Task migration method and mobile terminal

Abstract

This application discloses a task migration method and a mobile terminal, relating to the field of artificial intelligence technology. In response to changes in the screen form of the mobile terminal, task context data is acquired; based on the task context data, task planning information is generated; wherein, the task planning information includes inference information of the current task and at least one of the predicted subsequent executable tasks; thereby, based on the task planning information, a corresponding task management window is output in the changed screen form to form an interactive interface associated with the current task.

Description

TECHNICAL FIELD

[0001] The present application mainly relates to the field of artificial intelligence, and more particularly to a task migration method and a mobile terminal. BACKGROUND

[0002] With the maturity of flexible screen technology, mobile terminals with deformable screens have become one of the mainstream device forms. Their unique physical structure allows users to freely switch between different screen physical forms according to use scenarios and personal habits, providing users with more flexible and immersive interactive experiences.

[0003] However, at present, when the screen form changes, the operating system or application usually adopts a simple response strategy: according to the perceived changes in screen size and proportion, the user interface (UI) is simply rearranged, the controls are shrunk or the information is folded. This way is similar to the processing logic of traditional multi-window, split-screen or landscape / portrait screen rotation scenarios, which only triggers passive adaptation at the UI level, has a single response dimension, and causes the task execution process to be forcibly interrupted, resulting in a significant break in user experience. SUMMARY

[0004] Therefore, the present application provides the following technical solutions:

[0005] The first aspect of the present application provides a task migration method, comprising:

[0006] In response to a change in the screen form of the mobile terminal, task context data is obtained;

[0007] Based on the task context data, task planning information is generated; wherein the task planning information includes at least one of inference information of the current task and predicted subsequent executable tasks;

[0008] Based on the task planning information, a corresponding task management window is output under the changed screen form to form an interactive interface associated with the current task.

[0009] In one possible implementation method, based on the task context data, the task planning information is generated, comprising:

[0010] Based on the task context data, the task planning information is generated by a generative model, and the task planning information includes inference information of the current task and a task flow formed by the subsequent executable tasks;

[0011] Based on the task planning information, a corresponding task management window is output under the changed screen form, comprising:

[0012] The system outputs the reasoning result or execution process of the current task in the task flow under the changed screen format, and provides the control entry point corresponding to the subsequent executable task.

[0013] In one possible implementation, the screen shape change is the switching of the folding screen of the mobile terminal between any two of the following: closed shape, unfolded shape, and at least one intermediate shape.

[0014] The task context data includes at least one of the following:

[0015] Interaction behavior data is used to characterize historical operation trajectories across applications, application switching behavior, or input semantic intent;

[0016] Scene content data is used to characterize the content features or data types of the interface output before the screen shape changes;

[0017] Task status data is used to characterize the current task execution process, generated intermediate results, or the status of interaction sessions with artificial intelligence services.

[0018] The task flow includes multiple consecutive tasks, and the current task is the task being executed in the task flow.

[0019] In one possible implementation, generating task planning information based on the task context data includes:

[0020] Semantic analysis is performed on the task context data to obtain the corresponding semantic analysis results;

[0021] The semantic analysis results are matched with predefined task flow templates to generate task planning information;

[0022] The subsequent executable tasks in the task planning information correspond to the task nodes to be executed in the task flow template.

[0023] In one possible implementation, the generation of task planning information further includes:

[0024] For the subsequent executable tasks included in the task planning information, determine the corresponding candidate executors;

[0025] The candidate executor includes at least one of the following types:

[0026] Callable applications or functional services;

[0027] Applicable content templates or data templates;

[0028] The data object to be processed.

[0029] In one possible implementation, based on the task planning information, a corresponding task management window is output in the changed screen format, including:

[0030] Based on the task planning information, generate interface element information for constructing the task management window;

[0031] The interface element information includes:

[0032] The first type of element information is determined based on the task state data in the task context data and is used to characterize the reasoning result or execution process of the current task.

[0033] The second type of element information is generated based on the candidate executors corresponding to the subsequent executable tasks included in the task planning information, and is used to characterize the control entry point corresponding to the subsequent executable tasks.

[0034] In one possible implementation, the step of outputting a corresponding task management window based on the task planning information in the changed screen format further includes:

[0035] Based on the interface element information, a corresponding task management window is generated to output the task management window in the changed screen form.

[0036] The task management window is configured to display:

[0037] Visual feedback generated based on the information of the first type of elements;

[0038] The control entry point is generated based on the information of the second type of elements.

[0039] In one possible implementation, the control entry point is associated with at least one of the following types of information:

[0040] The position identifier of the subsequent executable task in the task flow;

[0041] The interface description of the candidate executors to be invoked to execute the subsequent executable tasks;

[0042] The execution parameters used to invoke the candidate executor, determined based on the task context data.

[0043] In one possible implementation, the interactive interface and the source application interface associated with presenting the current task are simultaneously presented in a segmented display format under the changed screen form.

[0044] The display area of ​​the source application interface is adjusted in response to the output of the interactive interface.

[0045] A second aspect of this application provides a mobile terminal, comprising:

[0046] A screen capable of switching between at least two different physical forms;

[0047] At least one memory, and a computer program stored in the memory;

[0048] At least one processor capable of running intelligent programs;

[0049] The intelligent program can execute the computer program through the processor to implement the steps of any of the above-described task migration methods. Attached Figure Description

[0050] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0051] Figure 1 This is a flowchart illustrating the task migration method proposed in Embodiment 1 of this application;

[0052] Figure 2 This is an optional schematic diagram illustrating the task migration method proposed in this application, in which a task management window is output to form an interactive interface associated with the current task;

[0053] Figure 3 This is another optional schematic diagram illustrating the task migration method proposed in this application, in which a task management window is output to form an interactive interface associated with the current task;

[0054] Figure 4 This is a flowchart illustrating the task migration method proposed in Embodiment 2 of this application;

[0055] Figure 5 In the task migration method proposed in this application embodiment, the AI ​​workflow and the task flow diagram of the AI ​​workflow are displayed through the task management window in the interactive interface associated with the current task;

[0056] Figure 6 This is a schematic diagram illustrating the regional presentation of the interactive interface and source application interface of the visual task flow in the task migration method proposed in this application embodiment.

[0057] Figure 7 This is a schematic diagram illustrating the generation and visualization of a task flow in an optional application scenario within the task migration method proposed in this application embodiment.

[0058] Figure 8This is a schematic diagram illustrating the generation and visualization of a task flow in another optional application scenario within the task migration method proposed in this application embodiment.

[0059] Figure 9 This is a flowchart illustrating the task migration method proposed in Embodiment 3 of this application;

[0060] Figure 10 This is a schematic diagram of the task migration device proposed in the embodiments of this application;

[0061] Figure 11 This is a schematic diagram of the hardware structure of a mobile terminal proposed in this application. Detailed Implementation

[0062] The embodiments of this application are described below with reference to the accompanying drawings. The terminology used in the implementation section of this application is only for explaining specific embodiments and is not intended to limit the application. The embodiments of this application are described below with reference to the accompanying drawings. It will be understood by those skilled in the art that, with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0063] The terms "first," "second," etc., used throughout this application and in the foregoing figures are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms are interchangeable where appropriate; this is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, so that a process, method, system, product, or apparatus that comprises a series of units is not necessarily limited to those units, but may include other units not explicitly listed or inherent to those processes, methods, products, or apparatuses.

[0064] It is understood that before using the technical solutions disclosed in the embodiments of this application, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this application in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained. For example, in response to receiving a user's active request, a pop-up window may be used, and a textual prompt message may be presented in the pop-up window to clearly inform the user that the requested operation will require the acquisition and use of the user's personal information. The user can choose whether to provide personal information to the electronic device, application, server, or storage medium or other software or hardware that performs the operation of this application based on the prompt message. This application does not limit the prompt message and the method of user authorization implementation. The data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws and regulations and related provisions (such as corporate regulations).

[0065] Furthermore, the models involved in this application (such as generative models) can be general AI (Artificial Intelligence) models. These AI models can employ, but are not limited to, Transformer or its architectural variants (such as encoder-only / decoder-only, encoder-decoder, or MoE (Mixture of Experts, a neural network architecture) or other infrastructures). They learn the features and patterns of natural language by training on large amounts of diverse data, thereby enabling them to understand and generate natural language. They typically have hundreds of millions to trillions of model parameters (model parameters are variables that control the behavior of the target model) and are capable of capturing complex relationships and patterns in natural language.

[0066] The AI ​​models can include, but are not limited to, generative models and generative language models (GLMs). For example, one or more of the following: large language model (LLM), GPT (Generative Pre-trained Transformer) series models, T5 (Text to Text Transfer Transformer) model, BERT (Bidirectional Encoder Representation from Transformers) series models, large visual models, and multimodal large models.

[0067] Depending on actual needs, the models involved in this application's embodiments can also be expert large models fine-tuned from general AI models based on actual business requirements. For example, they can be proprietary models trained on sample datasets from general AI models (pre-trained models) within specific scenarios (such as professional scenarios in vertical fields like healthcare, finance, law, biology, or remote sensing, or general scenarios). The models involved in this application can also be lightweight models compressed from general AI models or expert large models through lightweight methods such as quantization, knowledge distillation, or pruning, to better meet the needs of edge deployments with limited computing resources and improve data security. This application does not limit the type of model involved in the context and can flexibly determine it according to actual business needs.

[0068] As described in the background section, when using mobile terminals with deformable screens, users' true intention when changing screen orientation is often to continue the core task at hand, rather than to start anew. However, current UI (User Interface) reflow processes abruptly interrupt the user's visual focus and workflow, resulting in a loss of task context and a strong sense of experience disruption. For example, when switching from an outer screen to an expanded large screen, the user's true intention shifts from "rapid information consumption" to "deep content production" (such as moving from browsing summaries to editing long articles), not just to see "larger buttons and icons." Currently, users typically need to manually reposition and reorganize information on the new screen orientation to "recall" and "rebuild" their previous workflow, severely impacting the continuity and smoothness of core tasks.

[0069] Therefore, this application aims to proactively understand and capture the task context behind screen shape changes, and intelligently maintain or migrate task continuity. By adapting the interactive interface to the new screen shape and associating it with the current task, it achieves a seamless, task-centric, and continuous experience across screen shapes, thereby solving the user experience fragmentation problem caused by simple UI redesign. It provides intelligent and fitting interactive support, truly unleashing the interactive potential of deformable screen devices. Therefore, this application proposes a new task migration method and terminal device. The task migration method proposed in this application will be described in detail below with reference to the accompanying drawings.

