Method and apparatus for classifying utterance intent in consideration of context around vehicle and driver

Abstract

The invention relates to a method and an apparatus for classifying utterance intent in consideration of the context around a vehicle and a driver. A computer-implemented method for determining an intent of a user's utterance includes obtaining utterance data representing an utterance occurring within a vehicle and contextual information related to the utterance; generating a prompt based on the utterance data and contextual information; obtaining a context-aware sentence from the output of the generative large language model by providing the cue to the generative large language model; the context-aware sentences are provided to an intent classification model to determine the intent of the utterance.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] The present application claims priority to Korean Patent Application No. 10-2023-0120613 filed on September 11, 2023 and Korean Patent Application No. 10-2024-0005376 filed on January 12, 2024, the entire contents of which are incorporated herein for all purposes by this reference. Technical Field

[0003] The present invention relates to a method and apparatus for classifying the intent of an utterance taking into account the context surrounding the vehicle and the driver. Background Art

[0004] The content described below merely provides background information related to the present embodiment and does not constitute related art.

[0005] With the advent of software-defined vehicles, speech recognition is growing in importance.

[0006] Vehicle voice recognition systems enable drivers to interact with their vehicles through voice commands. The ability to perform functions within a vehicle through simple voice commands is more than just a luxury, it is a safety and convenience feature designed to allow drivers to focus on the road ahead.

[0007] Advances in deep learning have enabled speech recognition systems to process natural language. Nevertheless, speech recognition systems still face an ongoing challenge: drivers often use short, truncated, or unclear words when using speech recognition systems, making it difficult to understand the driver's true intent and associate it with specific infotainment or driving-related functions.

[0008] Traditional speech recognition systems rely heavily on single utterances and are insufficient to capture the driver's subtle intentions.

[0009] The information included in this background technology section of the present invention is only used to deepen the understanding of the general background technology of the present invention and should not be regarded as an admission or any form of suggestion that the information constitutes the prior art already known to those skilled in the art. Summary of the invention

[0010] According to an exemplary embodiment of the present invention, a context-rich sentence may be obtained from a user's ambiguous utterance and vehicle context information using a large language model, and the intention of the ambiguous utterance may be determined based on the context-rich sentence.

[0011] Objectives to be achieved as exemplary embodiments of the present invention are not limited to the above-mentioned objectives, and other objectives not mentioned can be clearly understood from the following description by those skilled in the art.

[0012] Various aspects of the present invention are intended to provide a computer-implemented method for determining the intent of a user's utterance, the method comprising: obtaining utterance data representing an utterance occurring in a vehicle and contextual information associated with the utterance; generating prompts based on the utterance data and the contextual information, the prompts comprising a task description, a function catalog, a guided learning example, contextual information, and the utterance data; obtaining context-aware sentences from the output of the generative large language model by providing the prompts to a generative large language model; and providing the context-aware sentences to an intent classification model to determine the intent of the utterance.

[0013] According to another exemplary embodiment of the present invention, the present invention provides a computing device, the computing device comprising at least one processor and a memory operably connected to the at least one processor, wherein the memory stores instructions, the instructions causing the at least one processor to perform operations in response to execution of the instructions by the at least one processor; wherein the operations include: obtaining speech data representing speech occurring in a vehicle and context information related to the speech; generating prompts based on the speech data and the context information, the prompts comprising a task description, a function catalog, a guided learning example, context information and the speech data; obtaining context-aware sentences from the output of the generative large language model by providing the prompts to a generative large language model; and providing the context-aware sentences to an intent classification model to determine the intent of the speech.

[0014] According to another exemplary embodiment of the present invention, the present invention provides a non-volatile computer-readable recording medium storing instructions, which, when executed by a computer, causes the computer to perform the following operations: obtain speech data representing speech occurring in a vehicle and context information related to the speech; generate prompts based on the speech data and the context information, the prompts including a task description, a function catalog, a guided learning example, context information, and the speech data; obtain context-aware sentences from the output of the generative large language model by providing the prompts to a generative large language model; and provide the context-aware sentences to an intent classification model to determine the intent of the speech.

[0015] According to an exemplary embodiment of the present invention, the accuracy of intent classification for utterances can be improved by utilizing a large-scale language model without the need for additional training of the intent classification model.

