words is greater than the first confidence threshold (for example, the extracted words include "open the main driving window", and the confidence of "window" is greater than the first confidence threshold (0.9)), it can be considered that the vehicle is able to clearly indicate the intention. Therefore, when the recognition result meets the preset intention generation condition, the intention recognition result can be generated based on the speech recognition result.

[0122] See also Fig.11 In some embodiments, the recognition result includes a classification result and an extracted word, the classification result includes whether there is an intention, and the extracted word includes at least one of an entity word, a verb, and a directional word. The method further includes:

[0123] Step 016: When the recognition result meets the preset stop condition, stop the intention recognition;

[0124] The recognition result satisfies the preset stop condition, including at least one of the following:

[0125] (1) The classification result is that there is no intention.

[0126] When the vehicle recognizes the received voice command, it can classify whether the voice command has an intention. For example, in a scenario where the user is chatting in the vehicle, the user does not expect to control the vehicle. At this time, the vehicle classifies the received voice and confirms that the classification result is that there is no intention. If the recognition result meets the preset stop condition that the classification result is that there is no intention, the intention recognition is stopped.

[0127] (2) The extracted words include no entity words, verbs and directional words.

[0128] When the extracted words do not contain entity words, verbs, and directional words, it can be considered that the user does not need the vehicle to execute the corresponding instructions, and the intention recognition can be stopped.

[0129] (3) The confidence of the entity word in the extracted word is less than the third confidence threshold.

[0130] If the confidence of the entity word in the extracted word is less than the third confidence threshold, it can be considered that there is a misjudgment of whether the voice command has an intention, or it is considered that the user does not need the vehicle to execute the intention at this time, and the intention recognition can be stopped.

[0131] See also Fig.12In order to better implement the intention recognition method of the embodiment of the present application, the embodiment of the present application also provides an intention recognition device 300. The intention recognition device 300 may include a speech recognition module 301, a guidance module 302 and an intention recognition module 303. The speech recognition module 301 is used to recognize the voice instruction when receiving the voice instruction to obtain the speech recognition result; the guidance module 302 is used to issue guidance information when the speech recognition result meets the preset guidance condition, and the guidance information is generated based on the speech recognition result; the intention recognition module 303 is used to recognize the feedback voice when receiving the feedback voice for the guidance information to obtain the intention recognition result.

[0132] In one embodiment, the speech recognition results include extracted words, which include at least one of entity words, verbs and directional words. The intention recognition device 300 may also include a first generation module 304, which is used to generate guidance information based on at least one of the confidence levels of target words and entity words missing from the extracted words, and the target words include at least one of verbs and directional words.

[0133] In one embodiment, the first generation module 304 is specifically used to generate guidance information based on the confidence of the target word and the entity word missing in the extracted word when the extracted word contains the entity word; and to generate guidance information based on the target word missing in the extracted word when the extracted word does not contain the entity word.

[0134] In one embodiment, the guidance information includes direction guidance information, action guidance information, action and direction guidance information, entity guidance information, action and entity guidance information, and direction and entity guidance information; the first generation module 304 is also specifically used to generate action guidance information or action and direction guidance information when there are only entity words in the extracted words and the confidence of the entity words is greater than the first confidence threshold; generate entity guidance information or direction and entity guidance information when there are only entity words and verbs in the extracted words and the confidence of the entity words is between the second confidence threshold and the first confidence threshold; generate action and entity guidance information when there are only entity words in the extracted words and the confidence of the entity words is between the second confidence threshold and the first confidence threshold; generate action and entity guidance information when there are only direction words in the extracted words; generate direction and entity guidance information when there are only verbs in the extracted words; generate entity guidance information when there are only direction words and verbs in the extracted words.

[0135] In one embodiment, the speech recognition result includes a classification result and extracted words, the classification result includes whether there is an intention, the extracted words include at least one of entity words, verbs and directional words, and the intention recognition device 300 may also include a second generation module 305, the second generation module 305 is used to generate an intention recognition result based on the speech recognition result when the recognition result meets a preset intention generation condition; wherein the intention generation condition includes the classification result that there is an intention, the extracted words include verbs and entity words, and the confidence of the entity words is greater than a first confidence threshold, or the classification result is that there is an intention, the extracted words include verbs, directional words and entity words, and the confidence of the entity words is greater than the first confidence threshold.

