Computer-readable data carrier, vehicle, apparatus, computer program and methods for adapting a user interface of a vehicle

Abstract

The present disclosure relates to a computer-readable data carrier, a vehicle, an apparatus, a computer program, and methods for adapting a user interface of a vehicle. In particular, the concept proposed herein relates to speech-based user interface adaption using artificial intelligence. The method comprises obtaining information on a voice input of a user, determining configuration data from the voice input to adapt the user interface using a machine-learning-based model, and adapting the user interface based on the configuration data.

Description

[0001] Description

[0002] Computer-readable data carrier, vehicle, apparatus, computer program and methods for adapting a user interface of a vehicle

[0003] The present disclosure relates to a computer-readable data carrier, a vehicle, an apparatus, a computer program, and methods for adapting a user interface of a vehicle. In particular, the concept proposed herein relates to speech-based user interface adaption using artificial intelligence.

[0004] A user interface (III), also referred to herein as “human-machine interface” (HMI), is a medium through which humans interact with computers, systems, or devices. It includes elements which users can manipulate to control or interact with these systems, such as e.g., screens, buttons, icons, menus, and voice commands. In general, the main objective of a III is to facilitate user-friendly and efficient interaction, thereby enhancing the user experience.

[0005] User interfaces are applied across a wide range of fields. They are integral to computers and mobile devices, where they support operating systems, applications, and websites. They are also crucial in consumer electronics like e.g., TVs, home appliances, and gaming consoles, as well as in automotive systems for e.g., managing infotainment, navigation, and driver assistance features. In industrial contexts, Ills are essential for e.g., operating machinery, control panels, and robotics. Additionally, they are used in smart home devices such as e.g., thermostats, lighting controls, and the like.

[0006] In automotive applications, user interfaces present specific challenges. One major issue is distraction; complex interfaces can divert the driver's attention from the road, increasing the risk of accidents. Usability problems arise when poorly designed controls confuse drivers, leading to operational errors. Inconsistent Uls across different models or systems can frustrate users. Accessibility is another concern, as interfaces must be usable by a diverse range of people, including those with disabilities or limited technical skills. These challenges necessitate careful design to ensure that automotive Ills are both safe and user-friendly.

[0007] Several approaches provide static user interfaces. Particularly in automotive applications, static user interfaces that cannot adapt to user preferences present significant problems. Firstly, they often fail to accommodate the diverse needs and preferences of different drivers. For instance, a non-customizable interface might not display certain information important to some drivers and / or it may display information such that it is difficult to perceive or understand for some drivers.

[0008] Static interfaces also struggle to address different situational needs; what works for a casual drive may not be suitable for navigation during heavy traffic or when the driver requires quick access to critical functions. Moreover, such interfaces can become outdated quickly, unable to integrate new technologies or improvements in user experience, thus, lagging behind evolving driver expectations. This rigidity can result in a cluttered or overly simplified layout that either overwhelms or under-serves the user, impacting both usability and safety. Additionally, a lack of adaptability can lead to accessibility issues, as the static design might not be easily usable by individuals with disabilities or specific needs. Overall, static user interfaces can significantly hinder the effectiveness, enjoyment, and safety of automotive systems by failing to provide a personalized and context-aware user experience.

[0009] Interface adaptation based on predefined commands can be problematic in automotive applications due to its limited flexibility and potential safety risks. Such systems rely on specific commands or gestures, which can be unintuitive or difficult for some users to remember or execute correctly, thereby complicating the interaction and diminishing user experience. This rigidity may frustrate users who expect more natural and intuitive interactions, leading to increased cognitive load and distraction as they struggle to recall or properly use these commands while driving. The inability to intuitively adapt to a driver’s natural behavior or changing situational demands can result in delayed or inappropriate responses from the system, potentially creating hazardous scenarios. For instance, if a voice command is misinterpreted or requires repeated attempts to be recognized, the driver’s attention may be diverted from the road, elevating the risk of accidents. Additionally, in emergency situations where quick and accurate interaction is crucial, reliance on predefined commands can slow down critical decision-making and responses. Thus, the reliance on rigid, predefined commands can compromise both the user experience and driving safety by failing to provide seamless, intuitive, and adaptable interaction.

[0010] Hence, there may be a demand for an improved concept for an HMI. This demand may be satisfied by the subject-matter of the appended claims. Exemplary / optional embodiments are disclosed by the appended dependent claims.