[0070] Reference Figure 1 This is a flowchart illustrating the task migration method proposed in Embodiment 1 of this application. The method can be applied to mobile terminals with variable-form screens (such as foldable screens, rollable screens, or stretchable screens), such as smartphones, tablets, smart wearable devices, or other business terminals. This application does not limit the product form or screen type of the mobile terminal; it can be determined as appropriate. Figure 1 As shown, the task migration method proposed in this embodiment may include, but is not limited to:

[0071] Step S11: In response to changes in the screen shape of the mobile terminal, obtain task context data;

[0072] In this embodiment of the application, the screen shape change of the mobile terminal (referring to the physical shape change of the screen) can be a hardware-level, intentionally defined input signal / event. After the system listens to and captures the input signal / event, it will immediately trigger the task context acquisition module to actively acquire task context data, instead of just sensing the shallow data of screen size. This provides a data foundation for subsequent intelligent planning to maintain or migrate task continuity.

[0073] The task context data can be multi-dimensional data related to the user's actions (current task) performed in the previous screen configuration. This application does not restrict the active acquisition method or content of the task context data. Taking a foldable phone as an example, if the phone is in the outer screen state, and the user is using a note-taking application to write meeting minutes and has already generated a text draft, the user unfolds the foldable screen from the outer screen state to the inner screen state, immediately triggering the task context acquisition module. It detects that the user's current application is the "note-taking application," the user's last edit position is at the end of the third paragraph, and there is an unsaved text draft; it detects that the user copied a meeting recording in the past 3 minutes, converted it to text, pasted it into the notes, and bolded the title in the text draft; it also detects that the user edited the text as meeting minutes and just entered the keyword "to-do event," etc., but is not limited to these. Therefore, this application provides a data foundation for the system to understand "what the user is doing, where they are, and what they want to do next" by acquiring task context data.

[0074] Step S12: Generate task planning information based on task context data; wherein, the task planning information includes at least one of the inference information of the current task and the predicted subsequent executable tasks.

[0075] The task context data output from the event perception and data acquisition layers can be input into the task planning engine / inference engine to perform semantic analysis and intent recognition. This allows for the inference of the user's current task (which could be an ongoing task, such as organizing meeting minutes as shown in the example above, or preparing for an upcoming task), generating inference information for the current task to understand its current state, what has been accomplished, what problems have been encountered, and its core objectives. For example, in the example above, it could be inferred that "the user is organizing meeting minutes and has currently completed a draft text." This application does not restrict the method or content of obtaining the inference information for the current task; it can be determined based on the specific data content of the currently acquired task context data.

[0076] Optionally, based on task context data, the system can also predict the user's next needs in the changed screen format to plan the tasks most likely to be executed in the future (denoted as subsequent executable tasks, which may be several executable operation steps), achieving proactive prediction at the task level. For example, in the example above, the predicted subsequent executable task could be "converting meeting minutes into a PPT presentation," etc. It should be noted that the subsequent executable tasks predicted in this application can be the next task to be executed directly and continuously with the current task, and can also include the next task to be executed after that or more consecutive tasks to be executed, that is, it can be at least one task to be executed directly or indirectly continuous with the current task. For example, if the task context data includes a PPT request received by a social application, the predicted subsequent executable tasks could include "extracting key information," "generating a PPT outline," and "writing a draft response," etc. This application does not limit the content and number of predicted subsequent executable tasks.

[0077] Based on the above analysis, the inference information of the current task represents the system's understanding of the past and present, while subsequent executable tasks represent future planning. In practical applications, the prediction of subsequent executable tasks can be achieved based on the inference information of the current task or directly based on task context data; this application does not restrict the implementation method. Furthermore, as the content of the acquired task context data changes, the content of the generated task planning information will also differ, further affecting the content displayed in the changed screen format.

[0078] Therefore, this application can define / encapsulate (this method can obtain structured data) the reasoning information of the current task generated by reasoning from task context data as task planning information; it can also define / encapsulate the subsequent executable tasks predicted by reasoning from task context data as task planning information; or it can define / encapsulate the reasoning information of the current task generated by reasoning from task context data and the predicted subsequent executable tasks together as task planning information. Of course, depending on actual needs, task planning information can also include other content, such as the results of executed tasks, various types of execution bodies prepared in advance to realize subsequent executable tasks, etc. This application does not limit the content and data format of task planning information.

[0079] Step S13: Based on the task planning information, output the corresponding task management window in the changed screen format to form an interactive interface associated with the current task.

[0080] This application can transform task planning information into at least one visual task management window, which is a UI container specifically generated for managing and guiding tasks, independent of the source application (the application displayed before the screen shape change). Based on the above description of task planning information, at least one task management window can be generated corresponding to its different contents. For example, a window displaying the progress of meeting minutes writing in the form of a progress bar, a window displaying the current task: organizing meeting minutes, a window displaying tools for completing the current task, and a window displaying subsequent executable tasks, etc.

[0081] As can be seen, each task management window obtained in this application is directly or indirectly related to the current task, and the interactive interface formed by the display of each task management window in the changed screen format is also related to the current task. For different contents included in the task planning information, one or more task management windows in the interactive interface can display the reasoning information of the current task, such as... Figure 2 As shown in Figure (a), users can intuitively see the status of the task being executed before the screen shape change (the interrupted current task) through the task management window (interactive interface), ensuring the continuity of the user experience; one or more task management windows can also display predicted subsequent executable tasks, such as Figure 2 As shown in Figure (b), each subsequent executable task corresponds to a task management window. Of course, multiple consecutive subsequent executable tasks can also be displayed in a single task management window, etc. Optionally, such as... Figure 3 As shown, this application can also present relevant information about different tasks (current task and predicted subsequent executable tasks, etc.) through different task management windows, without any limitation.

[0082] Based on this, users no longer need to switch between multiple applications to find functions. They can directly trigger subsequent tasks with a single click through the corresponding task management window presented in this interactive interface. Users do not need to think about what to do next or learn new interactions, and can enjoy intelligent assistance, greatly improving the efficiency of human-computer interaction. Specifically, for task management windows presented through an interactive interface (especially when multiple task management windows are presented in parallel), the layout, information density, and interaction methods of each task management window need to intelligently adapt to the changed screen format. For example, on a small screen, the task management windows may be presented in a compact list format, while on a large screen, they can expand into a detailed view with flowcharts and previews, making it convenient for users to operate the task management windows presented in the changed screen format, thus ensuring a balance between information density and operability.

[0083] In summary, by acquiring "task context data" and generating "task planning information," this application enables the system to perceive the user's true intention to change the screen format. The "task management window" presented in the output interactive interface acts as a "bridge," seamlessly connecting the task logic before and after the screen format change. Users do not need to manually rebuild the context, achieving seamless "zero-interruption" task migration across screen formats. This solves the problem of user experience disruption (task continuity disruption) caused by traditional methods during screen format switching.

[0084] By introducing "predicted subsequent executable tasks," this application gains a forward-looking capability. In the instant the screen changes shape (within a very short time), it anticipates the next task for the user and proactively presents it through the "task management window." This significantly reduces the user's cognitive load and operational costs. Users no longer need to plan themselves; direct visual guidance reduces decision-making costs, achieving an upgrade from a "passive response" to a "proactive service" interaction mode and improving human-computer interaction efficiency.

[0085] Furthermore, by displaying the reasoning information of the current task and subsequent executable tasks simultaneously through different task management windows in the interactive interface, the large screen space advantage can be leveraged to encapsulate complex cross-application operations (such as "copying text → opening PPT → pasting → formatting") within a single interactive interface, fully utilizing the productivity advantages of the transformed large screen format. The displayed reasoning information of the current task transparently reveals the agent's thought process to the user. This interpretable design reduces user concerns about agent misjudgments. Simultaneously, the accuracy of predicted executable tasks generated based on task context data is higher, enhancing user reliance on and satisfaction with the agent's functionality.

[0086] Furthermore, this application provides a standardized task takeover framework, using the task management window as a standardized output interface to offer a consistent user experience for task migration across different applications. Application developers do not need to handle form changes independently; they only need to provide task context data to access this intelligent task succession system, which has excellent ecosystem scalability.

[0087] Reference Figure 4 This is a flowchart illustrating the task migration method proposed in Embodiment 2 of this application. This embodiment can describe a possible refined implementation of the task migration method described above, such as... Figure 4 As shown, the task migration method proposed in this example may include:

[0088] Step S41: In response to the screen shape change of the mobile terminal, obtain task context data;

[0089] Based on the above description of task context data, during the collection of task context data, the system does not simply rely on screenshots, but rather captures as many raw data snapshots as possible of the moment the screen changes, interrupting the task. This includes the type of foreground application (e.g., social, office), the specific content displayed on the interface (text, image, video), the user's input state (e.g., text being edited, cursor position), and the status of background services, providing raw and rich input features for subsequent generative models. Therefore, task context data can include, but is not limited to, at least one type such as unstructured text data, image data, video data, and operation logs. This application does not restrict the types of data it includes or the methods / sources of its collection.

[0090] Step S42: Based on the task context data, generate task planning information through a generative model; wherein, the task planning information includes the inference information of the current task and the task flow formed by the predicted subsequent executable tasks;

[0091] In this embodiment, unlike matching methods based on rule engines, decision trees, or traditional machine learning classifiers, this embodiment utilizes generative models (such as large language models or multimodal large models, etc.) to achieve intelligent planning. Specifically, it leverages the powerful semantic understanding, content generation, logical reasoning, and multi-step planning capabilities of generative models to perform deep semantic analysis and inference on rich task context data, thereby achieving continuous task planning. This application does not limit the type of generative model or its input information processing methods; implementation can be based on the computational principles of the generative model itself.

[0092] In this application, based on the relevant description of task context data, the task context data or input data processed to meet the model input format requirements is directly input, or prompt information is constructed by combining prompt word templates and then input into the generative model. The generative model infers "what the user is doing" (the current task) to obtain the inference information of the current task. Therefore, this inference information is a readable conclusive description generated by the generative model after considering the task context data, which can be the inference result or execution process of the current task.

[0093] For example, if a user is identified as viewing a long text about "product launch" in WeChat, the inferred information could be "extracting key information and planning the launch process"; or it could be a sentence generated through inference, such as "You are writing a product design report emphasizing 'continuous experience' based on a tech news article about foldable screens, and you have just completed the pain point analysis section, extracting three key arguments" or a summary description, etc. This application does not restrict the implementation method of generative models obtaining inference information for the current task based on task context data, nor the content of the inference information.

[0094] Furthermore, during the aforementioned reasoning process, generative models can also analyze semantic relationships within the task context based on a deep understanding of the current task, simulate the user's thought process, and predict a series of subsequent tasks to generate a logically ordered task flow. For example: generating a summary → creating a PPT outline → composing an email notification. It is evident that the task flow is not an isolated single task, but a reasonable, ordered, and executable sequence / set of tasks to ensure semantic coherence. Understandably, a task flow can be formed by multiple predicted consecutive executable tasks; it can also be formed by the current task and at least one predicted subsequent executable task, according to the logical order of these tasks. If the preceding task (executed task) of the current task is also obtained based on the task context data, the task flow can also include that preceding task so that the user clearly understands the background and development of the current task. Therefore, the task flow of this application typically includes multiple consecutive tasks, with the current task being the task being executed within the task flow. This application does not limit the content, number, or form of the tasks forming the task flow.

