Intelligent memorandum task generation method and device, electronic equipment and storage medium

Abstract

The invention discloses an intelligent memorandum task generation method and device, electronic equipment and a storage medium, and relates to the technical field of computers, and the method comprises the following steps: obtaining natural language initial memorandum data input by a user, and carrying out semantic analysis; based on the semantic analysis result, generating a structured cue word from a predefined scene classification template library through dynamic instantiation; inputting the structured cue word into a large language model, and obtaining extended content which is output by the model and comprises a plurality of subtasks and supplementary information; analyzing the extended content into a structured task object with a set membership and a priority attribute, and displaying the structured task object; and in response to a refining request of a user for the target sub-task, triggering recursive decomposition of the target sub-task to generate a subordinate sub-task of the target sub-task. The method has the technical effects that the manual input burden of the user is relieved, the integrity and the structuring level of task planning are improved, the personalized task generation requirement of the user is met, multi-level deep refinement of the task is supported, and the task execution convenience and comprehensiveness are improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, electronic device, and storage medium for intelligent generation of memo tasks. Background Technology

[0002] With the deep integration of digital office work and daily life management, memo tools have gradually evolved from traditional text recording devices into intelligent task management platforms. Leveraging advancements in natural language processing and large-scale artificial intelligence models, memo tools are no longer limited to simple information storage; they now undertake core functions such as task breakdown, process planning, and progress tracking, becoming key tools for improving individual and team efficiency. In particular, breakthroughs in semantic understanding and content generation using large-scale language models have provided core support for the intelligent upgrade of memo tools, making it possible to automatically identify task intent and break down complex requirements.

[0003] However, existing memo task generation solutions still have many limitations, making it difficult to meet users' needs for intelligent, structured, and flexible task management: they lack deep semantic mining capabilities for the initial natural language memo content input by users, failing to accurately capture core intent; prompt word generation lacks dynamic adaptability, relying heavily on fixed templates and making it difficult to match diverse scenarios; the extended content output by the large language model lacks completeness and relevance, and cannot be effectively parsed into a structured form with clear parent-child relationships and priority attributes; at the same time, they do not support further recursive refinement of generated subtasks, making it difficult to adapt to the multi-level decomposition requirements of complex tasks, resulting in heavy manual input burdens for users, inadequate task planning, and low execution efficiency. Summary of the Invention

[0004] This application provides a method, apparatus, electronic device, and storage medium for intelligent generation of memo tasks. It can solve the problems of low efficiency, omission of details, and poor adaptability in related technologies, caused by task content relying entirely on manual input, rigid generation methods unable to be personalized, lack of clear hierarchical relationships and priority divisions, and difficulty in deeply refining tasks.

[0005] According to a first aspect of this application, a method for intelligently generating memo tasks is provided, comprising: Obtain initial memo data in natural language form input by the user, perform semantic analysis on the initial memo data, and generate semantic analysis results; Based on semantic analysis results, structured prompt words are dynamically instantiated from a predefined scene classification template library; Input the structured prompts into the large language model and obtain the extended content output by the large language model, which includes multiple sub-tasks and supplementary information; The extended content is parsed into structured task objects with parent-child relationships and priority attributes, and then the structured task objects are displayed. In response to a user's request for more detailed information about the target subtask, a recursive decomposition of the target subtask is triggered, generating a subtask of the target subtask.

[0006] According to a second aspect of this application, a memo task intelligent generation device is provided, comprising: The first generation module is configured to acquire initial memo data in natural language form input by the user, perform semantic analysis on the initial memo data, and generate semantic analysis results. The second generation module is configured to dynamically instantiate and generate structured prompt words from a predefined scene classification template library based on semantic analysis results. The acquisition module is configured to input structured prompts into a large language model and acquire extended content output by the large language model, which includes multiple subtasks and supplementary information. The parsing module is configured to parse the extended content into structured task objects with parent-child relationships and priority attributes, and then display the structured task objects. The response module is configured to respond to user requests for more detailed information about the target subtask, triggering a recursive decomposition of the target subtask and generating its subordinate subtasks.