[0016] According to an exemplary embodiment of the present invention, the usability of a speech recognition function may be improved by accurately determining the intention of an utterance in consideration of not only the utterance but also vehicle context information.

[0017] The effects of the present invention are not limited to the above-mentioned effects, and those skilled in the art will clearly understand other effects not mentioned from the following description.

[0018] The methods and apparatus of the present invention have other features and advantages, which will be apparent from, or will be described in detail in, the accompanying drawings and subsequent detailed descriptions incorporated herein, which together serve to explain the specific principles of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A schematic diagram for determining the intention of a user's utterance in consideration of context according to an exemplary embodiment of the present invention is shown.

[0020] Figure 2A , Figure 2B and Figure 2C is a schematic diagram for explaining a process of obtaining a context-aware sentence according to an exemplary embodiment of the present invention.

[0021] Figure 3 is a flowchart of a method for determining the intent of a user utterance according to an exemplary embodiment of the present invention.

[0022] Figure 4 is a block diagram illustrating an example of a computing device according to an exemplary embodiment of the present invention.

[0023] It is to be understood that the accompanying drawings are not necessarily drawn to scale, and that they present appropriately simplified representations of various features illustrating the basic principles of the present invention. The specific design features of the present invention as included herein (including, for example, specific dimensions, directions, locations, and shapes) will be determined in part by the environment in which they are to be applied and used.

[0024] In the drawings, like reference numbers refer to the same or equivalent parts of the present invention throughout the several figures of the drawing. DETAILED DESCRIPTION

[0025] Reference will now be made in detail to various embodiments of the present invention, examples of which are shown in the accompanying drawings and are described below. Although the present invention will be described in conjunction with exemplary embodiments of the present invention, it should be understood that this description is not intended to limit the present invention to those exemplary embodiments of the present invention. On the contrary, the present invention is intended to cover not only exemplary embodiments of the present invention, but also various alternative embodiments, modified embodiments, equivalent embodiments and other embodiments that may be included within the spirit and scope of the present invention as defined by the appended claims.

[0026] Hereinafter, various exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description, the same reference numerals preferably represent the same elements, although these elements are shown in different drawings. In addition, for the sake of clarity and brevity, when the detailed description of related known components and functions is considered to obscure the subject matter of the present invention, the following description of various exemplary embodiments will omit the detailed description of related known components and functions.

[0027] Various ordinal numbers or letter codes such as first, second, i), ii), a), b), etc. are used only as prefixes to distinguish one component from another, and do not teach or imply the substance, order or sequence of the components. Throughout the specification, when a component "includes" or "comprises" a component, the component is intended to further include other components, rather than excluding other components, unless specifically stated to the contrary. Terms such as "unit", "module", etc. refer to units that process at least one function or operation, and they can be implemented by hardware, software, or a combination thereof.

[0028] The description of the present invention presented below in conjunction with the accompanying drawings is intended to describe exemplary embodiments of the present invention, and is not intended to represent the only embodiment in which the technical idea of ​​the present invention can be practiced.

[0029] As used herein, the term “utterance data” refers to text data converted from a user's voice command through a speech recognition module (ie, a speech-to-text (STT) module).

[0030] As used herein, the term “ambiguous utterance data” refers to utterance data for which the user's intention cannot be determined using only the utterance data because the utterance is brief, incomplete, or unclear.

[0031] Large Language Model (LLM) is an advanced type of language model trained on large amounts of text data using deep learning techniques. LLMs include: causal language models, which generate human-like text by predicting subsequent words based on the context provided by previous words, and masked language models, which predict missing words based on the context provided by previous and subsequent words. This is achieved by using advanced deep learning techniques such as Transformer architecture and attention mechanisms, which enable the model to capture the complex relationship between words and the context in which they are used.

[0032] One of the main advantages of LLMs is their ability to reason and generate knowledge about the human world. This is not because LLMs inherently understand the world or have experience, but because they are trained on large amounts of human-generated text data. Since this text data contains a wide range of human knowledge, culture, and reasoning processes, LLMs can generate outputs that simulate human-like understanding and reasoning by learning patterns in this data.

[0033] Furthermore, another unique feature of LLMs is their ability to perform contextual learning. Unlike traditional machine learning models that require a separate training phase, LLMs can adjust their predictions based on the context provided in a conversation or interaction sequence. Therefore, LLMs are well suited for tasks such as zero-shot learning, one-shot learning, and few-shot learning, where the model is expected to perform based on a single example or a small number of examples, respectively.