[0095] In the process of generating task planning information through generative models, the granularity of tasks can be dynamically adjusted based on the content characteristics of user operations in the task context data. For example, if the user copies many technical points, the generative model will generate more detailed sub-tasks (corresponding to a task node); if the content is less, a higher-level, more general task (corresponding to a task node) can be generated. Thus, generative models can dynamically determine the level of task granularity based on the amount of data accumulated by the user in the task context data. Furthermore, customized task planning information can be generated based on the specific content and operation history of user operations to achieve personalized adaptation.

[0096] Based on the above analysis, the task flow of this application is not a simple application jump list, but an automated workflow driven by artificial intelligence (AI) (i.e., AI workflow). It can deconstruct the user's current interaction intent through generative models, breaking down a single complex goal into multiple fine-grained tasks (AI workflow nodes / task nodes) with clear logical dependencies. Each task can correspond to an execution entity (such as an application service, AI model, or data operation interface). The tasks are connected through data channels, with the output of the previous task automatically serving as the input of the next, thus forming a closed-loop task execution chain. The task flow of this application possesses state awareness and dynamic branching (data flow) capabilities, enabling real-time adjustment of the execution path of subsequent tasks based on changes in screen format and intermediate results of task execution (which represent the state of the corresponding task), achieving a continuous task experience across screen formats.

[0097] In this embodiment of the application, a generative model can be used as an "orchestration engine." Based on the intent understanding of the task context data, it retrieves or infers subsequent executable tasks starting from the current task, and links them in a logical order to form an AI workflow specific to the current scenario (a task execution blueprint described from the system implementation perspective). For example, in the same scenario of "expanding the screen to process social messages," for the "meal invitation" content included in the task context data, the task flow generated by the generative model based on this content could be: extract time and location → create calendar → navigation; for the "request materials" content included in the task context data, the task flow generated by the generative model based on this content could be: identify file type → cloud drive retrieval → send file. This demonstrates the real-time generation and highly customized (personalized) characteristics of the task flow in this application.

[0098] Step S43: Based on the task planning information, output the reasoning result or execution process of the current task in the task flow under the changed screen format, and provide the control entry corresponding to the subsequent executable tasks.

[0099] In this embodiment, the reasoning information of the current task includes the reasoning result, which can be a conclusive description of the current task generated by the generative model (e.g., a summary has been generated for you...). This reasoning result is presented visually. This can be achieved when the current task is interrupted in the "thinking" phase to help the user quickly recover the cognitive context of the current task, but is not limited to this. Thus, during the rendering process of the corresponding task management window output under the changed screen format, the reasoning result can be controlled to be highlighted in the task management window, achieving a differentiated presentation between the reasoning result of the current task and other content in the task management window. For example, the reasoning result can be displayed at the top of the task management window, or the display attributes of the reasoning result can be adjusted using any prominent method such as font enlargement, bolding, color change, or background color addition; there are no limitations on this.

[0100] Optionally, this application can also present the inference results of the current task separately through the task management window (e.g., Figure 2 As shown in Figure (a), in addition to adjusting the display attributes of the inference results according to the method above, the inference results of the current task can also be highlighted by adjusting the layout relationship between the task management window and other task management windows, or by adjusting the display attributes of the task management window itself. This application does not restrict the visualization presentation of the inference results of the current task.

[0101] Furthermore, if the inference information of the current task includes the execution process (such as an edited document preview or processed image effects), the execution process of the current task can also be presented in a visual form, following, but not limited to, the methods described above. That is, the execution process can be presented in the corresponding task management window output in the changed screen format. This can be achieved when the current task is interrupted in the "operation phase," helping users quickly locate the work site. Therefore, the visual presentation of the inference information of the current task in this application allows users to intuitively perceive the working status of the AI.

[0102] The execution process of the current task can be displayed by showing its current status, which can be presented in the form of a progress bar or a chart. For example, the task management window can display the requirements analysis progress: 65% (indicators extracted, constraints to be completed); or display at least one chart such as a pie chart, bar chart, or line chart reflecting the current task progress (65% complete). Optionally, depending on actual needs, the execution process presented in the task management window can also include the execution steps to achieve the current task (which may include completed steps, steps being executed, and steps to be executed) and their execution results, so that users can intuitively and clearly understand how the current task is completed.

[0103] In the embodiments of this application, the reasoning result and execution process of the current task can respectively represent different states of the current task, and the output content for the current task can be intelligently selected according to the actual situation. Moreover, in the process of visualizing the reasoning information of these two dimensions, it can be achieved through a task management window (a task management window for the current task), or through different task management windows (in which case a task can output multiple task management windows that implement different functions), and this application does not impose any restrictions on this.

[0104] When the task planning information includes a task flow (AI workflow) formed by subsequent executable tasks, the task management window, as the carrier (or visual container) of the task planning information, can present at least one control entry point, such as interactive controls / information in the form of buttons, icons, floating cards, timelines, or interactive list items. By configuring these control entry points within the task management window, a concrete mapping of the corresponding subsequent executable tasks onto the interactive interface is achieved. This allows users to intuitively understand the complex task flow logic by recognizing the control entry points. For example, if a task in the task flow is "call AI to generate a summary," this is mapped to a button or card labeled "Generate Summary" in the corresponding task management window. When the user needs to generate a summary, they can directly trigger this button / card to automatically call AI to generate the summary, improving the convenience and efficiency of human-computer interaction.

[0105] Therefore, the control entry point in this embodiment is not an independent element floating outside the task management window, but rather a part of the content presented in the task management window. The task management window provides a display carrier for the control entry point, ensuring that the control entry point and other information in the task flow (such as task status and inference results) maintain visual consistency and coherence. Moreover, as the tasks forming the task flow are executed, the display content of the corresponding task management window is updated accordingly (such as changes in the task progress bar, the lighting of flowchart nodes, etc.), and the control entry point can also change dynamically, such as "Execute" changing to "Complete," or the appearance of a new next step operation entry point, driving the task flow of the task flow, thereby achieving an upgrade in the interactive experience from "passive viewing" to "active guidance."

[0106] For example, such as Figure 5 As shown, if the task flow formed by multiple consecutive executable tasks generated by the generative model is: generate summary → create PPT outline → compose email notification, then in the changed screen format, a task management window will be output for each subsequent executable task. The control entry type, content, layout, etc., presented in this task management window can be dynamically generated by the generative model, such as... Figure 5 In Figure (a), the "Generate Summary" task management window presents three control entry points: "Start Generation," "Regenerate," and "Export Summary." Users can select a control entry point by touch (such as using a mouse and keyboard or voice commands; this application does not restrict the touch method) according to their actual needs and execute the task according to the corresponding instructions. For example, selecting "Start Generation" will immediately call the AI ​​service to generate a summary; selecting "Export Summary" will export the generated summary to a specified storage location or input it into other executors (such as as input for the next task) according to a predefined or selected file format. Alternatively, a pop-up floating window can be used to display the exported summary.

[0107] In addition to outputting the task management window corresponding to subsequent executable tasks, the inference information of the current task can also be presented through the corresponding task management window (such as...). Figure 5 (Extracting key information, etc.). After touching the control entry point corresponding to the "Generate Summary" task, the system automatically generates a summary and drives the task flow. The content displayed in the corresponding task management window is automatically updated to the generated summary content. The layout and other display attributes of each task management window can also be dynamically adjusted as needed, such as... Figure 5 As shown in Figure (b), the inference results / execution process of the executed task and the inference results / execution process of the current task are displayed in the upper display area of ​​the interactive interface. The way information is presented in the task management window of the subsequent executable task "Create PPT Outline" is adjusted, compared to... Figure 5 As shown in Figure (a), clicking the "Select Template" drop-down menu button provides access to various PPT templates that the user might need. This application can directly display each PPT template in the task management window for easy comparison and selection by the user, but it is not limited to this. Figure 5 The visualization method shown.

[0108] It should be noted that each task in the task flow corresponds to a task management window, and a control entry point is presented in the task management window. This control entry point can be an interactive control within the task management window, in which case it can be a part of the task management window; alternatively, the control entry point can be displayed within the display area of ​​the task management window, in which case the task management window itself is a control entry point (e.g., ...). Figure 2 and Figure 3 As shown in the diagram, it can be presented in the form of content cards or interactive list items. Users can trigger any position within the display range of the task management window to immediately trigger the execution of the corresponding subsequent executable tasks. This application does not restrict the implementation method of the control entry corresponding to each subsequent executable task.

[0109] In summary, this embodiment employs a generative model as the core engine for task migration. Through this model, the complex semantic relationships within task context data can be understood, generating personalized task flows and significantly improving the system's intelligence and scenario coverage. In other words, leveraging the powerful general understanding and reasoning capabilities of generative models, unprecedented and complex cross-domain task combinations can be handled. Whether users are processing work documents, planning trips, learning knowledge, or creating content, by processing the collected task context data through generative models, they can attempt to understand and generate reasonable task planning information, greatly expanding the applicability and practicality of the method described in this application.

[0110] Moreover, because generative models can perform multi-step reasoning and creative association, they may propose reasonable subsequent executable tasks that the user had not thought of, forming a logically coherent task flow (i.e., AI workflow). For example, when editing images, it can not only predict "filters" and "cropping," but also suggest tasks such as "generating a matching caption for social media posting." This makes task migration no longer a simple continuation of the current task, but rather an enhancer of work efficiency and creative inspiration. At the moment of screen form change (such as screen unfolding), users can see a clear task flow on the interactive interface, greatly reducing cognitive load and improving the coherence and efficiency of task processing. This application's presentation of complex task flows in a visual form maximizes the use of the interactive resources of the large screen, achieving a dual improvement in information density and interaction efficiency, fully releasing the productivity potential of foldable screen hardware. Furthermore, it provides a visual output of the corresponding control entry points for subsequent executable tasks, allowing users to trigger them with a single click, significantly reducing operational costs.

[0111] Furthermore, by outputting inference information such as the reasoning results or execution process of the current task, the AI's thinking process is made white-boxed. Users not only know what the system recommends, but also "why" the system recommends it (based on the inference results of the current task). This transparent interactive design allows users to clearly understand the completion status of the current task, reducing uncertainty and understanding the system's decision-making basis. Users retain the final choice, being able to select to execute or ignore multiple predicted subsequent executable tasks, enhancing their trust and sense of control over intelligent assistance. Moreover, the inference information generated by the generative model is described in natural language, making it easy to understand, much like a colleague updating you on project progress. This reduces the cognitive burden required for users to re-understand the task status after screen changes, making the user experience more seamless and natural.