[0007] According to a third aspect of this application, an electronic device is provided, comprising: At least one processor; and memory that is communicatively connected to at least one processor; The memory stores instructions that can be executed by at least one processor, which are executed by at least one processor to enable the at least one processor to perform the aforementioned memo task intelligent generation method of the first aspect.

[0008] According to a fourth aspect of this application, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause a computer to execute the memo task intelligent generation method of the first aspect described above.

[0009] According to a fifth aspect of this application, a computer program product is provided, including a computer program that, when executed by a processor, implements the memo task intelligent generation method as described in the first aspect above.

[0010] This application provides a method, apparatus, electronic device, and storage medium for intelligent generation of memo tasks, comprising: acquiring initial memo data in natural language form input by a user; performing semantic analysis on the initial memo data and generating semantic analysis results; dynamically instantiating structured prompt words from a predefined scene classification template library based on the semantic analysis results; inputting the structured prompt words into a large language model to obtain extended content output by the large language model, which includes multiple sub-tasks and supplementary information; parsing the extended content into structured task objects with parent-child relationships and priority attributes, and displaying the structured task objects; and responding to a user's request for refinement of a target sub-task, triggering recursive decomposition of the target sub-task to generate lower-level sub-tasks of the target sub-task. This application addresses the challenges of obtaining initial memo data in natural language input from users and performing semantic analysis. Based on this analysis, structured prompts are dynamically instantiated from a predefined scenario classification template library. These prompts are then input into a large language model to obtain extended content containing multiple sub-tasks and supplementary information. This extended content is then parsed into structured task objects with parent-child relationships and priority attributes and displayed. Furthermore, the application responds to user requests for refinement of target sub-tasks, triggering recursive decomposition of the target sub-tasks to generate lower-level sub-tasks. Therefore, it solves the problems of inefficiency, omissions of details, and poor adaptability in related technologies, which suffer from tasks relying entirely on manual input, rigid generation methods that cannot be personalized, lack of clear hierarchical relationships and priority divisions, and difficulty in deeply refining tasks. This achieves the technical effects of reducing the burden of manual input for users, improving the completeness and structure of task planning, meeting users' personalized task generation needs, supporting multi-level deep refinement of tasks, and improving the convenience and comprehensiveness of task execution.

[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0012] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A flowchart illustrating a method for intelligently generating memo tasks provided in an embodiment of this application; Figure 2 A flowchart illustrating another intelligent memo task generation method provided in this application embodiment; Figure 3 A flowchart illustrating another intelligent memo task generation method provided in this application embodiment; Figure 4 This is a schematic diagram of the structure of a memo task intelligent generation device provided in an embodiment of this application. Detailed Implementation

[0014] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0015] The following description, with reference to the accompanying drawings, outlines a method, apparatus, electronic device, and storage medium for intelligently generating memo tasks according to embodiments of this application.

[0016] Figure 1 This is a flowchart illustrating a method for intelligently generating memo tasks provided in an embodiment of this application.

[0017] like Figure 1 As shown, the method includes the following steps: Step 101: Obtain initial memo data in natural language form input by the user, perform semantic analysis on the initial memo data, and generate semantic analysis results.