[0011] In particular, the proposed approach aims to alleviate drawbacks of static configuration methods by providing a dynamic, Al-based system capable of interpreting user-spoken commands to adapt the HMI in real-time across multiple infotainment applications. It seeks to streamline and enhance user experience by enabling adaptable configurations through natural language interactions, thereby reducing the attention required for interacting with the HMI.

[0012] Embodiments provide a method for adapting a user interface. The method comprises obtaining information on a voice input of a user, determining configuration data from the voice input to adapt the user interface using a machine-learning based model, and adapting the user interface based on the configuration data. In this way, manual interaction with the HMI, e.g., using a touchscreen and / or buttons may be reduced or, ideally, replaced by the proposed speech-based adaption method. So, a user may need to spend less attention for interacting with the HMI and more attention to more important application-specific things.

[0013] A skilled person having benefit from the present disclosure will appreciate that the proposed concept may be applied in various applications including vehicles. Accordingly, the user interface may correspond to a user interface for a vehicle. In practice, the user interface may correspond to or comprise an infotainment system for a vehicle. In such applications, the proposed concept allows to spend less attention to the HMI in more attention to the traffic environment and navigating the vehicle. So, the proposed concept leads to safer driving behavior in automotive applications / vehicles.

[0014] The model allows to understand the semantic of the voice input. In this way, instructions by other voice input are not limited to only predefined instructions. This particularly allows a more intuitive speech-based recognition also of subtle and implicit commands and a more detailed and / or diverse / versatile input. In other words, the model provides a larger input space for the adaption of the user interface.

[0015] As well, the model provides a larger output space. For this, the model may be aware of any options to adapt the user interface. In practice, the model may be configured to program / configure the user interface via an application programming interface. In this way, the model may be able to freely (re-) program the user interface according to the voice input, as a software developer may be able to do. So, the proposed concept allows for a more individual user interface adaption. In this way, and information supply to the user may be individualized to user-specific preferences for a better communication of information to the user. In vehicles, this supplies a driver with information more individually and, therefore, helps driving more safely. So, in automotive applications, the proposed concept may lead to a higher level of safety. The same applies for other applications such as e.g., for user interfaces for (manufacturing) machines.

[0016] In practice, the machine-learning-based model may be configured to generate the configuration data such that it is indicative of one or more adjustment options of the user interface. In some embodiments, the model may be configured to interact with the user interface via an application programming interface of the user interface.

[0017] The configuration data may comprise one or more configuration parameters for a configuration of the user interface, and adapting the user interface may comprise applying the configuration parameters. The configuration parameters, e.g., determine the interaction between the user interface and the user. In some examples, e.g., the configuration parameters determine how the user interacts with the user interface visually and / or acoustically. For example, configuration parameters determine an appearance of a dashboard and / or icons displayed to the user.

[0018] Additionally, or alternatively, the configuration parameters may determine an acoustic output of the user interface, e.g., a voice of a speech output.

[0019] In practice, the user interface may be configured to execute one or more software applications, and adapting the user interface may comprise adapting one or more settings and / or functions of the software applications. The software applications, e.g., include a software application for playing music and adapting one or more settings and / or functions of the software applications may comprise adjusting the volume of played music.

[0020] In some embodiments determining the configuration data comprises converting the voice input to text using a speech-to-text module and determining the configuration data based on the text. A skilled person will appreciate that, for this, various speech-to-text modules for speech recognition may be applied.

[0021] In some embodiments, the method further comprises providing a response indicative of the adaption of the user interface. The response, e.g., comprises a visual or acoustic output confirming and / or summarizing the adaption. In practice, the response may include a visual, acoustic, and / or tactile notification.

[0022] In some embodiments, the response may be a speech-based.

[0023] The response, for example, includes an audio and / or text output to the user. Providing the response may comprise obtaining a text form of the response from the machine-learning-based model and converting the text form of the response into an audio signal using a text-to-speech module.

[0024] Further embodiments of the proposed concept provide a training method for training the machine-learning-based model for adapting a user interface of a vehicle. The training method comprises training the machine-learning-based model to determine a configuration of the user interface based on a user input indicative of a configuration data for adapting the user interface.

[0025] In some embodiments, the machine-learning-based model is trained based on training data including one or more exemplary user input samples and ground truth of respective configuration data for the exemplary user input samples.

[0026] Further embodiments provide a computer program comprising instructions which, when the computer program is executed by a computer, cause the computer to carry out an embodiment of any one of the proposed methods.