[0112] In some embodiments, the application interface (which is the original interface the user is currently operating, such as a social media chat interface or a document editing interface, and is the source of the user's task, denoted as the source application interface) and its displayed content output by the mobile terminal before the screen shape changes, according to the method of this application, will not be interrupted with the change of screen shape. Therefore, the source application interface is still presented in the changed screen shape, and the source application interface is associated with the current task, so that the displayed content of the source application interface is continuous with the interface content interrupted when the screen shape changes, providing the user with a continuous visual reference to ensure the continuity of the current task. This application does not limit the way the source application interface is output in the changed screen shape.

[0113] Since the interactive interface formed by the output task management window is associated with the current task in the changed screen form, and the source application interface is also associated with the current task, the interactive interface presented in this application is associated with the source application interface. These two are presented simultaneously in the form of regional display in the changed screen form. Users can obtain two kinds of information at the same time without switching views, maintaining the "visual anchor point" of task migration, solving the problem of context loss, and providing a natural interactive scenario for comparing the content of these two interfaces. This enables intuitive comparison and verification of source data (task context data) and AI processing results (inference information of the current task, task flow), improving the user experience.

[0114] Based on this, when the mobile terminal detects a change in screen form (such as unfolding a large screen), and the system decides to display the task migration results in a "split-screen / collaboration" mode, it can trigger the control interface and the source application interface in this embodiment to be displayed simultaneously in a segmented manner. The physical space under the changed screen form (large screen form) can be logically divided into at least two display areas, such as... Figure 6 As shown, the source application interface is displayed on the left side of the screen (the first display area). Figure 6 The video playback interface shown in Figure (a) is displayed, and an interactive interface (i.e., the task management window that forms the interactive interface, such as...) is displayed on the right side of the screen (second display area). Figure 6 (See Figure (b) for the AI ​​workflow interface). In this way, users do not need to remember or imagine the content of the previous screen. The visual focus can smoothly move between the original task (the interrupted current task, associated with / presented in the source application interface) and intelligent suggestions (task planning information, presented in the interactive interface), maintaining the visual continuity of the user experience.

[0115] Furthermore, the display area of ​​the source application interface adjusts in response to the output of the interactive interface. This adjustment is responsive; that is, when the interactive interface outputs, the source application interface responds immediately; if the interactive interface is closed or collapsed by the user, the source application interface can return to full-screen display. Therefore, the regional display method of the interactive interface and the source application interface in this application is not a simple overlay obscuring, but rather the two interfaces share screen space, do not obscure each other, and are in a parallel active state. It is not a rigid cutting or fixed partitioning of the original application interface, but a dynamic and intelligent collaborative process to ensure complete and clear content display. Users can continue to operate / view the content displayed on the source application interface without repositioning, achieving "zero-interruption" task migration. It does not affect the user's main operations on the source application interface; users can choose whether to use the intelligent assistance of the interactive interface, achieving a user-friendly interaction that assists without interfering.

[0116] Optionally, adjustments to the source application interface can be implemented based on at least one of the following: information density, interaction focus, or shape change. Specifically, during the implementation of information density-based adjustments, if the task planning information output by the interactive interface is rich in content, the display area (first display area) of the source application interface can be appropriately scaled, such as... Figure 6 As shown, when zooming to 40% of the left side of the screen, the layout can be automatically rearranged based on the adjusted first display area size, including text wrapping, image scaling, and table column widths. This ensures that the complete content remains readable even after being zoomed out, and maintains its original state (such as reading progress, highlighted marks, and annotations). This maximizes the display of content, allocating more space for the interactive interface (ensuring a sufficiently large second display area) to guarantee the readability and completeness of task planning information, and facilitating subsequent interactive operations. Conversely, if the interactive interface only needs to display the reasoning results or key execution processes of the current task (such as...), the layout can be adjusted accordingly. Figure 2 (as shown in Figure (a)), or the control entry point corresponding to one or two subsequent executable tasks (such as...) Figure 2 As shown in Figure (b), the source application interface can take center stage, with a larger first display area to facilitate user viewing or operation of the source application interface.

[0117] When users engage in deep interaction with the control entry point and interface (such as expanding the details of a subsequent executable task), the source application interface can be adjusted based on the interaction focus, such as dynamically highlighting or fine-tuning the ratio of two display areas to guide the user's visual flow. Furthermore, the ratio and arrangement of the divided areas (such as left-right, top-bottom, or picture-in-picture, etc.) can vary depending on the screen configuration (e.g., fully unfolded inner screen, half-folded and hovered). Figure 6 The presentation method shown can be intelligently adapted to achieve the optimal layout for this screen type.

[0118] Based on the above analysis, Figure 6 The scenario shown is used as an example for explanation. Before the mobile terminal screen is unfolded (e.g.) Figure 6 In the small screen / outer screen configuration shown in Figure (a), the user browses social media on the outer screen, sees AI learning videos, and performs actions such as liking / favoriting the video content. The user then unfolds the screen to its fully unfolded state (e.g., ...). Figure 6 (As shown in Figure (b) – large screen / inner screen configuration), the system captures screen configuration change events, obtains task context data, and generates task planning information that matches the user's true intent through a generative model. For example, it infers that the user wants to engage in deep learning and that the current task is to generate video summaries. During this process, it can quickly extract the core points of the video to generate video summaries and present the inference results or execution process in the corresponding task management window. Figure 6The AI ​​shown in Figure (b) can be applied to multiple fields, machine learning and deep learning (and related key points), AI decision-making ethics considerations (and related key points), etc. It can also accurately locate key knowledge points by combining core key points with timeline navigation.

[0119] Meanwhile, generative models can also infer and predict the tasks that the user may or needs to perform next (predicted subsequent executable tasks) based on task context data. In this example, this could involve extracting cases from a video or retrieving recommendations that match those cases, and then visually presenting the learning content through an image slideshow. Based on the above description of subsequent executable tasks, this application does not limit the presentation method, such as... Figure 6 As shown in Figure (b), a task management window can display one or more control entry points corresponding to subsequent executable tasks (recommended case tasks), such as prompts to create image slideshow cards, with one card corresponding to each case. During this process, the system can also recommend AI experience tools (TryAI, such as web-based tools) to present corresponding control entry points (which can be indicated by the area where the task name is displayed / function buttons, such as...). Figure 6 The "AI Experience Tools" shown in the image facilitate users to trigger this control entry point, which calls up and presents the recommended AI experience tools for users to choose from, or the system can automatically call up the recommended AI experience tools to process the output of the previous task.

[0120] In the example above, the generative model can also suggest a customized learning plan (calendar) to the user. This plan, as a subsequent executable task, can also provide corresponding control entry points in the deformed screen format, in ways not limited to those described above (e.g., Figure 6 The "Create a study plan" option shown can be displayed in the area below the source application interface if there is insufficient space in the second display area. Figure 6 As shown in Figure (b), the update includes, but is not limited to, updating the content displayed on the source application interface in the changed screen format. This allows users to create a learning plan by tapping the "Create Learning Plan" entry point. Based on the inference information and task context data for each task, a learning plan tailored to the user is generated, eliminating the need for manual creation based on recalling watched video content. Users can use the automatically generated learning plan as a reference and adjust it according to their needs, such as using input components (like a mouse, keyboard, or stylus) to edit the learning plan displayed via pop-ups or file export, thus preserving the final plan. Furthermore, the system can monitor the user's AI learning progress based on this learning plan and dynamically adjust it according to the user's actual progress, without any limitations.

[0121] Therefore, when users watch AI learning videos on the outer screen of their mobile phones and then unfold the phone screen to the inner screen, the task migration method proposed in this application can automatically connect different learning assistance tasks (subsequent executable tasks) without interrupting playback / viewing, and provide corresponding operation entry points. Furthermore, the playback interface of the AI ​​learning video and the interactive interface presenting the task flow are displayed simultaneously in separate areas. This lossless simultaneous screen comparison method of "on-site (source application interface) - planning (interactive interface)" simplifies the user's thinking mode from "recall-understand-decision" to a human-computer collaborative mode of "seeing-choosing," minimizing cognitive burden and significantly improving decision-making speed and quality. This reduces the burden of manual switching for users and improves learning efficiency. Moreover, this application supports multi-threaded parallel processing for complex tasks. For compound tasks that require reference to multiple sources and execution of multiple steps, the simultaneous screen display allows users to operate on the source application interface while simultaneously viewing task steps, calling tools, and consulting AI-generated reference content on the interactive interface. This breaks down application boundaries, ensuring that the user's attention remains focused on the task itself, rather than on the loss caused by switching between different applications.

[0122] In some embodiments, taking a foldable screen as an example, the screen form changes involved in this application can be switching between any two physical states of the foldable screen of the mobile terminal: closed state (outer screen form, usually a small screen), unfolded state (inner screen form, usually a large screen), and at least one intermediate state (such as hovering state, semi-unfolded state, etc.). The closed state typically corresponds to the outer screen working mode, suitable for lightweight interactions (such as browsing notifications, quick replies, etc.), allowing for convenient one-handed operation. The unfolded state typically corresponds to the large screen / inner screen working mode, suitable for immersive experiences (such as document editing, video watching, content creation, and other high-load tasks). An intermediate state can be any physical form between the closed and unfolded states when the user unfolds the screen, such as a laptop-style hovering mode, suitable for pausing, finishing, or switching scenes. This application does not limit the definition of the different physical forms of the screen. When the screen of the terminal device is of other types, such as a rollable screen or a stretchable screen, the definitions of these three physical forms of the screen can be adaptively adjusted, and this application will not provide detailed examples of each.

[0123] Furthermore, in practical applications of this application, different physical form switching paths (such as closed form → expanded form, expanded form → hovering form, hovering form → closed form, etc.) often imply different user intentions. For example, "closed form → expanded form" may mean that the user intention is shifting from "light consumption" to "deep production," while "expanded form → hovering form" may mean that the user intention is shifting to "semi-automatic continuous viewing," etc. Therefore, this application defines different triggering logics for different physical form switching paths to achieve differentiated intelligent responses, making the task migration triggering mechanism more refined and scenario-based.

[0124] When the user unfolds the foldable screen to approximately a 90-degree tent-like shape (an intermediate shape, but not limited to this), the system can still trigger the task migration process described in this application. Considering that the user might place the device on a desktop for hands-free operation, the layout of the task management window will be adaptively adjusted, such as displaying it in a floating format on the upper half of the screen for easy viewing and clicking from a distance. The display size of the control entry point can also be appropriately increased to reduce touch difficulty. The visual feedback content of the interactive interface can also be simplified, highlighting core information. Thus, by adapting to different intermediate shapes, the system flexibly responds to user needs in different usage scenarios, further improving the human-computer interaction experience of foldable screen devices. The implementation process is not detailed in this application.