[0018] In some embodiments, initial memo data in natural language form refers to task descriptions entered by users in a conversational, unstructured text format commonly used in daily communication. This type of data does not need to follow specific formatting specifications and can cover to-do items in various scenarios such as life, work, and study, such as "attend a project review meeting next Friday," "prepare a family dinner this weekend," and "complete the report writing by the end of the month." Its core characteristic is that it directly reflects the user's task intent and can be entered without special formatting. The process of acquiring this initial memo data can be completed through the interactive interface of the terminal device. The integrity of the original text is maintained during data transmission to ensure that key information is not lost in subsequent analysis. Semantic analysis is a process of deep parsing of the acquired initial memo data based on natural language processing technology. It goes beyond simply extracting keywords and uncovers the core intent, key entities, potential needs, and scenario attributes behind the text. Specifically, the text is first segmented, tagged with parts of speech, and analyzed using grammatical structure to eliminate redundant modifiers and accurately locate core actions and related objects. For example, from "attend a project review meeting next Friday," the core intent is identified as "participate in the review," the key entity as "project review meeting," and the time attribute as "next Friday." Then, combined with a scene feature library, the task's scene category is determined, such as travel, social, or work scenarios. Simultaneously, ambiguous information or implicit needs in the text are identified; for example, "preparing for a trip" implies potential needs such as "planning the itinerary" or "purchasing supplies." Finally, these analysis results are integrated to form a semantic analysis result, which clearly presents the core intent, key entities, scene category, and related attribute information of the initial memo data. This step, through precise semantic analysis, effectively captures the user's true needs, avoiding misjudgments caused by relying solely on surface keywords. This provides an accurate basis for the subsequent generation of structured prompts, improving the targeting and effectiveness of the entire task generation process.

[0019] Step 102: Based on the semantic analysis results, dynamically instantiate and generate structured prompt words from a predefined scene classification template library.

[0020] In some embodiments, the predefined scenario classification template library refers to a pre-built collection of prompt word templates categorized and stored according to life event scenario types (such as travel, social commemoration, shopping, work projects, study plans, etc.). Each scenario category's templates include task decomposition logic, guidance direction, and output format specifications adapted to that scenario. The templates reserve placeholders corresponding to the semantic analysis results, such as scenario core element placeholders and task type placeholders, which can be dynamically filled according to actual needs. The semantic analysis results provide the core basis for matching and instantiating scenario classification templates. The information they contain, such as the task's scenario category, core intent, and key entities, can accurately locate the most suitable target template in the template library, avoiding blind template selection. The process of dynamically instantiating and generating structured prompts is based on specific information from semantic analysis results. It involves personalized filling and logical adaptation of the target template, rather than simply calling a fixed template. Instead, the template content is adjusted according to the actual situation of the user's initial memo. For example, when the semantic analysis results show that the task scenario is "travel and business - banking business", the core intent is "transfer", and the key entities are "bank, ID card, bank card", the corresponding banking business prompt template will be matched from the scenario classification template library. The above information will be filled into the placeholders of the template. At the same time, the guidance logic will be adjusted according to the characteristics of the transfer business, and the large language model will be required to generate content focusing on business confirmation, material preparation, time planning and other dimensions. Structured prompts are text messages with clear formatting requirements, logical coherence, and specific guidance instructions. They not only guide the direction of task decomposition but also specify the structural norms of the output content. For example, they require the output to include fields such as subtask name, priority, and execution conditions, and to be presented in a preset data format (such as JSON or lists) to ensure the consistency and parsability of the extended content output by the large language model. This step, through precise matching and dynamic instantiation of scene classification templates and semantic analysis results, ensures that the generated structured prompts closely match the user's actual needs, effectively guiding the large language model to perform targeted task decomposition and content generation. This avoids the lack of adaptability of traditional fixed prompts, laying the foundation for the accuracy and structure of subsequent extended content and improving the efficiency and reliability of the entire task generation process.

[0021] Step 103: Input the structured prompt words into the large language model and obtain the extended content output by the large language model, which includes multiple sub-tasks and supplementary information.