[0027] Still further embodiments provide an apparatus comprising one or more interfaces for communication and a data processing circuit configured to execute an embodiment of any one of the proposed methods.

[0028] Other embodiments provide a computer-readable data carrier having stored thereon the proposed computer program.

[0029] Further, embodiments are now described with reference to the attached drawings. It should be noted that the embodiments illustrated by the referenced drawings show merely optional embodiments as an example and that the scope of the present disclosure is by no means limited to the embodiments presented:

[0030] Brief description of the drawings Fig. 1 shows a flow chart schematically illustrating an embodiment of a method for adapting a user interface;

[0031] Fig. 2 schematically shows an exemplary embodiment and use case;

[0032] Fig. 3 schematically shows another embodiment and use case; and

[0033] Fig. 4 shows a block diagram schematically illustrating an embodiment of an apparatus according to the proposed approach.

[0034] Embodiments of the present disclosure are based on the finding that artificial intelligence may be able to serve a user like a personal virtual software developer which is able to understand and implement language-based change requests in detail to customize a user interface.

[0035] Further aspects and details are described below with reference to Fig. 1 . It is noted that, even though embodiments may be described only in the context of automotive applications, features and aspects of embodiments may also apply to non-automotive applications.

[0036] Fig. 1 shows a flow chart schematically illustrating an embodiment of a method 100 for adapting a user interface.

[0037] As can be seen from the flow chart, method 100 comprises obtaining 110 information on a voice input of a user. For this, a microphone or any other type of sound recorder can be used to record the voice and / or sound of a user. The voice input may comprise any kind of verbal expressions of a user, e.g., commands, questions, expressions of individual needs or wishes, and / or the like. The microphone / audio recorder can provide an audio signal of the recorded voice input. The information on the voice input may represent a content of the voice input. Further, method 100 comprises determining 120 configuration data from the (information on the) voice input to adapt the user interface using a machine-learning-based model (also referred to herein as “(the) model”).

[0038] For this, the machine-learning-based model may be specifically trained to understand the content of the voice input and translate the content into corresponding adaptions of the user interface, as laid out in more detail later.

[0039] In practice, the model may correspond to or include a (large) language model configured to understand the voice input semantically as well as how a configuration of the user interface may be adjusted or set to adapt the user interface in accordance with the voice input.

[0040] In embodiments, the model may be configured to process text data. Accordingly, the information on the voice input may include text indicating the content of the voice input. For this, the audio signal may be processed using a speech-to-text (STT) module. The STT module may be implemented separate from the model.

[0041] Alternatively, in some embodiments, the STT module may be included in the model.

[0042] The configuration data, e.g., comprise or correspond to one or more parameters of the user interface’s configuration. In particular, the configuration data may determine what kind of information and how it is presented to the user. In embodiments, configuration data for example determines an appearance of information presented to a user visually and / or acoustically. In automotive applications, the configuration data, e.g., determines what kind of and / or how information on the vehicle (e.g., speed, acceleration, temperature, engine speed, torque, battery level, oil level, warnings, notifications, and / or the like) is presented to the user. In practice, configuration data, e.g., determines which widgets (displaying different information) are presented as well as their position and / or size on a screen of the user interface.

[0043] Further, the method 100 comprises adapting 130 the user interface based on the configuration data. In doing so, e.g., previous configuration data of the user interface is changed to or replaced by the determined (new) configuration data for adapting 130 the user interface.

[0044] Classic approaches for static user interfaces rely on predetermined user commands and only allow a few predefined options to adapt the user interface. They, e.g., offer a few predefined general layouts of a visual presentation of information. In contrast, the proposed model may be configured to exploit several or all available options to freely re-program the user interface’s configuration with respect to details in the voice input of the user.

[0045] At the same time, the proposed approach provides a hands-free customization of the user interface. So, the proposed solution leads to less user distraction and more attention of a user to application-related things.

[0046] Further details and aspects are described below in more detail with reference to exemplary embodiments schematically illustrated by Figs 2 and 3.

[0047] Fig. 2 schematically shows a use case of the proposed concept.

[0048] The use case includes a user 210 expressing a spoken user request (see step 1 ). The spoken user request is converted into an electric audio signal and fed into a speech-to-text (“STT”) module 220 which is configured to convert the electric audio signal into text representing or corresponding to the spoken user request. The text is provided to an Al service 230. The Al service 230 provides the text of the user request (e.g., as prompt) to an Al backend 240 (machine-learning-based model). So, the Al service 230 may be understood as an interface between an application of the proposed concept and the Al backend 240. In examples, the Al service 230 may correspond to or comprise an application programming interface (API).