[0125] In some embodiments, the task context data involved in this application may include at least one of the dimensions such as interaction behavior data, scene content data, and task status data, serving as a knowledge base for the agent's subsequent reasoning and thinking, and dynamically updated / accumulated as the user uses the mobile terminal or the mobile terminal task progresses. The implementation process is not detailed in this application.

[0126] In this embodiment, interactive behavior data is used to characterize at least one of the following: cross-application historical operation trajectory, application switching behavior, or input semantic intent. This data content is related to the current task and reflects what the user has done / how the user operates. It can be a collection of historical information representing the user's interaction with the mobile terminal over a past period. The cross-application historical operation trajectory can be the order and path of the user's operations across multiple applications (source application and other applications). For example, a user might search for "coffee shop" in a map application, then switch to a notes app to write a note, and then open WeChat (a social media app) to share their location. This application can collect the operation trajectory "Map → Notes → WeChat," which includes the user's historical operations within a single application. Therefore, this historical operation trajectory itself is a high-value task context, facilitating the analysis of user operating habits and potential intent chains, and determining whether different application jumps belong to the same task chain.

[0127] Application switching behavior can include at least one of the following: switching frequency, switching mode, and dwell time. This is clearly a strong intent signal. For example, a user spends 18 minutes in a document reading application (deep reading), briefly switches to an email application to check meeting notifications (dwelling for 5 seconds), and then returns to the reading application. Based on this, it can be inferred that the user's current task is focused on reading and analyzing reports. Input semantic intent can include stage-specific keywords in the user's input. For example, a user inputting "polish it again," "change the style," etc. (indicating editing intent), "export PPT," etc. (indicating sharing intent); or, stage-specific keywords can be obtained based on the user's historical operational behavior analysis. For example, if a user highlights "market share increased by 32%" in a report and adds the annotation "for Q3 report PPT," the system can identify stage-specific keywords such as "PPT" and "report" through semantic analysis, clearly indicating the user's intent to create a presentation, etc., so as to understand task coherence based on input semantic intent. Optionally, this understanding process can also be combined with the logical complexity of the input (such as the complexity gradually decreasing from "draft writing" to "format adjustment," etc., indicating task coherence).

[0128] Therefore, this embodiment can analyze the user's operation trajectory to determine whether the user is at a task node such as "collecting materials," "creating content," or "preparing to share" (inferring the current task). It can also predict the user's most likely next action based on historical behavior patterns, thus determining the direction of subsequent executable tasks. Since the input semantic intent directly reveals the user's specific expectations for the current task, the depth of intent can be understood, allowing for accurate prediction of subsequent executable tasks. Therefore, interactive behavior data enables the system to understand the process of the task and the user's macro-goals, not just the static content displayed on the source application interface; it is a key input for generative models to predict subsequent executable tasks.

[0129] The scene content data in the task context data includes or represents the content features or data features of the interface output before the screen shape changes. It reflects what the user saw and captures the user's current task focus and operation object to determine the current task and predict subsequent executable tasks through semantic understanding. In this application, the scene content data is not just a screenshot (a static, instantaneous, context-oriented snapshot that records the "situation" on the user interface at the moment the screen shape changes), but a deep analysis of the information output by the last frame of the interface before the screen shape changes. Among them, the content features can represent the type of content displayed on the interface, such as an article paragraph being read, an image being edited, or selected text, thereby determining whether the currently displayed content is "a long article," "a product details page," "a chat log," or "a video playback interface," thus determining the type of the current task (reading, shopping, communication, entertainment) and the mode of subsequent executable tasks (such as video content triggering "summary generation," and text content triggering "deep reading"). Data features can be data that users pay attention to / annotate in the interface content, such as whether the highlighted data in the interface is "address," "date," "name," "link," or "amount," which determines the target of subsequent executable tasks; it can also include the nature and format of the content, such as text, images, tables, PDFs, code, and chat logs. This application does not limit the feature content and representation methods included in each of the content features and data features.

[0130] Therefore, this application, by analyzing scene content data, can identify the task object (knowing what specific content the user is currently focusing on), determine the content type (different content types trigger different subsequent processing (text → polishable, image → editable, table → analyzable), connect the context, and provide specific source data for subsequent executable tasks. Furthermore, scene content data can complement the aforementioned interaction behavior data; that is, interaction behavior data provides the historical context and intent guidance of the task, while scene content data provides the current focus and operation object of the task. The combination of these two allows the system to comprehensively understand the current task from two dimensions: "how the user got to this point" and "what the user is looking at at this point," providing a complete data foundation for generating execution plans for subsequent executable tasks.

[0131] Task state data in task context data can include or represent at least one of the following: the execution process of the current task, generated intermediate results, or the interaction session state with artificial intelligence services. It reflects where the user is / where the task is progressing, accurately locating task progress and ensuring seamless migration of task progress. Users do not need to start from the beginning on a changed screen format; they can directly continue from the "interruption point" or "intermediate result," greatly improving the continuity of complex, long-cycle tasks. Generated intermediate results can include at least one of the following: inference results of executed tasks, user-edited content, intermediate files, temporary data, such as a written summary draft or adjusted image filters. These, along with the execution process of the current task (e.g., the section being edited in a document, the minute and second being edited in a video), jointly determine the completeness and continuation of the current task.

[0132] The interactive session state of the AI ​​service can include at least one of the following: multi-turn dialogue history (dialogue context memory) between the user and the AI ​​service (such as an AI agent or other intelligent assistant); submitted instructions and parameters (the original natural language instructions contained in each historical request issued by the user to the AI, as well as structured parameters carried by the user through interface interaction, such as format requirements, style settings, constraints, reference content or attachments); and the AI's historical response content (such as the output given by the AI ​​service in response to the user's input instructions, which serves as the operation object for subsequent executable tasks). Therefore, this application can treat an incomplete AI dialogue (such as "Help me think of a title, I haven't decided which one to use yet") as a subsequent task to be executed. Furthermore, by collecting the interactive session state, it indicates that the agent or other intelligent program was invoked before the screen shape changed, and when the screen shape change was perceived, task context data was collected through methods not limited to those described above.

[0133] In summary, this application integrates interactive behavior data, scene content data, and task status data to construct a three-dimensional, panoramic understanding of tasks, making task migration no longer a simple "interface relocation," but a "holistic migration of memory, state, and capabilities." In this implementation, by analyzing "cross-application historical trajectories" and "input semantic intent," it distinguishes between explicit and implicit user behaviors. For example, if a user receives and opens a file in WeChat (interactive behavior data), and the file content is a contract (scene content data), combining these two dimensions of data allows for a high-accuracy inference of the task intent of "reviewing a contract," rather than simply "browsing a file." Furthermore, by limiting task status data and triggering the summarization, continuation, or deepening of incomplete AI dialogues through screen shape changes, it opens a new model of "physical interaction driving AI workflow." This not only solves the core pain point of users being forced to interrupt deep collaboration with AI when switching devices or screen shapes, but also, by preserving complete instructions, responses, and context, allows the non-deterministic, context-dependent creative process to be suspended and resumed, achieving task flow continuity.

[0134] In the task migration method described in the above embodiments, one possible implementation method for generating task planning information based on task context data is to perform semantic analysis on the task context data to obtain corresponding semantic analysis results, and then match the semantic analysis results with a predefined task flow template to generate task planning information (such as through a generative model). The task flow template can consist of multiple logically connected task nodes, and its general task processing paradigm can be obtained through statistical / analysis of historical task processing records and experience in the corresponding scenario. This application does not impose any restrictions on the structure of the task flow template.

[0135] Based on the above description of the task context data, this embodiment can perform deep semantic analysis on it, such as using natural language processing techniques or pre-trained models to identify the deep semantics contained in the data. Optionally, at least one of named entity recognition, sentiment analysis, topic modeling, and intent recognition can be performed on the text content in the task context data; and at least one of OCR (Optical Character Recognition), scene recognition, and object detection can be performed on the interface screenshots or displayed content; pattern mining can also be performed on historical operation trajectories, user operation sequences, etc., to achieve task correlation analysis, so as to obtain semantic analysis results containing, but not limited to, structured and matchable multi-dimensional semantic features such as user intent, content type, core keywords, accumulated materials, and task connection points, providing accurate matching basis for subsequent template matching.

[0136] Optionally, this application can input the semantic analysis results into a template matching engine, and match the semantic features of each dimension contained in the semantic analysis results with the triggering conditions / task features of each task node contained in the task flow template, and determine the current task and subsequent executable tasks based on the matching degree. When the above task flow template is a general template for the same scenario, this matching method can accurately locate the specific task node currently in use to determine the current task, and can also determine the task nodes to be executed that connect to the current task node, so as to predict the tasks that the user may need to execute in the future (i.e., subsequent executable tasks). In other words, subsequent executable tasks correspond to the task nodes to be executed in the task flow template.

[0137] Based on the above analysis, when the predefined task flow template is a standardized process (task node sequence) for different scenarios, and it is: generate task node → edit task node → review task node → share task node, the content type / data type obtained from semantic analysis based on task context data can be used to determine which task node best matches the user's current actual progress, and thus determine the task node corresponding to the current task. For example: if the user is editing a generated image before the screen changes, semantic analysis can determine that the current task is the editing task node (the most matching task node), and combined with task context data, the current task is specifically an image editing task; if the user jumps from the generation interface to the WeChat dialog box, it can be determined that the current task is the sharing task node; if the user zooms in on the generated result, it can be determined that the current task is the review task node, and so on. Then, the semantic analysis results can be used to instantiate it as the current task. Furthermore, based on the logical order between the task nodes in the task flow template, the necessary task nodes in the task flow can be determined, such as one or more unexecuted task nodes connected to the current task node, to predict subsequent executable tasks that are logically consistent and highly feasible with the current task, adapting to actual scenarios and ensuring the integrity and standardization of the task flow logic. In particular, for the same task node in the task flow template, the instantiated task content is often different in different scenarios to form a personalized task flow.

[0138] For example, as shown above Figure 7As shown, a customer sends a message containing a request to create a PowerPoint presentation using instant messaging software. After quickly reading the message on the outer screen, the user expands the screen to enter the inner screen mode. The agent detects the change in screen mode, obtains task context data containing at least the dialogue context memory, and matches its semantic analysis results with each task node in the task flow template. If the matching degree does not exceed the matching threshold, the AI ​​workflow will not be triggered, and the instant messaging software's dialogue interface can continue to be displayed on the inner screen. If the matching degree exceeds the matching threshold, it means that the AI ​​workflow needs to be triggered, that is, the corresponding task flow needs to be generated, and the current task node is determined based on the matching degree to determine the current task. For example, in the outer screen mode, the agent has already quickly identified the message type (request or task). Extracting key information such as categories, deadlines, and output objects (e.g., requirements, deadlines, output objects) and key fields, a summary is obtained containing key information (core viewpoints of events / content), key figures (core roles / subjects), and key milestones (summarizing important time nodes or stages). This indicates that the summary generation task corresponding to the task generation node is an executed task. At this point, a floating layer at the bottom of the outer screen can display a message: "Summary generated, expand to continue processing," prompting the user to expand the screen and enter the inner screen mode. Optionally, before expanding, the user can touch the "Create AI Workflow" function button displayed on the outer screen, or input voice or operation commands to create an AI workflow, so that the corresponding task flow can be created and displayed directly according to the method proposed in this application after the screen is expanded.

