Task information identification method and device, equipment and storage medium

Abstract

The application discloses a task information identification method and device, equipment and a storage medium, and relates to the technical field of vehicles, and comprises the following steps: obtaining voice information, inputting the voice information into a preset deep learning model, obtaining an intention identification result, a vehicle use intention and slot information, obtaining an expanded intention identification result according to the vehicle use intention and the slot information in combination with the intention identification result, and obtaining task information according to the expanded intention identification result. The voice information is integrated through the deep learning model, intention and slot identification are realized, accurate task instructions are generated, the accuracy of voice recognition is improved, and user experience is enhanced.

Description

Technical Field

[0001] This invention relates to the field of vehicle technology, and in particular to a method, apparatus, device, and storage medium for identifying task information. Background Technology

[0002] With the development of intelligent vehicle technology, in-vehicle systems are increasingly integrating voice recognition functionality to provide a safer and more convenient interaction method. Users can control navigation, media playback, communication, and other in-vehicle services via voice commands. This interaction method reduces the driver's reliance on physical controls while driving, thereby contributing to improved driving safety. However, to achieve efficient and accurate voice recognition, the system needs to be able to understand and process complex language structures and diverse user intents.

[0003] The current vehicle-mounted task recognition process collects user voice commands through the vehicle's microphone, performs noise reduction and enhancement processing on the collected voice signals, converts the preprocessed voice signals into text information, analyzes the recognized text to determine the user's intent and related parameters, and finally executes the corresponding control commands based on the parsed intent and slot information.

[0004] Despite some progress in existing technologies, significant shortcomings remain. First, speech recognition and text processing tasks are typically performed by independent modules, leading to a significant gap between them and increasing communication and coordination costs. Furthermore, in noisy in-vehicle environments or when faced with voice input containing regional accents, accurate recognition and understanding of user intent is difficult. Therefore, a more efficient and accurate method is urgently needed to improve the quality of speech recognition. Summary of the Invention

[0005] The main objective of this application is to provide a method, apparatus, device, and storage medium for task information recognition, aiming to solve the technical problem of how to improve the accuracy of speech recognition task extraction.

[0006] To achieve the above objectives, this application proposes a task information identification method, the method comprising:

[0007] Obtain voice information;

[0008] The voice information is input into a preset deep learning model to obtain the intent recognition result, vehicle usage intent and slot information. The preset deep learning model includes at least a voice encoder, a predictor and a joint network layer, wherein the voice encoder and the predictor are both connected to the joint network layer.

[0009] Based on the vehicle usage intention and the slot information, combined with the intention recognition result, an extended intention recognition result is obtained;

[0010] Task information is obtained based on the results of the expanded intent recognition.

[0011] In one embodiment, the step of obtaining the intent recognition result, the vehicle usage intent, and the slot information by inputting the voice information into a preset deep learning model includes:

[0012] The speech information is input into a preset deep learning model, and the speech information is converted by the speech encoder to obtain a feature vector;

[0013] The feature vector is input into the predictor to obtain the predicted text information;

[0014] The predicted text information is input into the joint network layer for integration, generating an intent recognition result;

[0015] Based on the intent recognition results, the vehicle usage intent and slot information are obtained.

[0016] In one embodiment, the step of integrating the predicted text information input into the joint network layer to generate an intent recognition result includes:

[0017] The predicted text information is input into the joint network layer;

[0018] The predicted text information is integrated through the joint network layer to perform nonlinear transformation and fusion operations, thereby obtaining the intent recognition result.

[0019] In one embodiment, the step of obtaining an expanded intent recognition result by combining the vehicle usage intent and the slot information with the intent recognition result includes:

[0020] The in-vehicle usage intent and the slot information are expanded to obtain an extended text vocabulary. The in-vehicle usage intent includes at least navigation, media control, air conditioning adjustment and seat control. The slot information includes at least geographical location, song name and contact person.

[0021] The extended text vocabulary is then added to the intent recognition result to obtain the extended intent recognition result.