[0049] In practice, the application, e.g., is an automotive application (e.g., a vehicle) and the Al backend 240 may be implemented on an external server separate from the vehicle. In embodiments, for this, the application, and the Al backend 240 may be configured to communicate using a wireless communication technology (particularly in mobile applications). Alternatively, they may communicate using a wired connection.

[0050] The Al backend 240 is trained to understand the user request and to provide a user interface configuration for implementing the user request. So, in other words, the Al backend 240 is configured to translate the user request into respective configuration data in accordance with the user request. In examples, the Al backend 240 provides configuration data including one or more configuration files 260 (also referred to herein as “config files”) to the Al service 230 of the application. The config files or each of them may comprise one or more configuration parameters for a configuration of the user interface. In some examples, the configuration data may include software resources (e.g., code and / or data) for configuring the user interface via an application programming interface (API) of the user interface.

[0051] So, the configuration data may be also understood as instructions for adapting the user interface (e.g., via the API.)

[0052] The Al service 230, then, performs necessary expansions, enrichments, and validations, ensuring the configuration files align with the application’s requirements, if necessary. Then, the Al service 230 provides the processed (“expanded”) configuration data to one or more software applications 270 (see step 4) running on the user interface to adapt the user interface accordingly to adapt the user interface. In doing so, e.g., one or more settings and / or functions of the software applications are adapted. For this the configuration parameters may be applied. To this end, e.g., previous configuration parameters may be replaced by or changed to the Al-generated configuration parameters.

[0053] In implementations, these validated configurations (processed / expanded configuration data) are implemented in real-time, affecting the software applications. In this way, the software applications dynamically adjust their functionalities and update their User Interface / User Experience (lll / UX) according to the new config files, thereby fulfilling the user's request for adaptation. This proposed approach vastly improves user interaction and system adaptability within the automotive infotainment domain, ushering in an era of intuitive, voice-driven customization, and real-time adaptability.

[0054] The infusion of Al technology not only streamlines the process of user-system interaction but also enhances the adaptability and responsiveness of the system, providing a significant advantage over traditional manual configuration methods prevalent in the field.

[0055] Advantages include streamlined user interaction, real-time adaptability, improved user experience through natural language commands, reduction in manual configuration efforts, and enhanced efficiency across diverse infotainment applications.

[0056] Beyond automotive applications, this invention's concepts could be employed in smart home systems, personal electronic devices, or any domain requiring adaptable user interfaces via voice commands and Al-driven configurations.

[0057] Optionally, a history of one or more (previous) user requests may be provided to the Al backend 240 for a contextual understanding / interpretation of the latest user request in context of the request history and, thus, for even more precise adaptions of the user interface.

[0058] Additionally, the Al backend 240 may provide feedback indicating the adaption of the user interface to the user. In practice, e.g., the Al backend 240 provides text indicating the feedback. For a spoken response or audible response, the text for the feedback is, then, converted into an audio signal and played (see step 5) to the user 210 via one or more speakers of the user interface. In this way, the user 210 is informed / updated about the adaptions made to user interface.

[0059] For this, the application employs a text-to-speech (TTS) molecule to convert the Al-generated feedback from text into speech, presenting it as a response to the user's initial inquiry. This cohesive interaction engenders an experience akin to conversing with an intelligent assistant capable of adaptively reconfiguring the system as per user specifications.

[0060] In practice, the Al backend 240 may deploy an appropriately (pre-) trained foundation model. The Al backend 240, for example, deploys a robust generative Al system, akin to OpenAI’s GPT models (“ChatGPT”). This kind of Al possesses the capability to comprehend and interpret the user requests articulated in natural language, while also understanding the contextual nuances and the array of configuration options available within the system.

[0061] For training the Al backend 240, e.g., the Al backend 240 is provided with training data including one or more exemplary user input samples and ground truth of respective configuration data for the exemplary user input samples and / or any other type of information indicating available options to adapt the configuration of the user interface as well as samples of effects of changes to the configuration. In practice, prompt engineering may be used for this purpose. So, respective prompts 250 including said information on the available adaption options and their effects may be provided to the Al backend 240 for training. In examples, information on the available adaption options may include a scheme of the user interface’s configuration. In this way, the Al backend 240 is pre-trained to understand contextual user requests, made aware of a spectrum of configuration options of the user interface, and configured to create tailored config files for targeted software applications based on the user request.