[0139] During the above process, the task nodes can be arranged according to the logical order of the task flow template, or the PPT creation task corresponding to the editing task node can be designated as the current task. Based on the generated summary, a structured PPT framework can be automatically generated, transforming abstract key points into a demonstrable content structure, i.e., generating PPT content / outline. The reasoning results or execution process of the corresponding task are displayed in the task management window output in the inner screen mode, such as... Figure 7 As shown, the generated summary content is displayed in the summary window, and the PPT content / outline is displayed in the PPT creation window in the form of a key point list (a control entry point). It can also display charts, diagrams, or key illustrations that can be inserted into the PPT. Users can trigger AI to automatically generate the corresponding PPT by touching the control entry point. Afterwards, it can further predict the message reply task corresponding to the sharing task node (the next executable task), providing a corresponding "Reply Message" text label button (a control entry point). After completing the summary and PPT creation, clicking the "Reply Message" button will call the email sending API or instant messaging software to automatically generate a notification email or instant messaging message, synchronizing the created PPT to relevant personnel and completing the entire task flow loop. Optional, such as... Figure 7As shown, this application can also display the progress of the entire task flow in the form of a progress bar at the top of the interactive interface, such as... Figure 7 The four circles in the middle represent four task nodes in the task flow template. When entering a task node, the corresponding circle will be highlighted (or displayed in other differentiated ways) so that the user can clearly see the current task node, but this is not limited to this.

[0140] Similarly, such as Figure 8 As shown, a user takes a photo under the outer screen of their phone, hoping to capture inspiration for jewelry design. During the shooting process, the system can perform visual understanding on the real-time captured image. Then, the user unfolds the screen, intending to create a jewelry design. The intelligent agent senses the change in screen shape and acquires task context data containing at least the captured image and the visual understanding results. Based on this, through task node matching, it determines that the current state is a task generation node, requiring the triggering of the AI ​​workflow (which can also be called the AI ​​creation flow, i.e., the task flow). According to the task migration method proposed in this application, in the inner screen state, the source application interface can continue to be displayed, showing either the last frame image captured on the outer screen or directly displaying the last captured frame image, such as... Figure 8 The image shown is a real-life photograph of a firefly, its tail glowing. This serves as the source of visual inspiration, generating the initial creative material for the AI's creative flow. The AI ​​then uses this real visual element to generate designs. Simultaneously, the interactive interface on the right side of the inner screen displays the AI's creative flow. Its top task management window shows the reasoning results of the current task, such as the creative concept / target intent obtained through semantic analysis. For example, the core creative concept of designing jewelry inspired by firefly glowing is shown, thus transforming visual inspiration into textual creative direction, providing clear instructions for subsequent AI generation.

[0141] Based on the above analysis, the various control entry points provided in this application corresponding to the task of creating jewelry design sketches with fireflies as the theme can visualize the creative ideas through a generative model (AI image generation model) and output several jewelry design templates that the user may be satisfied with (such as...). Figure 8 The system provides four design drafts (e.g., to be selected by the user) for selection and can also display the generation process. Input boxes or suggestion tags can be provided as needed to supplement and modify creative instructions, allowing users to directly trigger the corresponding tags or input instructions directly into the input boxes to call the model and regenerate jewelry designs that match the user's actual intentions. After reviewing and confirming the jewelry design drawings required by the user, the generated jewelry design drawings and creative concepts can be packaged and exported by triggering the "Export Creative Results" control entry (corresponding to the sharing task node) for subsequent design, production, or sharing. Optionally, during this implementation, the sharing service associated with this control entry (such as social platform interfaces / social software / work software, etc.) can be called to directly share the exported content to the corresponding platform, completing the creative delivery.

[0142] In other application scenarios, the task migration method proposed in this application uses screen shape changes as the trigger source for AI workflows. Based on the currently collected task context data, it dynamically orchestrates and adapts AI workflows (task flows). Upon entering the changed screen shape, a visualized AI workflow takes over the current task, predicts subsequent executable tasks, and deploys adaptive task modes to ensure continuous and contextually consistent interactive experience. This achieves intelligent task migration, avoiding misjudgments of task nodes and unnecessary jumps. Furthermore, through the system scheduling layer, when task nodes change, the inference results of the current task are not interrupted. Instead, candidate executors corresponding to the next task are prepared directly, and the corresponding control entry points are displayed, improving the convenience and diversity of interaction.

[0143] In some embodiments, this application can also configure corresponding task flow templates for several common scenarios. For example, a task flow template suitable for the scenario of organizing meeting minutes: key point extraction → to-do list organization → responsibility allocation → minutes output; a task flow template suitable for the scenario of writing technical solutions from requirements documents: requirements analysis → architecture design → module division → technology selection → solution output; a task flow template suitable for the scenario of creating presentation PPTs from reports / documents: material extraction → outline generation → data chart creation → conclusion page insertion → format beautification. In this way, based on the scenario type contained in the semantic analysis results, the pre-configured task flow template for the corresponding scenario can be determined first. Then, according to the method described above, the task node currently in the task flow template (which can be any task node, so that no matter which link / task node the user interrupts from, the corresponding position can be found in the task flow template, and the correct guidance from the task node can be provided to continue the task anytime and anywhere), and its logically continuous unexecuted task nodes, specifically generating task planning information that conforms to the user's true intention.

[0144] In this system, the task flow templates define a standard sequence of task nodes, allowing users to clearly understand the tasks to be executed next, thus enhancing the system's predictability and user trust. This template matching method can complete task planning within milliseconds, eliminating the need for complex real-time generation calculations and ensuring fast response times. Furthermore, it eliminates the need for large model inferences, placing lower demands on the computing power of terminal devices, making it suitable for efficient operation on mobile devices. Additionally, developers can continuously enrich the template library based on business scenarios, covering more task types and improving the system's adaptability to different scenarios.

[0145] Preferably, in the process of generating task planning information, this application can use a generative model. In this case, task path constraints can be determined based on the task flow template, so that the generative template generates instantiated task planning information according to the logical order under the constraints. For example, the generative model needs to generate specific tasks corresponding to the task nodes in the task flow template, without skipping intermediate task nodes when predicting subsequent executable tasks, and finally outputting task planning information that has both a standard logical order and specific execution content. It can be seen that this method of task flow template matching and generative model collaboration solves the problems that may arise when using only a generative model, such as generating tasks with different logic (e.g., sharing before editing) or missing key steps, ensuring that the generated task planning information is rigorous, compliant and in line with user habits in terms of process, greatly improving the stability of system execution. Moreover, by predicting the direction of the task flow through a standardized task flow template, the interpretability of the task flow is realized, and when certain standard task nodes are not needed, the task node can be located and removed based on the template structure, which facilitates the user's visual intervention in the task flow.

[0146] Furthermore, unlike rule-based matching which can only handle fixed processes, this embodiment introduces a generative model. Based on task context data, it fills in the specific content of corresponding task nodes within the framework of the task flow template, dynamically generating corresponding tasks, such as predicted subsequent executable tasks. This design, with a fixed structure and flexible content, ensures both the standardization of the task flow and the personalization of the scenario. Moreover, by determining the framework result of the actual task flow through task flow template matching, the generative model only needs to focus on the instantiation details, reducing the computational resource consumption of the model, improving inference speed, and making the task transition response after screen changes more rapid, thus enhancing the smoothness of the user experience.

[0147] In some embodiments, the task planning information generated by any of the above embodiments may further include: determining corresponding candidate executors for the predicted subsequent executable tasks. The candidate executors may include, but are not limited to, at least one type such as callable applications or functional services, applicable content templates or data templates, and data objects to be processed. Thus, when generating task planning information, based on task context data, the system not only plans what to do but also pre-determines what to use to do it. This process can be the aforementioned process of mapping abstract task nodes to specific system capability units. It can also be achieved by querying platforms / databases such as service registries, capability graphs, or application stores to determine suitable execution carriers for each predicted subsequent executable task, identifying them as candidate executors to provide corresponding control entry points in the changed screen format, assisting users in intuitively selecting the required executor to complete the task.

[0148] In this embodiment, the callable application or functional service (“tool”) refers to the software capability unit (which can be programmatically invoked) or service interface (input / output interface) required to perform the task (corresponding subsequent executable task). The application can be a complete software APP provided by an app store or other platform, such as a WeChat app, image editing app, or document app; the functional service can refer to a more granular system service or API interface than the application. For example, it is not necessary to launch the entire translation app; only the “system translation service interface” needs to be called, or the “AI large model text polishing interface” needs to be called. Accordingly, for the subsequent executable task of “translating a document,” the candidate executor may be the xx translation app or the system's built-in translation service. The system service can be a specific data processing service provided internally by the enterprise or by a third party, such as AI generation service, data format conversion service, cross-application sharing service, and data analysis service; the API (Application Programming Interface) can include local APIs (such as the operating system's “share” service, file manager, etc.) and cloud APIs (such as AI model-based generation services, map services, payment services, etc.), which can be flexibly selected according to actual application needs.

[0149] The aforementioned applicable content templates or data templates (“templates”) are structured frameworks or format templates required to perform tasks, used to standardize the output or input formats of the tasks. Content templates guide the style of the generated content. For example, for the task of “creating a poster,” candidate executors could include “poster design templates”; for the task of “generating a PPT outline,” candidate executors could include “business report outline templates” (which include sections such as background introduction, market analysis, competitive landscape, and strategic recommendations); for the task of “creating data chart slides,” candidate executors could include “trend chart templates” (which include sections such as title area, data visualization area, and conclusion area), etc. Data templates can be used to standardize data flow formats, and can be determined based on at least one dimension such as style, visual effects, and layout. For example, for the task of "creating a PPT", the candidate executor could include "business report PPT template" or "minimalist PPT template". In this application, these content templates and data templates can themselves serve as a control entry point, allowing users to intuitively select and execute the corresponding task. For instance, the context data of the corresponding task can be automatically filled or applied to the corresponding position of the template to obtain a template instance, thus reducing the cognitive load on users after switching between different formats.