[0022] In one embodiment, before the step of expanding the extended text vocabulary to the intent recognition result to obtain the expanded intent recognition result, the method includes:

[0023] Based on the extended text vocabulary, the intent slot information is obtained;

[0024] Set the intent slot information as a preset label;

[0025] The step of expanding the extended text vocabulary to the intent recognition result to obtain the expanded intent recognition result includes:

[0026] When the preset tag is detected in the intent recognition result, the intent recognition result is expanded by combining the extended text vocabulary to obtain an expanded intent recognition result.

[0027] In one embodiment, the step of obtaining task information based on the extended intent recognition result includes:

[0028] The results of the expanded intent recognition are analyzed to obtain the user intent and the slot information corresponding to the user intent;

[0029] The task information is obtained by converting the user intent and the slot information.

[0030] In one embodiment, after the step of converting the user intent and the relevant slot information to obtain task information, the method includes:

[0031] Based on the task information, a task instruction is generated and sent to the vehicle-mounted device, so that the vehicle-mounted device receives the task instruction and performs the corresponding operation.

[0032] Furthermore, to achieve the above objectives, this application also proposes a task information identification device, which includes:

[0033] The acquisition module is used to acquire voice information;

[0034] The acquisition module is also used to input the voice information into a preset deep learning model to obtain the intent recognition result, vehicle usage intent and slot information. The preset deep learning model includes at least a voice encoder, a predictor and a joint network layer, wherein the voice encoder and the predictor are both connected to the joint network layer.

[0035] An expansion module is used to obtain an expanded intent recognition result based on the vehicle usage intent, the slot information, and the intent recognition result.

[0036] The analysis module is used to obtain task information based on the expanded intent recognition results.

[0037] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the task information identification method described above.

[0038] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the task information recognition method described above.

[0039] This application acquires voice information, inputs it into a preset deep learning model, and obtains intent recognition results, in-vehicle usage intent, and slot information. Based on the in-vehicle usage intent and slot information, combined with the intent recognition results, an expanded intent recognition result is obtained. Task information is then derived from this expanded intent recognition result. By integrating voice information through a deep learning model, intent and slot recognition are achieved, generating precise task instructions, thus improving the accuracy of voice recognition and enhancing the user experience. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a flowchart illustrating the first embodiment of the task information identification method of this application;

[0042] Figure 2 This is a flowchart illustrating the second embodiment of the task information identification method of this application;

[0043] Figure 3 This is a schematic diagram of the module structure of the task information recognition device according to an embodiment of this application;

[0044] Figure 4 This is a schematic diagram of the device structure of the hardware operating environment involved in the task information identification method in this application embodiment;

[0045] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0046] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0047] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0048] The main solution of this application embodiment is: by acquiring voice information, inputting the voice information into a preset deep learning model, obtaining the intent recognition result, the vehicle usage intent and slot information, and combining the vehicle usage intent and slot information with the intent recognition result to obtain the extended intent recognition result, and obtaining the task information based on the extended intent recognition result.

[0049] Based on this, embodiments of this application provide a task information identification method, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the task information identification method of this application.

[0050] In this embodiment, the task information identification method includes steps S10 to S40:

[0051] Step S10: Obtain voice information.

[0052] It should be noted that the system captures the user's raw voice commands through an in-vehicle microphone array. These commands contain complex semantics and diverse intentions. To improve recognition accuracy, the acquired raw voice signal undergoes a series of preprocessing steps, including noise suppression, echo cancellation, and volume normalization, to ensure the quality of the voice data. Subsequently, the preprocessed voice data is input into a preset deep learning model. This model is a complex neural network, such as a convolutional neural network, a recurrent neural network, a Transducer model, or an attention mechanism model. In this embodiment, the deep learning model adopts a Transducer model.

[0053] Step S20: Input the voice information into the preset deep learning model to obtain the intent recognition result, vehicle usage intent and slot information. The preset deep learning model includes at least a voice encoder, a predictor and a joint network layer, wherein the voice encoder and the predictor are both connected to the joint network layer.

[0054] It's important to note that the system inputs the collected speech information into the Transducer model, which comprises three core components: a speech encoder, a predictor, and a joint network layer. The speech encoder converts the speech signal into a high-dimensional feature representation that can be further processed. These features include not only the physical properties of the speech but also the speaker's identity characteristics and emotional state. The predictor utilizes these high-dimensional features to predict possible text outputs, gradually constructing possible sentence sequences by learning the complex mapping relationship between speech and text. Most importantly, the joint network layer acts as a bridge between the speech encoder and the predictor, integrating their outputs and performing information fusion. Through complex algorithms, such as attention mechanisms or memory networks, the joint network layer analyzes and understands the intent and slot information in the speech to obtain the intent recognition result.