[0034] In the context of LLMs, prompt engineering is a key aspect to consider. Prompt engineering refers to crafting initial inputs (i.e., prompts) to guide the model to generate desired outputs. As LLMs become larger and more complex, prompt engineering becomes increasingly important. It has been observed that the same model can generate very different outputs based on prompts, making prompts a key aspect to achieve desired performance. In fact, in the field of prompt engineering, various studies focusing on strategies to elicit the most effective responses from LLMs have gained significant attention recently.

[0035] The present invention relates to a technology that generates prompts by combining ambiguous utterances and vehicle context information, provides the prompts to an LLM to obtain a context-rich sentence as an output, and determines the user's intention based on the context-rich sentence, so that the user's intention can be determined even when ambiguous utterances are input to a vehicle voice recognition system.

[0036] The present invention utilizes the reasoning capabilities of LLM to generate comprehensive and coherent sentences based on in-vehicle specific cues. The present invention provides a model that can replace traditional encoder-based intent classification models (which often encounter difficulties due to short and truncated sentences) by using LLM to generate context-rich sentences (hereinafter referred to as "context-aware sentences"). Accordingly, the intent of ambiguous utterances can be accurately determined.

[0037] The large language model used in the exemplary embodiment of the present invention is a language model trained with a large amount of data, and may refer to a generative large language model that can appropriately perform tasks when only a small number of samples are provided. In other words, the large language model is an autoregressive model, and may refer to a language model configured for reasoning without fine-tuning using methods such as few-sample learning. Compared with existing general language models, the large language model may have more than 10 times the parameters (e.g., more than 100 billion parameters). The large language model used in the exemplary embodiment of the present invention may include, for example, a generative pre-trained Transformer 3 (GPT-3), etc., but is not limited thereto.

[0038] Figure 1A schematic diagram for determining the intention of a user's utterance in consideration of context according to an exemplary embodiment of the present invention is shown. Figure 2A , Figure 2B and Figure 2C is a schematic diagram for explaining a process of obtaining a context-aware sentence according to an exemplary embodiment of the present invention.

[0039] The process of determining the intention of the user's utterance in consideration of the context according to the exemplary embodiment of the present invention can be performed in the server of the vehicle voice recognition system, and the reference Figure 1 and FIG. 2A to FIG. 2C The above contents are described in detail.

[0040] refer to Figure 1 , the prompt module 110 generates structured input (i.e., prompts) for the generative large language model 120. The prompts provide a structured set of instructions to the generative large language model 120, thereby guiding the generative large language model 120 to infer the hidden intentions behind the ambiguous utterances and subsequently generate corresponding context-rich sentences.

[0041] The main components of the prompt include task description 210, function catalog 220, context information 230, guided learning examples 240 and user speech data 250. Personalized prompts can be generated based on user speech data and context information. Examples of prompts are shown in Figure 2A Shown in.

[0042] The task description 210 describes the tasks that the generative large language model 120 can perform. This corresponds to a series of guidelines that show the role of the generative large language model 120 and help the generative large language model 120 understand its mission. The task description 210 can be pre-stored in a database, etc., and the database, etc. can be referenced when the prompt module 110 generates a prompt.

[0043] The function catalog 220 lists the range of in-vehicle functions that can be accessed through the vehicle voice recognition system. This provides a clear idea of ​​the range of commands available to the generative large language model 120 and helps the vehicle voice recognition system to provide the best match to the user's utterance among the various in-vehicle functions. The function catalog 220 can be pre-stored in a database or the like, and the database or the like can be referenced when the prompt module 110 generates a prompt.

[0044] The vehicle speech recognition system can manage multiple intent categories, which can be classified based on functional similarity or functional domains, some examples of functional domains are shown in Table 1. The description of the functional domain is not only used as an intent category, but also can be used to form a functional catalog of input prompts for the generative large language model 120.