[0150] The aforementioned data objects to be processed can be input data / specific materials for task operations. They are the material basis for task execution and are typically raw or generated intermediate data extracted from task context data that has not yet been processed. These can include, but are not limited to, any form such as a message to be replied to (the message extracted from chat history that requires a user response), a summary to be organized (a set of key paragraphs extracted by AI from a long article), or data to be formatted (a bunch of unprocessed text and numbers scraped from a webpage), depending on the situation. When this application identifies the data object as the executor, it means that the system has recognized it as the core operation object of the next task (a subsequent executable task) and is ready to pass it to subsequent tools or templates. For example, for the "reply to message" task, the identified candidate executor is the text content of the "message to be replied to." Clicking its control entry point allows direct entry into the editing state of the message, thereby improving the automation level and success rate of task execution (data object ready).

[0151] Optionally, in the process of determining the candidate executors described above, one or more suitable candidate executors can be filtered, sorted, and associated from an available resource pool (which may be local, cloud-based, or a combination of both) based on the type and context of the candidate executable tasks. This process may involve semantic matching between task descriptions and executor functional descriptions; considering the adaptability of input data formats, processing capabilities, user permissions, current network environments, etc., to achieve context adaptation; prioritizing based on factors such as performance, cost, and user preferences, etc. This application does not limit the method for determining at least one candidate executor corresponding to each subsequent executable task.

[0152] Therefore, this application ensures that each subsequent executable task has all the necessary resources during execution by dynamically determining candidate executors during the task planning phase, thus preventing interruptions due to lack of tools, templates, or data. When the user clicks the control entry, the system has already pre-determined all the elements required to execute the task, achieving true "one-click completion." Users no longer need to manually select tools, find templates, or locate data. The system does not need to perform software resource searches or parameter matching again and can directly call the determined candidate executors for execution, ensuring fast task response after screen changes and achieving a seamless experience. Moreover, the data objects in the candidate executors directly originate from dynamically accumulated task context data, ensuring the continuity and personalization of task execution.

[0153] Reference Figure 9This is a flowchart illustrating the task migration method proposed in Embodiment 3 of this application. This embodiment describes a possible implementation method of how to output the corresponding task management window in the changed screen form based on the task planning information in the task migration method described above. Figure 9 As shown, the implementation method may include:

[0154] Step S91: Based on the task planning information, generate interface element information for constructing the task management window; the interface element information includes a first type of element information for representing the reasoning result or execution process of the current task; and a second type of element information for representing the control entry point corresponding to the subsequent executable tasks.

[0155] Since task planning information does not directly generate the final display interface, this embodiment will first generate structured interface element information, i.e., data structures recognizable by the UI rendering engine, such as a UI construction blueprint or a virtual DOM (Document Object Model), as an intermediate bridge from task planning information to interface presentation. Based on the above description of the content included in task planning information, the first type of element information (information flow) defines how to visualize abstract "reasoning information," which may include converting data about the current task in the task planning information into structured content recognizable by the UI rendering engine, such as reasoning structure text (text summary), progress bar data (obtained from the execution process), key data previews (thumbnail URLs (Uniform Resource Locators), etc.), status icon identifiers, etc. The second type of element information (control flow) defines the corresponding control entry point on the interface for each subsequent executable task, which may include task identifiers (such as task name), associated candidate information, entry type (such as buttons, icons, etc.), and triggering behaviors (such as calling an AI service to generate an outline and opening a PPT application), etc.

[0156] Therefore, the interface element information generated in this application decouples the logical content of task planning from the specific form of interface presentation, allowing the same set of interface element information to adapt to different screen formats (and different visual styles). The interface element information adopts a structured data format, which facilitates the addition of new element types (such as audio feedback and animated prompts) without affecting the core logic, enhancing scalability. Furthermore, as an intermediate layer, the interface element information allows developers to test the task planning logic and interface rendering logic separately, reducing system complexity.

[0157] Step S92: Generate visual feedback for the current task based on the information of the first type of elements;

[0158] In this embodiment, the first type of element information can be input into the visualization rendering engine to generate static or dynamic UI components for displaying information, serving as user-perceptible visual feedback. For example, text labels (displaying inference results), progress rings (or other types of progress display controls to show the execution process), status icons (displaying success / failure), and card lists (displaying key statistical information) can be generated, thereby transforming the abstract "inference results" and "execution process" into graphical elements. Users can quickly understand the current task status without reading long texts. Through components such as card lists and progress bars, multi-dimensional information is efficiently presented within a limited space. This visual feedback utilizes the rapid recognition capabilities of the human visual system, allowing users to obtain task progress information within milliseconds without having to read every word.

[0159] Step S93: Generate the control entry point for subsequent executable tasks based on the second type of element information;

[0160] Step S94: Output the task management window in the changed screen format, and present the visual feedback and control entry in the task management window.

[0161] In this embodiment, the second type of element can be input into the interactive control generation engine to generate an interactive control entry point (UI control), and the underlying execution logic (such as the logic of calling the corresponding candidate executor determined above) can be bound to it. The user can directly touch the control entry point to trigger the call of the corresponding candidate executor. This application does not limit the type and presentation form of the control entry point, such as the button, icon, card, controllable list item or various types of data label as shown in the example above.

[0162] Optionally, based on the above analysis, the control entry point generated in this application may be associated with (bound to) at least one of the following: the position identifier of the subsequent executable task in the task flow (task sorting), the interface description of the candidate executor to be called to execute the subsequent executable task, and the execution parameters used to call the candidate executor, determined based on the task context data. This application does not restrict the implementation method (underlying logic) of how the user touches the control entry point and uses its associated information; this information may also exist as part of the second type of element information.

[0163] The interface description of the candidate executor can be at least one of the following: call address, protocol, method, parameter format, etc., which enables the invocation and execution of the candidate executor. The execution parameters of the candidate executor can include at least one of the following: data source required to execute the subsequent executable task, output target format requirements, content or style settings, or key information extracted based on task context data. These parameters are usually related to the task flow scenario. For example, in the scenario of calling AI to generate PPT, they can include at least one parameter such as theme, deadline, style, content source, etc.; in the scenario of unfolding a large screen deep learning, they can include at least one parameter such as timeline, format, language, etc.; in the scenario of unfolding a large screen creation, they can include at least one parameter such as image data, platform parameters, tag parameters, etc. There are no restrictions on these parameters.

[0164] It should be noted that the application scenarios for AI workflows triggered by changes in screen shape (i.e., the method of this application) can be configured based on experience or data statistics, or determined through model analysis; this application does not impose any restrictions on this. Examples include: an AI workflow where a user sees inspiration while browsing social media on the outer screen → expands to the large screen for in-depth editing; an AI workflow where a user searches for information on the outer screen (light browsing) → expands to the large screen for in-depth reading or summarizing; a WeChat customer submits a request → quickly reads it on the outer screen → expands to begin processing and generate PPT / text / graphics; an AI workflow where a user watches a video tutorial on the outer screen → expands to the large screen to obtain a summary and organization of information; and temporary needs for travel / meeting / life assistance → expands to a structured task flow, etc., but not limited to these examples. In this way, when a change in screen shape is detected, the method described in the above embodiments can be used to determine whether the current scene needs to trigger the AI ​​workflow based on task context data. If so, the method of this application can be executed again, or based on its semantic analysis results, it can be determined that the current task is a certain task node in the task flow template, indicating that the AI ​​workflow needs to be triggered. The current task is determined and its subsequent executable tasks are predicted to form a task flow (AI workflow) that conforms to the current scene, which is then displayed through a visual interface.

[0165] Furthermore, this application does not restrict the generation order of the aforementioned visual feedback and control entry points. They can be generated in the order described above, or simultaneously, etc. These two generation steps can be implemented using a generative model, which is very quick. Thus, within a very short time after the screen shape changes, the generative model is invoked to process the task context data according to the above method, and the generated visual feedback and control entry points are presented in the task management window output under the changed screen shape. However, this generation method is not limited to this one. In other words, in the process of generating the corresponding task management window based on interface element information and outputting the task management window under the changed screen shape, the method described above can be followed, but is not limited to, to configure the task management window to present the aforementioned visual feedback and control entry points. The presentation method can refer to, but is not limited to, the content shown in the attached figures.

[0166] In some embodiments, the above-mentioned interface element information may also include layout information, which can be used to indicate the regional display method (such as vertical / horizontal arrangement, picture-in-picture, etc.) of the task management window and the source application interface associated with presenting the current task in the changed screen form, so that when the task management window is output, the task management window and the source application interface are presented collaboratively in the changed screen form based on the layout information.

[0167] In summary, this embodiment adopts a three-layer architecture of "planning → element information → interface presentation," decoupling business logic from interface presentation. This allows the system to flexibly adapt to different screen shapes, resolutions, and user preferences without modifying the core task planning logic. Compared to traditional hard-coding of the interface, this application dynamically generates the task management window based on task planning information. This allows the system to flexibly adjust the window content according to different task types. For example, for simple tasks, the task management window may only display a single button; for complex task flows, it may display a flowchart containing multiple steps. This mechanism ensures the universality of the task management window, enabling it to handle various complex business scenarios. Furthermore, by distinguishing between "first-type element information" and "second-type element information," a visual separation between the "status area" and the "operation / interaction area" is naturally formed on the interface. This allows users to quickly understand the current task progress and make decisions, reducing cognitive burden and ensuring that users are not distracted by cluttered information. Users can quickly locate the control entry point, significantly reducing the learning cost and error rate of interaction.

[0168] Optionally, after the tasks included in the above task flow are executed continuously to produce corresponding inference results or the final task results, they can be synchronized to the source application or the source application interface for display, so that the user can return to the external screen mode and continue to view the task results or perform other operations on the task results.

[0169] In conjunction with the task migration methods described in the above embodiments, in addition to generating a task management window and providing an operation entry point corresponding to subsequent executable tasks to facilitate user interaction, this application can also generate executable instructions for executing task flows based on task planning information, and directly run the executable instructions to call the executor corresponding to the candidate executable task (which can be selected from the candidate executor) to complete the corresponding subsequent executable tasks, so as to advance the task flow until the end. This process does not require manual operation by the user and is suitable for application scenarios that require a high degree of automation.

[0170] Reference Figure 10 The diagram shown is a structural schematic of a task migration device proposed in an embodiment of this application. Figure 10 As shown, the task migration device may include:

[0171] The acquisition module 101 is used to acquire task context data in response to changes in the screen shape of the mobile terminal;

[0172] The generation module 102 is used to generate task planning information based on the task context data; wherein the task planning information includes at least one of the inference information of the current task and the predicted subsequent executable tasks.

[0173] The output module 103 is used to output the corresponding task management window in the changed screen form based on the task planning information, so as to form an interactive interface associated with the current task.

[0174] In one possible implementation, the generation module 102 described above may include:

[0175] The first generation unit is used to generate task planning information based on the task context data through a generative model. The task planning information includes the reasoning information of the current task and the task flow formed by the subsequent executable tasks.