[0055] Furthermore, the speech information is input into a pre-defined deep learning model, where it is transformed by a speech encoder to obtain feature vectors. The speech information is then input into the Transducer model, where it is first fed into the speech encoder for transformation. The speech encoder, as part of the Transducer model, is responsible for converting the input raw speech signal into multiple high-dimensional feature vectors. These feature vectors are not merely simple digitizations of the speech signal; they contain rich information, such as basic speech attributes like pitch, timbre, and prosody, and can even capture the speaker's emotional characteristics, environmental background noise, and subtle changes in pronunciation.

[0056] Further, step S20 specifically includes: inputting the feature vector into the predictor to obtain predicted text information. The feature vector is then input into the predictor, which, as a core component of the Transducer model, begins to generate predicted text information based on the received feature vector. By combining contextual information and symbols predicted based on tone, an output symbol sequence is gradually constructed, and then integrated to obtain the predicted text information. The predictor employs recurrent neural network technology, which means it can effectively handle long-distance dependencies and maintain high prediction accuracy even when facing complex speech input. In addition, the predictor also has a certain degree of flexibility, dynamically adjusting its prediction strategy according to different speech input environments (such as speech speed, accent differences, etc.).

[0057] Further, step S20 specifically includes: inputting the predicted text information into the joint network layer, and performing nonlinear transformation and fusion operations through the joint network layer to obtain the intent recognition result. As an important component of the Transducer model, the joint network layer is responsible for comprehensively processing the feature vectors from the speech encoder and the text information generated by the predictor. Specifically, the predicted text information is input into the joint network layer, and nonlinear transformation and fusion operations are performed through the integration of the predicted text information to obtain the intent recognition result. Through nonlinear transformation and multi-level fusion mechanisms, the joint network can fully utilize speech features and contextual information to generate intent recognition results. This integration method not only ensures that the system's output at each time step accurately reflects the true text of the current speech signal, but also extracts the in-vehicle usage intent and slot information related to the speech segment.

[0058] Step S30: Based on the vehicle usage intention and slot information, combined with the intention recognition result, an extended intention recognition result is obtained.

[0059] It's important to understand that in-vehicle usage intent refers to commands or requests issued by the user via voice. These commands or requests aim to instruct the vehicle to perform a specific function or service, such as navigating to a location, playing specific music, or adjusting the interior temperature. Slot information refers to the specific parameters or data required to complete these commands, such as the destination address, the song title, or the set temperature value. Based on the in-vehicle usage intent and slot information, combined with the intent recognition results, the system generates a more detailed, expanded intent recognition result.

[0060] Further, step S30 specifically includes: expanding the in-vehicle usage intent and slot information to obtain an expanded text vocabulary. Expanding the in-vehicle usage intent and slot information into an expanded text vocabulary allows the system to more comprehensively understand the user's voice commands. The expanded text vocabulary not only includes basic words and phrases but also specifically adds all intent words related to in-vehicle functions, such as navigation, media control, air conditioning adjustment, and seat control. Simultaneously, it also covers various specific slot information that users might mention when using these functions, such as the destination's geographical location, the name of the song they want to play, or the name of a contact.

[0061] Furthermore, step S30 specifically includes: expanding the extended text vocabulary to the intent recognition result to obtain an expanded intent recognition result. This process involves combining the identified user intent with specific slot information. For example, if the user says "navigate to Plaza A," the system will not only recognize the action intent of "navigation" but also extract the geographical entity "Plaza A" as slot information. This information together constitutes a complete expanded intent recognition result. In this way, the system not only understands what the user wants to do but also clarifies the specific operational details that the user wishes to perform.