[0022] In some embodiments, structured prompts serve as input data for the large language model. Containing explicit guidance logic, scenario constraints, and output format requirements, they provide the large language model with precise task execution direction, preventing the model-generated content from deviating from user needs. The large language model is a core AI component with a vast amount of common sense, scenario-based rule reserves, and powerful logical reasoning capabilities. It can deeply understand the core instructions in the structured prompts and, based on its understanding of task execution processes in various life and work scenarios, extend and expand upon the initial memo requirements. During data transmission, the structured prompts are input in a text format that the model can directly parse, ensuring that all constraints, guidance directions, and output specifications are transmitted completely without information loss or deviation. After receiving the input, the large language model first accurately analyzes the scenario category, core task intent, and output structure requirements in the prompts. Then, it invokes its built-in knowledge graph and reasoning algorithms to break down multiple logically related subtasks from the entire process of task initiation, execution, and completion, while supplementing information such as prerequisites, precautions, and resource requirements for completing each subtask. Here, subtasks refer to specific operational items that can be executed independently and closely adhere to the core requirements of the initial memo, without redundant or irrelevant content. Supplementary information effectively supports each subtask, helping users avoid risks and improve efficiency during execution. For example, for the structured prompt "Go to the bank tomorrow," the large language model combines the general process and common scenarios of banking transactions to generate subtasks such as "Confirm the specific business type" and "Prepare your ID card and bank card," and supplements supplementary information such as "Check the branch's lunch break time" and "Determine if an appointment is needed in advance," ultimately forming extended content containing the above. This step leverages the intelligent reasoning capabilities of the large language model to generate comprehensive and practical extended content without manual user intervention, effectively solving the problems of incomplete task decomposition and missing supplementary information in traditional solutions. It provides rich and high-quality basic data for subsequent structured task processing, improving the efficiency and comprehensiveness of task planning.

[0023] Step 104: Parse the extended content into a structured task object with parent-child relationship and priority attributes, and then display the structured task object.

[0024] In some embodiments, extended content refers to non-standardized text data output by the large language model based on structured prompts, containing multiple sub-tasks and supplementary information. Its format may be common text formats such as JSON strings or Markdown lists. The core content includes specific operation items related to the initial memo content and supplementary information to assist execution. A structured task object refers to a task data carrier with clear data relationships and attribute identifiers. The parent-child relationship is reflected in the initial memo content as the parent task, and the extended sub-tasks as its subordinate related tasks, forming a clear hierarchical logic. Priority attributes are identifiers based on the urgency and importance of task execution, such as high, medium, and low levels, used to guide users to prioritize critical tasks. The parsing process first requires validating the extended content, filtering out logically consistent and non-redundant or conflicting subtasks and supplementary information, and excluding data with format errors or invalid content. Then, based on the initial memo content and the inherent logical relationship between each subtask, a correspondence between parent and child tasks is established, clarifying the subordinate status of each subtask. Simultaneously, priority information contained in the extended content is extracted or confirmed. If the extended content does not directly indicate priority, it is supplemented and improved based on common-sense logic of the task scenario, ensuring that each subtask has a clear priority attribute, ultimately integrating to form a structured task object. The presentation process must present the structured task object in an intuitive and easy-to-understand form, clearly demonstrating the hierarchical relationship between parent and child tasks and the priority differences of each subtask, allowing users to quickly identify the subordinate relationships and execution order between tasks. This step, by transforming non-standardized extended content into structured, hierarchical task data and displaying it clearly, solves the problems of messy content, lack of clear relationships and priorities in traditional task generation schemes, improves the organization of task management, facilitates users to quickly grasp the overall task situation and rationally arrange the execution order, and reduces the risk of task omissions or execution chaos.

[0025] Step 105: In response to the user's request for refinement of the target subtask, a recursive decomposition of the target subtask is triggered, generating the lower-level subtasks of the target subtask.