[0176] Based on this, the output module 103 mentioned above may include:

[0177] The first output unit is used to output the reasoning result or execution process of the current task in the task flow under the changed screen form, and to provide the control entry corresponding to the subsequent executable task.

[0178] Optionally, the above-mentioned screen form change is the switching of the foldable screen of the mobile terminal between any two forms in the closed form, unfolded form and at least one intermediate form.

[0179] In some embodiments, the task context data described above includes at least one of the following:

[0180] Interaction behavior data is used to characterize historical operation trajectories across applications, application switching behavior, or input semantic intent;

[0181] Scene content data is used to characterize the content features or data types of the interface output before the screen shape changes;

[0182] Task status data is used to characterize the current task execution process, generated intermediate results, or the status of interaction sessions with artificial intelligence services.

[0183] The task flow includes multiple consecutive tasks, and the current task is the task being executed in the task flow.

[0184] In one possible implementation, the generation module 102 described above may also include:

[0185] A semantic analysis unit is used to perform semantic analysis on the task context data and obtain corresponding semantic analysis results.

[0186] The matching unit is used to match the semantic analysis results with a predefined task flow template to generate task planning information;

[0187] The subsequent executable tasks in the task planning information correspond to the task nodes to be executed in the task flow template.

[0188] In one possible implementation, the generation module 102 may further include:

[0189] The first determining unit is used to determine the corresponding candidate executor for the subsequent executable tasks included in the task planning information;

[0190] The candidate executor includes at least one of the following types:

[0191] Callable applications or functional services;

[0192] Applicable content templates or data templates;

[0193] The data object to be processed.

[0194] In one possible implementation, the output module 103 described above may include:

[0195] The second generation unit is used to generate interface element information for constructing a task management window based on the task planning information.

[0196] The interface element information includes:

[0197] The first type of element information is used to characterize the reasoning result or execution process of the current task;

[0198] The second type of element information is used to characterize the control entry point corresponding to the subsequent executable task.

[0199] Based on this, the second generation unit mentioned above may include:

[0200] The first determining subunit is used to determine the first type of element information based on the task status data in the task context data;

[0201] The first generation subunit is used to generate a second type of element information based on the candidate executors corresponding to the subsequent executable tasks included in the task planning information.

[0202] In one possible implementation, the output module 103 may further include:

[0203] The third generation unit is used to generate a corresponding task management window based on the interface element information;

[0204] The second output unit is used to output the task management window in the changed screen form to form an interactive interface associated with the current task.

[0205] The task management window is configured to display:

[0206] Visual feedback generated based on the information of the first type of elements;

[0207] The control entry point is generated based on the information of the second type of elements.

[0208] In some embodiments, the aforementioned control entry point may be associated with at least one of the following types of information:

[0209] The position identifier of the subsequent executable task in the task flow;

[0210] The interface description of the candidate executors to be invoked to execute the subsequent executable tasks;

[0211] The execution parameters used to invoke the candidate executor, determined based on the task context data.

[0212] Optionally, in the above embodiments, the interactive interface and the source application interface associated with the current task are simultaneously presented in a segmented display format under the changed screen form; wherein, the display area of ​​the source application interface is adjusted in response to the output of the interactive interface.

[0213] In some embodiments, the above-described task migration apparatus may further include:

[0214] The instruction generation module is used to parse task planning information and generate executable instructions.

[0215] The execution module is used to run executable instructions to call the executor corresponding to the candidate executable task (which can be selected from the candidate executors) and complete the corresponding subsequent executable task.

[0216] In some embodiments, the above-described task migration apparatus may further include:

[0217] The synchronization processing module is used to synchronize the task results to the source application or source application interface associated with the screen before the screen shape changes after the task flow is continuously executed and the task results are generated.

[0218] This application also provides a computer-readable storage medium that carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any of the task migration methods provided in this application.

[0219] The computer-readable storage medium can be any available medium that an electronic device can store, or a data storage device such as a training device or data center that integrates one or more available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

[0220] This application also provides a computer program product including computer-readable instructions, which, when executed on an electronic device, cause the electronic device to implement any of the task migration methods provided in this application.

[0221] When computer-readable instructions are loaded and executed on an electronic device, all or part of the processes or functions described in the embodiments of this application are generated. The electronic device may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer-readable instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer-readable instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means, depending on the actual application scenario.

[0222] This application also provides an intelligent program (such as an agent or intelligent assistant) to sense changes in screen shape, obtain task context data, and receive user input to implement the task migration method proposed in this application. The implementation process can be referred to the description of the corresponding part of the method embodiment above. In this implementation process, other components of the application or operating system can also be controlled through interface calls or other interactive methods to respond to user input or perception results, without limitation.

[0223] Reference Figure 11 This is a schematic diagram of the hardware structure of a mobile terminal proposed in this application, as shown below. Figure 11 As shown, the mobile terminal may include, but is not limited to: a screen 111 capable of switching between at least two different physical forms, at least one memory 112, a computer program 113 stored on the memory 112, and at least one processor 115 capable of running an intelligent program 114. The intelligent program 114 can execute the computer program 113 through the processor 115 to implement the steps of the task migration method proposed in any embodiment of this application. The implementation process can be referred to the description of the corresponding part of the above embodiment.

[0224] The screen 111 may include, but is not limited to, a foldable screen, a rollable screen, or a stretchable screen. It can switch between any two of the following forms: a closed form, an unfolded form, and at least one intermediate form. The specific form definition is not limited in this application.

[0225] The intelligent program can be an AI assistant embedded in the operating system, activated by voice or specific operations (such as screen shape changes, trigger buttons displayed in the previous screen shape, or other wake-up operations). Alternatively, it can be an intelligent agent or other program that enables human-computer dialogue interaction, invoking one or more models to perform corresponding tasks as needed through interface calls or other interactive methods. Of course, the intelligent program can also have one or more models; in this case, the models are part of the intelligent program, and can be activated to perform tasks as needed during the operation of the intelligent program. It should be noted that the task migration method proposed in this application includes, but is not limited to, implementation through intelligent programs.

[0226] Furthermore, the mobile terminal may also include an input component, facilitating user interaction with the control interface displayed on screen 111. This input component may include, but is not limited to, one or more of the following: an audio receiver, a finger, a stylus, a mouse and keyboard, or a joystick. Depending on the needs, the mobile terminal may also include at least one output component, such as a speaker, to output the interface content or key information via voice playback.

[0227] It should be understood that,Figure 11 The structure of the mobile terminal shown does not constitute a limitation on the mobile terminal in the embodiments of this application. In practical applications, the mobile terminal may include more than [other components]. Figure 11 The electronic device may include more or fewer components, or combinations of certain components, and when the electronic device is a terminal device, it may also include sensing units such as gyroscopes, accelerometers and gravity sensors for obtaining sensing parameters, power management modules, antennas or other communication elements, etc., which will not be listed in detail in this application.

[0228] Finally, it should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the accompanying drawings of the device embodiments provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.

[0229] In the above embodiments, the implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented using software plus necessary general-purpose hardware, or it can be implemented using dedicated hardware including dedicated integrated circuits, dedicated CPUs, dedicated memory, dedicated components, etc. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The various embodiments in this specification are described in a progressive or parallel manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to mutually. For the apparatus and mobile terminal disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and relevant parts can be referred to the method section.

Claims

1. A task migration method, comprising: Responding to changes in the screen shape of the mobile terminal, acquire task context data; Based on the task context data, task planning information is generated; wherein, the task planning information includes at least one of the inference information of the current task and the predicted subsequent executable tasks; Based on the task planning information, a corresponding task management window is output in the changed screen format to form an interactive interface associated with the current task.

2. The method according to claim 1, wherein generating task planning information based on the task context data includes: Based on the task context data, a generative model is used to generate task planning information, which includes the reasoning information of the current task and the task flow formed by the subsequent executable tasks. Based on the task planning information, a corresponding task management window is output in the changed screen format, including: The system outputs the reasoning result or execution process of the current task in the task flow under the changed screen format, and provides the control entry point corresponding to the subsequent executable task.

3. The method according to claim 2, wherein the screen shape change is the switching of the folding screen of the mobile terminal between any two of the following: closed shape, unfolded shape, and at least one intermediate shape; The task context data includes at least one of the following: Interaction behavior data is used to characterize historical operation trajectories across applications, application switching behavior, or input semantic intent; Scene content data is used to characterize the content features or data types of the interface output before the screen shape changes; Task status data is used to characterize the current task execution process, generated intermediate results, or the status of interaction sessions with artificial intelligence services. The task flow includes multiple consecutive tasks, and the current task is the task being executed in the task flow.

4. The method according to claim 1 or 2, wherein generating task planning information based on the task context data includes: Semantic analysis is performed on the task context data to obtain the corresponding semantic analysis results; The semantic analysis results are matched with predefined task flow templates to generate task planning information; The task planning information includes subsequent executable tasks that correspond to the task nodes to be executed in the task flow template.

5. The method according to claim 1 or 2, wherein the generation of task planning information further includes: For the subsequent executable tasks included in the task planning information, determine the corresponding candidate executors; The candidate executor includes at least one of the following types: Callable applications or functional services; Applicable content templates or data templates; The data object to be processed.

6. The method according to any one of claims 1-3, wherein based on the task planning information, a corresponding task management window is output in the changed screen configuration, comprising: Based on the task planning information, generate interface element information for constructing the task management window; The interface element information includes: The first type of element information is determined based on the task state data in the task context data and is used to characterize the reasoning result or execution process of the current task. The second type of element information is generated based on the candidate executors corresponding to the subsequent executable tasks included in the task planning information, and is used to characterize the control entry point corresponding to the subsequent executable tasks.

7. The method according to claim 6, wherein outputting the corresponding task management window based on the task planning information in the changed screen format further includes: Based on the interface element information, a corresponding task management window is generated to output the task management window in the changed screen form. The task management window is configured to display: Visual feedback generated based on the information of the first type of elements; The control entry point is generated based on the information of the second type of elements.

8. The method according to claim 7, wherein the control entry is associated with at least one of the following types of information: The position identifier of the subsequent executable task in the task flow; The interface description of the candidate executors to be invoked to execute the subsequent executable tasks; The execution parameters used to invoke the candidate executor, determined based on the task context data.

9. The method according to claim 7, wherein the interactive interface and the source application interface associated with presenting the current task are simultaneously presented in a segmented display format under the changed screen form; in, The display area of ​​the source application interface is adjusted in response to the output of the interactive interface.

10. A mobile terminal, comprising: A screen capable of switching between at least two different physical forms; At least one memory, and a computer program stored in the memory; At least one processor capable of running intelligent programs; The intelligent program can execute the computer program through the processor to implement the steps of the task migration method as described in any one of claims 1 to 9.

Images