[0062] Before obtaining the expanded intent recognition result, intent slot information is obtained based on the extended text vocabulary. This intent slot information is then set as preset tags. When a preset tag is detected in the intent recognition result, it is expanded by combining it with the extended text vocabulary to obtain an expanded intent recognition result. When the system detects these preset tags during intent recognition, it automatically triggers a process that combines the information from the extended text vocabulary with the current intent recognition result to generate a more detailed expanded intent recognition result. For example, if the user's voice command is "navigate to the nearest hospital," the system will recognize "navigation" as the intent and "nearest hospital" as the slot information. At this time, the system will check if there are any preset tags related to navigation. Once confirmed, it will use relevant words and information from the extended text vocabulary to supplement and improve this intent recognition result, so that the final expanded intent recognition result not only includes the user's original intent but also more specific operational details, such as the specific destination of the navigation. This approach, which comprehensively considers acoustic features, historical text, and historical slot information, ensures that the system can more accurately grasp the user's true intent during intent recognition, thereby generating more accurate and consistent intent recognition results.

[0063] Step S40: Obtain task information based on the expanded intent recognition result.

[0064] It's important to note that the task information obtained from the expanded intent recognition results refers to the system generating specific, executable operation guidelines by comprehensively analyzing the intent and slot information in the user's voice commands. After obtaining the task information, the system also generates task instructions based on this information and sends them to the in-vehicle device, allowing the device to receive and execute the corresponding operations. For example, when the system recognizes a user saying "Navigate to the nearest hospital," it not only understands the intent of "navigation" but also captures the geographical location slot for "nearest hospital." The system can further refine and convert this into actual task information, such as launching a navigation application, searching for nearby hospital locations, and loading the navigation route onto the in-vehicle display for the driver's reference.

[0065] This embodiment provides a task information recognition method. By acquiring voice information and inputting it into a preset deep learning model, it obtains intent recognition results, in-vehicle usage intent, and slot information. Based on the in-vehicle usage intent and slot information, combined with the intent recognition results, an expanded intent recognition result is obtained. Finally, task information is derived from the expanded intent recognition result. By integrating voice information through a deep learning model, intent and slot recognition are achieved, generating accurate task instructions, improving the accuracy of voice recognition, and enhancing the user experience.

[0066] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 2 The task information identification method step S40 further includes steps S401 to S402:

[0067] Step S401: Analyze the extended intent recognition results to obtain the user intent and the slot information corresponding to the user intent.

[0068] It's important to note that this process involves in-depth analysis of the intent recognition results obtained from deep learning models, especially the joint network layers, to extract the user's true intent and relevant slot information. Enhancing the intent recognition results typically includes richer semantic content and contextual information, enabling the system to more accurately understand user commands. During the analysis, the system employs Natural Language Processing (NLP) techniques, such as semantic role labeling and dependency parsing, to identify the various components of a sentence and their interrelationships. For example, the system can identify "play" as an action intent, "jazz" as a music genre slot, and "volume up" as a command to adjust the volume. Furthermore, the system may utilize machine learning algorithms, such as classifiers or regression models, to predict and validate the accuracy of intents and slots.

[0069] Step S402: Convert the user intent and slot information to obtain task information.

[0070] It's important to understand that after obtaining user intent and relevant slot information, the system can transform the data to generate more precise task instructions. During the transformation process, the system first verifies and calibrates the identified intent and slot information to ensure accuracy. For example, if a user says, "I want to listen to the latest pop songs," the system not only recognizes the intent to "play music," but also extracts the user's specific preferences from "latest" and "popular songs" to generate a playback task, thereby providing a personalized music playback service.

[0071] Through this precise conversion of task information, the in-vehicle system can provide users with more intelligent and personalized services. Users can control various vehicle functions, such as adjusting the air conditioning temperature, making phone calls, and sending messages, simply by using voice commands, greatly improving driving convenience and safety.

[0072] This embodiment analyzes the results of extended intent recognition to obtain user intent and related slot information. Based on the user intent and related slot information, it then converts them into task information. By analyzing the results of extended intent recognition, user intent and slot information are accurately extracted and converted into specific task information, thus improving accuracy.

[0073] This application also provides a task information identification device, please refer to... Figure 3 The device includes:

[0074] Acquisition module 10 is used to acquire voice information.

[0075] The acquisition module 10 is also used to input voice information into a preset deep learning model to obtain intent recognition results, vehicle usage intent and slot information. The preset deep learning model includes at least a voice encoder, a predictor and a joint network layer, wherein the voice encoder and the predictor are both connected to the joint network layer.