[0136] In one embodiment, the recognition result includes a classification result and extracted words, the classification result includes whether there is an intention, the extracted words include at least one of entity words, verbs and directional words, and the intention recognition device 300 may also include a stop module 306, and the stop module 306 is used to stop the intention recognition when the recognition result meets the preset stop condition; wherein the recognition result meets the preset stop condition, including at least one of the following: the classification result is that there is no intention; the extracted words have no entity words, verbs and directional words; the confidence of the entity words in the extracted words is less than the third confidence threshold.

[0137] In one embodiment, the speech recognition module 301 is further used to recognize the speech command based on a preset deep learning model to obtain a speech recognition result.

[0138] In the above text, the intention recognition device 300 is described from the perspective of functional modules in conjunction with the accompanying drawings. The functional modules can be implemented in hardware form, can be implemented by instructions in software form, or can be implemented by a combination of hardware and software modules. Specifically, the steps of the method embodiments in the embodiments of the present application can be completed by the hardware integrated logic circuit in the processor and / or the instructions in software form. The steps of the method disclosed in the embodiments of the present application can be directly embodied as a hardware coding processor for execution, or can be completed by a combination of hardware and software modules in the coding processor. Optionally, the software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps in the above method embodiments in conjunction with its hardware.

[0139] The electronic device of the embodiment of the present application includes a processor, the processor is connected to a memory, a computer program is stored in the memory, and the processor executes the computer program to implement any one of the intention recognition methods described above. For the sake of brevity, it will not be repeated here.

[0140] See also Fig.13, wherein the electronic device can be used as a processor of the vehicle, and the electronic device can be installed in the vehicle so that the vehicle can implement the intention recognition method of any of the above-mentioned embodiments through the electronic device.

[0141] The vehicle of the embodiment of the present application includes the intention recognition device or electronic device of the above embodiment, such as the intention recognition device or electronic device is a processor of the vehicle. The vehicle implements any one of the above intention recognition methods through the intention recognition device or electronic device.

[0142] In one embodiment, the internal structure diagram of the vehicle can be as follows: Fig.13 As shown, it includes a processor 402, a memory 403, a network interface 404, a display screen 401 and an input device 405 connected through a system bus.

[0143] Among them, the vehicle's processor 402 is used to provide computing and control capabilities. The vehicle's memory 403 includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The vehicle's network interface 404 is used to connect with external devices through a network for intention recognition. When the computer program is executed by the processor, the intention recognition method of any of the above-mentioned embodiments is implemented. The vehicle's display screen 401 can be a liquid crystal display screen or an electronic ink display screen, and the vehicle's input device 405 can be a touch layer covering the display screen 401, or it can be a button, trackball or touchpad set on the vehicle, or it can be an external keyboard, touchpad or mouse, etc.

[0144] Those skilled in the art will understand that Fig.13 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0145] See also Fig.14 The embodiment of the present application also provides a computer-readable storage medium 600 on which a computer program 610 is stored. When the computer program 610 is executed by a processor 620, the steps of the intent recognition method of any of the above-mentioned embodiments are implemented. For the sake of brevity, they are not repeated here.

[0146] In the description of this specification, the descriptions with reference to the terms "certain embodiments", "in an example", "exemplarily", etc., mean that the specific features, structures, materials or characteristics described in conjunction with the embodiments or examples are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

[0147] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0148] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for intention recognition, characterized in that: include: When receiving a voice command, the voice command is recognized to obtain a voice recognition result, wherein the voice recognition result includes an extracted word, the extracted word includes at least one of an entity word, a verb, and a directional word, and the entity word is used to represent an object operated by the voice command; Generate guidance information based on at least one of the confidences of the target word missing from the extracted words and the entity word, the target word includes at least one of a verb and a directional word, the confidence of the entity word is determined based on a first confidence and a second confidence, the first confidence is determined based on a maximum similarity between each matching entity word matching the entity word and the entity word in a preset intention configuration table, and the second confidence is determined based on a similarity between the pinyin of the entity word and the pinyin of the matching entity word corresponding to the maximum similarity; When the speech recognition result satisfies a preset guidance condition, issuing guidance information, wherein the guidance information is generated based on the speech recognition result; When receiving a feedback voice for the guidance information, recognizing the feedback voice to obtain an intention recognition result; Wherein, generating the guidance information based on at least one of the target word missing from the extracted word and the confidence of the entity word comprises: In the case where the extracted words include entity words, generating the guide information based on the target words missing from the extracted words and the confidence of the entity words; In a case where the extracted words do not include entity words, the guide information is generated based on target words missing from the extracted words.