[0062] Examples of the Al backend 240 (machine-learning-based model) may not only comprise foundation models or transformer networks but may alternatively or also correspond to or include other types of neural networks including Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTM) networks, Gated Recurrent Units (GRUs), and Convolutional Neural Networks (CNNs). Appropriate training techniques for configuring the Al backend 240 (machine-learning-based model) may (alternatively) include supervised learning with large, annotated datasets, transfer learning from pre-trained models, and unsupervised learning through self-supervised approaches like masked language modeling. Fine-tuning on task-specific datasets further enhances performance, and techniques such as data augmentation, regularization, and gradient clipping are often used to improve model robustness and generalization.

[0063] The software applications may comprise a suite of one or more customizable apps within the user interface, allowing runtime configurations through textual input (here: config files). Finally, it integrates a speech-to-text and text-to-speech processor responsible for the seamless translation of user inquiries into textual data for the Al to process, and subsequently converts the Al-generated feedback text into speech for user response.

[0064] The proposed method allows that the user interacts with the system through natural language, expressing desired adaptations. The speech-to-text module converts these spoken requests into text, transferring this textual data directly to the generative Al via a dedicated API. These generated config files, along with pertinent feedback text, are relayed back to the application and to the user.

[0065] As indicated, the user 210 may express the spoken user request “I want to control music”. The Al backend 240 recognizes from such user request that the user 210 wants to have an option to control different properties of the music played (in the vehicle). From training data indicative of a configuration scheme of the user interface, the Al backend 240 is, e.g., aware of an option to display (on a touchscreen) a widget or any other type of graphic object which allows the user to control e.g., the volume, the sound, the song played, and / or the radio channel. Also, properties of graphic objects, e.g., their size, position, color, and the like may be adapted.

[0066] Likewise, the user 210 can make safety-critical user interface adaptions, e.g., impacting the presentation of safety-critical or safety-relevant information (e.g., speed, acceleration, temperature, engine speed, torque, battery level, oil level, warnings, notifications, and / or the like). For example, the user request relates to a presentation or communication of safety-critical or safety-relevant information. In this way, the user interface may be customized to user preferences leading to a (more precisely) personalized presentation of safety-critical information for a higher level of safety while reducing the distraction of the user for adapting the user interface.

[0067] Further details are described below with reference to Fig. 3 visualizing an exemplary application of the proposed approach for a user interface (e.g., an infotainment system) for a vehicle.

[0068] As shown in Fig. 3, the Al service 230 may include a function or module for chaining / dispatching prompts from user requests which are provided to the Al backend and a function or module for chaining / dispatching the configuration data (also referred to herein as “state updates”). The configuration data is provided to an SDV (software-defined vehicle) platform 370 for adapting the user interface.

[0069] The skilled person having benefit from the present disclosure will appreciate that the proposed approach may be also applied for controlling or adapting any (other) function of the vehicle. In practice, e.g., the proposed may be similarly applied for opening or closing a window of the vehicle and / or adjusting an air condition. The SDV platform 370, e.g., my be configured to control the user interface as well as other functions (e.g., the window and / or the air condition) based on the configuration data. Accordingly, the model may be trained to provide appropriate configuration data for this based on the user request.

[0070] It is noted that the term “vehicle / s” is to be interpreted broadly in context of the present disclosure. So, vehicles according to the present disclosure may include any kind of vehicle, including various ground vehicles (e.g., car, truck, bus, motorcycle, and / or the like), various watercrafts (e.g., boat), various aircrafts (e.g., planes, helicopters), and / or a combination thereof (e.g., amphibious vehicle).

[0071] Fig. 4 shows a block diagram schematically illustrating an embodiment of such an apparatus 400. The apparatus comprises one or more interfaces 410 for communication and a data processing circuit 420 configured to execute the proposed method. In embodiments, the one or more interfaces 410 may comprise wired and / or wireless interfaces for transmitting and / or receiving communication signals in connection with the execution of the proposed concept. In practice, the interfaces, e.g., comprise pins, wires, antennas, and / or the like. As well, the interfaces may comprise means for (analog and / or digital) signal or data processing in connection with the communication, e.g., filters, samples, analog-to-digital converters, signal acquisition and / or reconstruction means as well as signal amplifiers, compressors and / or any encryption / decryption means.