[0076] The expansion module 20 is used to obtain an expanded intent recognition result by combining the vehicle usage intent and slot information with the intent recognition result.

[0077] Analysis module 30 is used to obtain task information based on the results of extended intent recognition.

[0078] The task information recognition device provided in this application, employing the task information recognition method described in the above embodiments, can solve the technical problem of how to improve the accuracy of speech recognition task extraction. Compared with the prior art, the beneficial effects of the task information recognition device provided in this application are the same as those of the task information recognition method described in the above embodiments, and other technical features in the task information recognition device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0079] In one embodiment, the acquisition module 10 is further configured to input voice information into a preset deep learning model, convert the voice information through a voice encoder to obtain a feature vector; input the feature vector into a predictor to obtain predicted text information; input the predicted text information into a joint network layer for integration to generate an intent recognition result; and analyze the intent recognition result to obtain the vehicle usage intent and slot information.

[0080] In one embodiment, the acquisition module 10 is further configured to input the predicted text information into the joint network layer; and to perform nonlinear transformation and fusion operations by integrating the predicted text information through the joint network layer to obtain the intent recognition result.

[0081] In one embodiment, the expansion module 20 is further used to expand the vehicle usage intent and slot information to obtain an expanded text vocabulary. The vehicle usage intent includes at least navigation, media control, air conditioning adjustment and seat control, and the slot information includes at least geographical location, song name and contact person. The expanded text vocabulary is then expanded to the intent recognition result to obtain an expanded intent recognition result.

[0082] In one embodiment, the expansion module 20 is further configured to obtain intent slot information based on the expanded text vocabulary and set the intent slot information as a preset tag;

[0083] In one embodiment, the expansion module 20 is further configured to expand the intent recognition result by combining an expanded text vocabulary when a preset label is detected in the intent recognition result, thereby obtaining an expanded intent recognition result.

[0084] In one embodiment, the analysis module 30 is further configured to analyze the extended intent recognition result to obtain the user intent and the slot information corresponding to the user intent; and to convert the user intent and the slot information to obtain task information.

[0085] In one embodiment, the analysis module 30 is further configured to generate task instructions based on task information and send them to the vehicle-mounted device, so that the vehicle-mounted device receives the task instructions and performs corresponding operations.

[0086] This application provides a task information recognition device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the task information recognition method in the first embodiment described above.

[0087] The following is for reference. Figure 4 The diagram illustrates a structural schematic of a task information recognition device suitable for implementing embodiments of this application. The task information recognition device in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 4 The task information recognition device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0088] like Figure 4As shown, the task information recognition device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the task information recognition device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. Communication device 1009 allows the mission information identification device to communicate wirelessly or wiredly with other devices to exchange data. Although mission information identification devices with various systems are shown in the figures, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0089] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0090] The task information recognition device provided in this application, employing the task information recognition method described in the above embodiments, can solve the technical problem of how to improve the accuracy of speech recognition task extraction. Compared with the prior art, the beneficial effects of the task information recognition device provided in this application are the same as those of the task information recognition method described in the above embodiments, and other technical features of this task information recognition device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0091] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0092] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0093] This application provides a computer-readable medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the task information recognition method in the above embodiments.

[0094] The computer-readable medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0095] The aforementioned computer-readable medium may be included in the task information recognition device; or it may exist independently and not be assembled into the task information recognition device.

[0096] The aforementioned computer-readable medium carries one or more programs that, when executed by a task information identification device, enable the task information identification device to write computer program code for performing the operations of this application in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—and conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0097] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0098] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0099] The readable medium provided in this application is a computer-readable medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-described task information recognition method, and can solve the technical problem of how to improve the accuracy of speech recognition task extraction. Compared with the prior art, the beneficial effects of the computer-readable medium provided in this application are the same as the beneficial effects of the task information recognition method provided in the above embodiments, and will not be repeated here.

[0100] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the task information recognition method described above.

[0101] The computer program product provided in this application can solve the technical problem of how to improve the accuracy of speech recognition task extraction. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the task information recognition method provided in the above embodiments, and will not be repeated here.