[0045] Table 1

[0046]

[0047] The context information 230 provides important information about the situation around the driver and the vehicle. The context information 230 provides important details about the situation or environment around the vehicle and the driver, and helps the generative large language model 120 understand the speech according to the specific situation of the driver. The context information 230 may include vehicle status information. Here, the vehicle status information may include, but is not limited to, the driving information of the vehicle, the operating status of the vehicle-mounted device, various setting information, and information about the user's previous speech. Here, the operating status of the vehicle-mounted device may include, but is not limited to, data about the operation of the vehicle-mounted device, such as the user's navigation operation, radio operation, air conditioning operation, heater operation, and heating / seat operation. The context information 230 collected from the vehicle can be sent to the server of the vehicle speech recognition system using a wireless communication network.

[0048] The guided learning examples 240 help train the generative large language model 120 through a process called few-shot learning. The guided learning examples 240 may include examples that demonstrate the reasoning process for understanding the user's true intent and revealing any hidden intent behind the utterance. The guided learning examples 240 also provide guidelines for constructing appropriate sentences that accurately capture the intended meaning of the utterance. In-context learning (ICL) through guided learning examples 240 can particularly improve the model's proficiency in understanding ambiguous or unclear utterances.

[0049] The guided learning example 240 may include example utterance data, example context-aware sentences, and an example process of inferring the example context-aware sentences from the example utterance data.

[0050] According to another exemplary embodiment of the present invention, the guided learning examples 240 may include example utterance data and example context-aware sentences.

[0051] The user speech data 250 is a text converted from a user voice command by a speech recognition module (i.e., an STT module). The speech recognition module can be set in a vehicle or included in a server of a speech recognition system. When the speech recognition module is set in a vehicle, the speech data can be sent to the server of the speech recognition system using a wireless communication network. In the process of the generative large language model 120 reasoning the context-aware sentences corresponding to the user speech data 250, the above-mentioned task description 210, function catalog 220, context information 230 and guided learning example 240 are used.

[0052] The generated prompts are input into the generative large language model 120 to obtain generated output data including context-aware sentences. This process effectively converts the ambiguous speech data into clear and feasible sentences, which contributes to a more intuitive and effective vehicle speech recognition system. An example of the output data generated by the generative large language model 120 is shown in Figure 2B Output data may be generated in the format of guided learning example 240, such as Figure 2B shown.

[0053] Obtaining context-aware sentences from the output data Since the output data includes context-aware sentences and reasoning processes, this is used to extract only context-aware sentences to be provided to the intent classification model 130 from the output data. Figure 2C Examples of context-aware sentences extracted from the output data are shown.

[0054] The obtained context-aware sentences are provided to the intent classification model 130 to determine the intent of the utterance. An encoder-based intent classification model (e.g., Electra, etc.) can be utilized to ensure robustness in selecting intent from a set of predefined categories. The intent classification model 130 can be trained on a domain-specific dataset related to the vehicle and is specifically designed for the intent classification task. Since the limitations of ambiguous utterances have been properly addressed in the previous process of obtaining the context-aware sentences, the intent classification model 130 can classify intents with minimal additional training. By using context-aware sentences instead of ambiguous utterance data to classify accurate intents, the overall efficiency of the vehicle speech recognition system can be improved.

[0055] Figure 3 1 is a flow chart of a method for determining the intention of a user's utterance according to an exemplary embodiment of the present invention. The method may be executed in a server of a vehicle speech recognition system.

[0056] refer to Figure 3 The method receives speech from a user in a vehicle at one or more input devices (e.g., microphones). The method obtains speech data representing the speech and context information related to the speech (S310). The speech data may refer to text data converted from the user's speech by a speech recognition module (e.g., a speech-to-text module).

[0057] According to another exemplary embodiment of the present invention, a step of determining whether the speech data is ambiguous may be additionally performed in step S310. That is, first, the method obtains speech data and provides the speech data to the intent classification model 130 to determine the intent of the speech. If the intent classification model 130 cannot determine the intent of the speech, the method obtains context information related to the speech.

[0058] The method generates a prompt corresponding to the utterance data and the context information (S320). The prompt includes a task description, a function catalog, a guided learning example, the context information and the utterance data.

[0059] The method provides a prompt to the generative large language model 120 to obtain a context-aware sentence from the output of the generative large language model (S330).

[0060] The method provides the context-aware sentence to an intent classification model to determine the intent of the utterance based on an output of the intent classification model (S340).

[0061] The method may generate response data in response to determining the intent of the utterance, the response data including the intent of the utterance and an in-vehicle function corresponding to the intent. The response data may be at least one of audio data, image data, or text data. The method may output the response data to the user at one or more output devices (e.g., a speaker, a display), and execute the corresponding in-vehicle function based on a confirmation response from the user.