2. The method for intention recognition according to claim 1, characterized in that: The speech recognition result includes a classification result and an extracted word, the classification result includes whether there is an intention, and the speech recognition result satisfies a preset guidance condition, including at least one of the following: The classification result is that there is an intention, only entity words exist in the extracted words, and the confidence of the entity words is greater than a first confidence threshold; The classification result is that there is an intention, only entity words and verbs exist in the extracted words, and the confidence of the entity words is between a second confidence threshold and the first confidence threshold, and the second confidence threshold is less than the first confidence threshold; The classification result is that there is an intention, only entity words exist in the extracted words, and the confidence of the entity words is between a second confidence threshold and the first confidence threshold, and the second confidence threshold is less than the first confidence threshold; The classification result is that there is an intention and only directional words exist in the extracted words; The classification result is that there is an intention and the extracted words only contain verbs; The classification result is that there is an intention and the extracted words only contain directional words and verbs.

3. The method for identifying intention according to claim 1, characterized in that: The guidance information includes at least one of action guidance information, action and orientation guidance information, entity guidance information, action and entity guidance information, orientation and entity guidance information; In the case where the extracted word includes an entity word, generating the guide information based on the target word missing from the extracted word and the confidence of the entity word includes: If only entity words exist in the extracted words and the confidence of the entity words is greater than a first confidence threshold, the action guidance information or the action and orientation guidance information is generated; In a case where only entity words and verbs exist in the extracted words and the confidence of the entity words is between a second confidence threshold and the first confidence threshold, generating the entity guide information or the direction and entity guide information; In a case where only entity words exist in the extracted words and the confidence of the entity words is between a second confidence threshold and the first confidence threshold, generating the action and entity guidance information; In the case where the extracted words do not contain entity words, generating the guide information based on the target words missing from the extracted words includes: In the case that only directional words exist in the extracted words, generating the action and entity guidance information; In the case that the extracted word only contains a verb, generating the direction and entity guidance information; The extracted words only contain directional words and verbs, and entity guidance information is generated.

4. The method for identifying intention according to any one of claims 1 to 3, characterized in that: The speech recognition result includes a classification result and an extracted word, the classification result includes whether there is an intention, and the method further includes: In the case where the speech recognition result satisfies a preset intention generation condition, generating the intention recognition result based on the speech recognition result; Among them, the intention generation condition includes that the classification result is the existence of intention, the extracted words contain verbs and entity words, and the confidence of the entity words is greater than the first confidence threshold, or the classification result is the existence of intention, the extracted words contain verbs, directional words and entity words, and the confidence of the entity words is greater than the first confidence threshold.

5. The intention recognition method according to claim 1, characterized in that: The speech recognition result includes a classification result and an extracted word, the classification result includes whether there is an intention, and the extracted word includes at least one of an entity word, a verb, and a directional word. The method further includes: When the speech recognition result meets the preset stop condition, stop the intention recognition; The speech recognition result satisfies a preset stop condition, including at least one of the following: The classification result is that there is no intention; The extracted words do not contain entity words, verbs, or directional words; The confidence of the entity word in the extracted word is less than a third confidence threshold.

6. The method for identifying intention according to claim 1, characterized in that: The step of recognizing the voice command to obtain a voice recognition result includes: Based on a preset deep learning model, the voice command is recognized to obtain the voice recognition result.

7. An electronic device, characterized in that: include: A processor, the processor is connected to a memory; a computer program is stored in the memory, and the processor executes the computer program to implement instructions of the intent recognition method described in any one of claims 1 to 6.

8. A vehicle, characterized in that: include: The electronic device according to claim 7.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the intention recognition method as described in any one of claims 1 to 6 is implemented.

Images