[0072] The data processing circuit 420 may correspond to or comprise any type of programable hardware. So, examples of the data processing circuit 420, e.g., comprise a memory, microcontroller, field programable gate arrays, one or more central, and / or graphical processing units. To execute the proposed method, the data processing circuit 420 may be configured to access or retrieve an appropriate computer program for the execution of the proposed method from a memory of the data processing circuit 420 or a separate memory which is communicatively coupled to the data processing circuit 420.

[0073] In practice, the proposed apparatus or at least parts thereof may be installed on or in a vehicle. So, embodiments may also provide a vehicle comprising the proposed apparatus. In implementations, the apparatus, e.g., is part or a component of the infotainment system.

[0074] However, in implementations, computing resources for the vehicle may be outsourced to an external server separate from the vehicle. In such implementations, the proposed approach may be also implemented outside of the vehicle, e.g., on a separate server.

[0075] In the foregoing description, it can be seen that various features are grouped together in examples for the purpose of streamlining the disclosure. This method of disclosure is not to be interpreted as reflecting an intention that the claimed examples require more features than are expressly recited in each claim. Rather, as the following claims reflect, subject matter may lie in less than all features of a single disclosed example. Thus, the following claims are hereby incorporated into the description, where each claim may stand on its own as a separate example. While each claim may stand on its own as a separate example, it is to be noted that, although a dependent claim may refer in the claims to a specific combination with one or more other claims, other examples may also include a combination of the dependent claim with the subject matter of each other dependent claim or a combination of each feature with other dependent or independent claims. Such combinations are proposed herein unless it is stated that a specific combination is not intended. Furthermore, it is intended to include also features of a claim to any other independent claim even if this claim is not directly made dependent to the independent claim.

[0076] Although specific embodiments have been illustrated and described herein, it will be appreciated by those of ordinary skill in the art that a variety of alternate and / or equivalent implementations may be substituted for the specific embodiments shown and described without departing from the scope of the present embodiments. This application is intended to cover any adaptations or variations of the specific embodiments discussed herein. Therefore, it is intended that the embodiments be limited only by the claims and the equivalents thereof.

Claims

Patent claims1 . A method (100) for adapting a user interface, the method (100) comprising: obtaining (110) information on a voice input of a user; determining (120) configuration data from the voice input to adapt the user interface using a machine-learning based model (240); and adapting (130) the user interface based on the configuration data.

2. The method (100) of claim 1 , wherein the user interface corresponds to a user interface for a vehicle.

3. The method (100) of claim 2, wherein the user interface comprises an infotainment system for a vehicle.

4. The method (100) of any one of the preceding claims, wherein the machine-learning-based model is configured to generate the configuration data such that it is indicative of one or more adjustment options of the user interface.

5. The method (100) of any one of the preceding claims, wherein the configuration data comprises one or more configuration parameters for a configuration of the user interface, and wherein adapting the user interface comprises applying the configuration parameters.

6. The method (100) of any one of the preceding claims, wherein determining the configuration data comprises converting the voice input to text using a speech-to-text module and determining the configuration data based on the text.

7. The method (100) of any one of the preceding claims, wherein the user interface is configured to execute one or more software applications, and wherein adapting the user interface comprises adapting one or more settings and / or functions of the software applications.

8. The method (100) of any one of the preceding claims, wherein the method (100) further comprises providing a response indicative of the adaption of the user interface.

9. The method (100) of claim 8, wherein the response includes an audio and / or text output to the user.

10. The method (100) of claim 8 or 9, wherein providing the response comprises: obtaining a text form of the response from the machine-learning-based model; and converting the text form of the response into an audio signal using a text-to-speech module.11 . A training method for training a machine-learning-based model (240) for adapting a user interface of a vehicle, wherein the method comprises: training the machine-learning-based model to determine a configuration of the user interface based on a user input indicative of a configuration data for adapting the user interface.

12. The training method of claim 11 , wherein the machine-learning-based model is trained based on training data including one or more exemplary user input samples and ground truth of respective configuration data for the exemplary user input samples.

13. A computer program comprising instructions which, when the computer program is executed by a computer, cause the computer to carry out any one of the methods of any one of the claims 1 to 12.

14. An apparatus (400) comprising: one or more interfaces (410) for communication; and a data processing circuit (420) configured to execute any one of the methods (100) of any one of the claims 1 to 12.

15. A computer-readable data carrier having stored thereon the computer program of claim 13.

Images