[0062] Examples of where the technology disclosed in this specification can be used include: cases where the object of the user's utterance is unclear, cases of indirect quotations, cases of habitually expressed utterances, and the like. For example, if the user says "turn it off", the air conditioner in the vehicle is on, the gear is in D, the driver's window is closed, and the radio is off, the user's intention can be predicted by reconstructing a context-aware sentence of "turn off the air conditioner". For example, if the user says "turn it on", the air conditioner in the vehicle is on, the gear is in D, the driver's window is closed, and the radio is off, the user's intention can be predicted by reconstructing a context-aware sentence of "turn on the radio". For example, FIG. 2A to FIG. 2C As shown, when the user says "I want to eat pizza", the context-aware sentence of "navigate to the nearest restaurant" can be reconstructed by referring to the function catalog 220, context information 230 and guided learning examples 240 to predict the user's intention.

[0063] Figure 4 is a block diagram showing an example of a computing device according to an exemplary embodiment of the present invention. The method for determining the intention of a user's speech according to various exemplary embodiments of the present invention may be performed by Figure 4 The computing device 400 shown is implemented.

[0064] like Figure 4As shown, computing device 400 may include at least one processor 410, memory 420, network interface 430, and input / output interface 440. Bus 450 provides a mechanism for enabling the components of computing device 400 to communicate with each other as intended. Although bus 450 is schematically shown as a single bus, alternative embodiments may use multiple buses.

[0065] The processor 410 may be configured to process instructions of a computer program by performing basic arithmetic, logic, and input / output operations. Instructions may be provided to the processor 410 via the memory 420 or the network interface 430. For example, the processor 410 may be configured to execute received instructions according to a program code stored in a recording device such as the memory 420.

[0066] The memory 420 is a computer-readable recording medium and may include a random access memory (RAM) and a permanent mass storage device such as a read-only memory (ROM) or a disk drive. Here, a permanent mass storage device such as a disk drive may be included in the computing device 400 as a separate permanent storage device separated from the memory 420.

[0067] In addition, an operating system and at least one program code may be stored in the memory 420. Such software components may be loaded into the memory 420 from a computer-readable recording medium separate from the memory 420. Such separate computer-readable recording media may include computer-readable recording media such as floppy disk drives, magnetic disks, magnetic tapes, DVD / CD-ROM drives, and memory cards. In another exemplary embodiment of the present invention, the software components may be loaded into the memory 420 via the network interface 430 instead of the computer-readable recording medium. For example, the software components may be loaded into the memory 420 of the computing device 400 based on a computer program provided by a file received via the network interface 430.

[0068] The network interface 430 may provide a function for the computing device 400 to communicate with other external devices (e.g., a terminal including a voice recognition module provided in a vehicle, etc.) through a wired or wireless communication network. For example, a request, instruction, data, file, etc. generated by the processor 410 of the computing device 400 according to a program code stored in a recording device such as the memory 420 may be sent to other external devices through a wired or wireless communication network under the control of the network interface 430. Conversely, signals, instructions, data, files, etc. may be sent from other external devices to the computing device 400 through the network interface 430 of the computing device 400 via a wired or wireless communication network. Signals, instructions, data, etc. received through the network interface 430 may be sent to the processor 410 or the memory 420, and files, etc. may be stored in a storage medium (the aforementioned permanent storage device), which may be additionally included in the computing device 400.

[0069] The input / output interface 440 may be a device for engaging with an input / output device. For example, the input device may include devices such as a microphone, a keyboard, and a mouse, and the output device may include devices such as a display and a speaker. As an exemplary embodiment of the present invention, the input / output interface 440 may be a device for engaging with a device (e.g., a touch screen) that integrates input and output functions. The input / output device may be integrated with the computing device 400.

[0070] In addition, in other exemplary embodiments of the present invention, the computing device 400 may include a Figure 4 The components shown may be fewer or more components. For example, the computing device 400 may be implemented to include at least some of the above-mentioned input / output devices, or may further include other components such as a transceiver, a database, and the like.

[0071] The device or method according to the exemplary embodiment of the present invention may include a corresponding component arranged to be implemented as hardware or software or a combination of hardware and software. In addition, each component may be functionally implemented by software, and a microprocessor may perform functions by the software for each component when implemented.