[0102] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A task information recognition method characterized by comprising: The method comprises: acquiring voice information; inputting the voice information into a preset deep learning model to obtain an intent recognition result, a vehicle use intent, and slot information, wherein the preset deep learning model at least comprises a voice encoder, a predictor, and a joint network layer, the voice encoder and the predictor are connected with the joint network layer, the voice information is inputted into the preset deep learning model, the voice information is converted by the voice encoder to obtain a feature vector, the feature vector is inputted into the predictor to obtain predicted text information, the predicted text information is inputted into the joint network layer, the joint network layer integrates the predicted text information to perform nonlinear change and fusion operation to obtain the intent recognition result, the joint network layer is an important component of a Transducer model, and is responsible for comprehensive processing of the feature vector from the voice encoder and the text information generated by the predictor; combining the vehicle use intent and the slot information with the intent recognition result to obtain an extended intent recognition result; obtaining task information according to the extended intent recognition result.

2. The method of claim 1, wherein, The step of inputting the voice information into a preset deep learning model to obtain an intent recognition result, a vehicle use intent, and slot information comprises: inputting the voice information into the preset deep learning model to convert the voice information by the voice encoder to obtain a feature vector; inputting the feature vector into the predictor to obtain predicted text information; integrating the predicted text information into the joint network layer to generate an intent recognition result; analyzing the intent recognition result to obtain a vehicle use intent and slot information.

3. The method of claim 1, wherein, The step of combining the vehicle use intent and the slot information with the intent recognition result to obtain an extended intent recognition result comprises: extending the vehicle use intent and the slot information to obtain an extended text vocabulary, wherein the vehicle use intent at least comprises navigation, media control, air conditioning adjustment, and seat control, and the slot information at least comprises a geographic location, a song name, and a contact person; extending the extended text vocabulary to the intent recognition result to obtain an extended intent recognition result.

4. The method of claim 3, wherein, Before the step of extending the extended text vocabulary to the intent recognition result to obtain an extended intent recognition result, the method comprises: obtaining intent slot information according to the extended text vocabulary; setting the intent slot information as a preset label; The step of extending the extended text vocabulary to the intent recognition result to obtain an extended intent recognition result comprises: when the preset label is detected in the intent recognition result, the intent recognition result is extended in combination with the extended text vocabulary to obtain an extended intent recognition result.

5. The method of claim 1, wherein, The step of obtaining task information according to the extended intent recognition result comprises: analyzing the extended intent recognition result to obtain a user intent and slot information corresponding to the user intent; converting the user intent and the slot information to obtain task information.

6. The method of claim 5, wherein, After the step of converting the user intent and the slot information to obtain task information, the method comprises: According to the task information, a task instruction is generated and sent to the vehicle-mounted device, so that the vehicle-mounted device receives the task instruction and performs corresponding operations.

7. A task information recognizing apparatus characterized by comprising: The device comprises: An acquisition module is configured to acquire voice information. The acquisition module is further configured to input the voice information into a preset deep learning model to obtain an intent recognition result, a vehicle-mounted use intent, and slot information, wherein the preset deep learning model at least includes a voice encoder, a predictor, and a joint network layer, the voice encoder and the predictor are both connected to the joint network layer, the voice information is input into the preset deep learning model, the voice information is converted by the voice encoder to obtain a feature vector, the feature vector is input into the predictor to obtain predicted text information, the predicted text information is input into the joint network layer, the joint network layer integrates the predicted text information to perform nonlinear change and fusion operations to obtain the intent recognition result, the joint network layer is an important component of a Transducer model and is responsible for comprehensive processing of the feature vector from the voice encoder and the text information generated by the predictor; An expansion module is configured to obtain an expanded intent recognition result according to the vehicle-mounted use intent, the slot information, and the intent recognition result. An analysis module is configured to obtain task information according to the expanded intent recognition result.

8. A task information recognizing apparatus characterized by comprising: The device comprises a memory, a processor, and a task information recognition program stored on the memory and executable on the processor, and the task information recognition program is configured to implement the steps of the task information recognition method according to any one of claims 1 to 6.

9. A storage medium, characterized by The storage medium stores a task information recognition program, and the task information recognition program is executed by the processor to implement the steps of the task information recognition method according to any one of claims 1 to 6.

Images