[0072] Various illustrative embodiments of the systems and methods described herein can be implemented by digital electronic circuits, integrated circuits, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include those implemented in one or more computer programs executable on a programmable system. A programmable system includes at least one programmable processor, the at least one programmable processor being connected to receive data and instructions from a storage system, at least one input device, and at least one output device, and to send data and instructions to a storage system, at least one input device, and at least one output device, wherein the programmable processor may be a dedicated processor or a processor. A computer program (also referred to as a program, software, software application, or code) includes instructions for a programmable processor and is stored in a "computer-readable recording medium".

[0073] Computer-readable recording media include any type of recording device on which data that can be read by a computer system can be recorded. Examples of computer-readable recording media include non-volatile or non-temporary media such as ROM, CD-ROM, magnetic tape, floppy disk, memory card, hard disk, optical / magnetic disk, storage device, etc. Computer-readable recording media may further include volatile media such as data transmission media. In addition, computer-readable recording media may be distributed in computer systems connected via a network, where computer-readable codes may be stored and executed in a distributed manner.

[0074] Although the steps in the various flowcharts are described as being performed sequentially, they only illustrate the technical ideas of various exemplary embodiments of the present invention. Therefore, those skilled in the art, including those skilled in the relevant art, can perform the steps by changing the order described in the corresponding flowchart or by performing two or more of the steps in parallel, and therefore the steps in the corresponding flowchart are not limited to the time sequence shown.

[0075] In various exemplary embodiments of the present invention, the memory and the processor may be provided as one chip, or provided as separate chips.

[0076] In various exemplary embodiments of the present invention, the scope of the present invention includes software or machine-executable commands (e.g., operating systems, applications, firmware, programs, etc.) for enabling operations of methods according to various embodiments to be performed on a device or computer, and non-transitory computer-readable media including such software or commands stored thereon and executable on a device or computer.

[0077] In various exemplary embodiments of the present invention, the control device may be implemented in the form of hardware or software, or may be implemented as a combination of hardware and software.

[0078] Furthermore, terms such as “unit”, “module” and the like included in the specification refer to a unit for processing at least one function or operation, which may be implemented by hardware, software or a combination thereof.

[0079] In an exemplary embodiment of the present invention, a vehicle may be referred to as being based on a concept including various vehicles. In some cases, a vehicle may be interpreted as being based on a concept including not only various land vehicles such as cars, motorcycles, trucks, and buses traveling on roads, but also various vehicles such as airplanes, drones, ships, etc.

[0080] For ease of explanation and accurate definition of the appended claims, the terms "upper", "lower", "inner", "outer", "above", "below", "upward", "downward", "front", "rear", "back", "inside", "outside", "inward", "outward", "inner", "external", "inside", "outside", "forward" and "rearward" are used to describe the features of the exemplary embodiments with reference to the positions of such features as shown in the accompanying drawings. It will be further understood that the term "connect" or its derivatives refer to both direct and indirect connections.

[0081] The term "and / or" may include a combination of a plurality of related listed items or any one of a plurality of related listed items. For example, "A and / or B" includes three cases, such as "A", "B" and "A and B".

[0082] In an exemplary embodiment of the present invention, “at least one of A and B” may refer to “at least one of A or B” or “at least one of a combination of at least one of A and B”. Furthermore, “one or more of A and B” may refer to “one or more of A or B” or “one or more of a combination of one or more of A and B”.

[0083] In this specification, a singular expression includes a plural expression unless the context clearly indicates otherwise.

[0084] In exemplary embodiments of the present invention, it should be understood that terms such as “include” or “have” are intended to indicate the presence of features, values, steps, operations, elements, parts, or a combination thereof described in the specification, and do not exclude the possibility of adding or existing one or more other features, values, steps, operations, elements, parts, or a combination thereof.

[0085] According to the exemplary embodiments of the present invention, components may be combined with each other to be implemented as one, or some components may be omitted.

[0086] In the following, the fact that parts of hardware are operably connected may include that a direct and / or indirect connection between the parts of hardware is established by wired and / or wireless means.

[0087] For the purpose of illustration and description, the foregoing description of specific exemplary embodiments of the present invention has been presented. The foregoing description is not intended to be exhaustive or to limit the present invention to the precise form disclosed, and it is apparent that many modifications and variations can be made based on the above disclosure. The exemplary embodiments are selected and described in order to explain certain principles of the present invention and their practical applications, so that other persons skilled in the art can realize and utilize various exemplary embodiments of the present invention and their different alternative embodiments and modified embodiments. The scope of the present invention is intended to be limited by the appended claims and their equivalents.

Claims

1. A computer-implemented method for determining the intent of a user utterance, the method comprising: obtaining, by a processor, speech data representing speech occurring in a vehicle and context information associated with the speech; generating, by a processor, prompts based on the utterance data and the context information, the prompts comprising a task description, a function catalog, a guided learning example, the context information, and the utterance data; Obtaining, by a processor, a context-aware sentence from an output of the generative large language model by providing the prompt to the generative large language model; The context-aware sentence is provided by a processor to an intent classification model to determine the intent of the utterance.

2. The method according to claim 1, wherein: Obtaining discourse data representing discourse and context information associated with the discourse includes: Obtaining discourse data; Feed the utterance data to the intent classification model to determine the intent of the utterance; In response to being unable to determine the intent of the utterance, contextual information associated with the utterance is obtained.

3. The method according to claim 1, wherein: The context information includes status information of the vehicle.

4. The method according to claim 1, wherein: The function catalog includes at least one in-vehicle function accessible via a vehicle voice recognition system.

5. The method according to claim 1, wherein: The guided learning example includes: example discourse data, example context-aware sentences, and an example process of inferring example context-aware sentences from the example discourse data.

6. The method according to claim 1, wherein: The guided learning examples include example discourse data and example context-aware sentences.

7. A computing device, comprising: at least one processor; and a memory operatively connected to the at least one processor; wherein the memory stores instructions for causing at least one processor to perform operations in response to the instructions being executed by the at least one processor; The operations include: obtaining speech data representing speech occurring in the vehicle and context information associated with the speech; generating a prompt based on the utterance data and the context information, the prompt comprising a task description, a function catalog, a guided learning example, the context information and the utterance data; Obtaining a context-aware sentence from an output of the generative large language model by providing the prompt to the generative large language model; The context-aware sentences are fed to the intent classification model to determine the intent of the utterance.

8. The computing device according to claim 7, wherein: Obtaining discourse data representing discourse and context information associated with the discourse includes: Obtaining discourse data; Feed the utterance data to the intent classification model to determine the intent of the utterance; Obtain contextual information related to an utterance when the intent of the utterance cannot be determined.

9. The computing device according to claim 7, wherein: The context information includes status information of the vehicle.

10. The computing device according to claim 7, wherein: The function catalog includes at least one in-vehicle function accessible via a vehicle voice recognition system.

11. The computing device according to claim 7, wherein: The guided learning example includes: example discourse data, example context-aware sentences, and an example process of inferring example context-aware sentences from the example discourse data.

12. The computing device according to claim 7, wherein: The guided learning examples include example discourse data and example context-aware sentences.

13. A non-transitory computer-readable recording medium storing instructions which, when executed by a computer including a processor, cause the computer to perform the following operations: obtaining speech data representing speech occurring in the vehicle and context information associated with the speech; generating a prompt based on the utterance data and the context information, the prompt comprising a task description, a function catalog, a guided learning example, the context information and the utterance data; Obtaining a context-aware sentence from an output of the generative large language model by providing the prompt to the generative large language model; The context-aware sentences are fed to the intent classification model to determine the intent of the utterance.

14. The non-volatile computer-readable recording medium according to claim 13, wherein: Obtaining discourse data representing discourse and context information associated with the discourse includes: Obtaining discourse data; providing the utterance data to an intent classification model to determine the intent of the utterance; In response to being unable to determine the intent of the utterance, contextual information associated with the utterance is obtained.

15. The non-volatile computer-readable recording medium according to claim 13, wherein: The context information includes status information of the vehicle.

16. The non-volatile computer-readable recording medium according to claim 13, wherein: The function catalog includes at least one in-vehicle function accessible via a vehicle voice recognition system.

17. The non-volatile computer-readable recording medium according to claim 13, wherein: The guided learning example includes: example discourse data, example context-aware sentences, and an example process of inferring example context-aware sentences from the example discourse data.

18. The non-volatile computer-readable recording medium according to claim 13, wherein: The guided learning examples include example discourse data and example context-aware sentences.

Images