[0026] In some embodiments, a target subtask refers to a specific subtask selected by the user from the displayed structured task objects with parent-child relationships and priority attributes, which requires further subdivision to obtain more detailed operational guidance. It is itself a subordinate task formed after the initial decomposition of the initial memo content, possessing a clear task core and execution direction. The user's request for refinement of the target subtask is a further decomposition instruction initiated by the user based on actual execution needs, addressing the current lack of granularity and specific operational guidance of the target subtask. The core of this request is the user's expectation of obtaining more detailed operational items that better fit the actual execution scenario. Upon receiving this refinement request, the system first accurately identifies the complete information of the target subtask, including its task content, its parent task relationship, and priority attributes, ensuring that the subsequent decomposition process does not deviate from the core intent of the task. Recursive decomposition refers to the system treating the target subtask as new initial input data and repeatedly executing the complete process of semantic analysis, structured prompt word generation, large language model invocation, and extended content parsing, forming a cyclical task decomposition mechanism. This means treating the execution requirements of the target subtask as a new analysis object, using semantic analysis to uncover its underlying execution steps, preparatory steps, operational details, and other deep requirements, then dynamically generating appropriate structured prompt words based on the analysis results. These are then input into the large language model to obtain extended content containing multiple subdivided operation items, ultimately parsing it into a structured task object with a clear parent-child relationship (the target subtask is the parent, and the newly generated subtasks are the subordinate subtasks). For example, if the target subtask is "prepare a barbecue grill and charcoal," the subordinate subtasks generated after recursive decomposition may include "check if the barbecue grill is intact," "confirm if the charcoal supply is sufficient," and "prepare an igniter and combustion aids," etc. These subordinate subtasks directly correspond to the completion steps of the target subtask, offering finer granularity and greater operability. This step, through a recursive decomposition mechanism, enables multi-level splitting of tasks from coarse to fine, solving the problem that traditional task decomposition can only be split once and cannot meet the detailed requirements of complex needs. It makes task planning more in line with the user's actual execution scenario, significantly improving the operability of task execution and the comprehensiveness of planning.

[0027] Compared with related technologies, in this embodiment, initial memo data in natural language form input by the user is obtained, semantic analysis is performed on the initial memo data, and semantic analysis results are generated. Based on the semantic analysis results, structured prompt words are dynamically instantiated from a predefined scene classification template library. The structured prompt words are input into a large language model to obtain extended content output by the large language model, which includes multiple sub-tasks and supplementary information. The extended content is parsed into structured task objects with parent-child relationships and priority attributes, and the structured task objects are displayed. In response to the user's request for refinement of the target sub-task, the recursive decomposition of the target sub-task is triggered, generating the lower-level sub-tasks of the target sub-task. This can solve the problems of low efficiency, omission of details, and poor adaptability caused by the reliance on completely manual input for task content, rigid generation methods that cannot be personalized, lack of clear hierarchical relationships and priority division of tasks, and difficulty in deep refinement of tasks in related technologies. It achieves the technical effects of reducing the user's manual input burden, improving the completeness and structure level of task planning, meeting the user's personalized task generation needs, supporting multi-level deep refinement of tasks, and improving the convenience and comprehensiveness of task execution.

[0028] Figure 2 A flowchart illustrating another intelligent memo task generation method provided in this application embodiment includes the following steps: Step 201: Obtain initial memo data in natural language form input by the user, perform semantic analysis on the initial memo data, and generate semantic analysis results.

[0029] For a description of step 201, please refer to the description of step 101 in the above embodiment. This embodiment will not repeat the description in detail.

[0030] Step 202: Match the corresponding scene classification template from the scene classification template library based on the semantic intent of the initial memo data.

[0031] In some embodiments, semantic intent refers to the core user needs extracted from initial memo data through semantic analysis. It is the essential task objective behind the user's input text. For example, the semantic intent of a user inputting "organize a company team building activity next weekend" is "team activity planning," and the semantic intent of inputting "complete the annual summary report by the end of the month" is "work document writing." The scenario classification template library is a pre-built collection of prompt word templates categorized by semantic intent type. Each scenario category in the library corresponds to a core semantic intent and includes task decomposition logic, core focus dimensions, and guidance direction adapted to the scenario. For example, the "team activity planning" category template focuses on dimensions such as venue booking, process design, and participant coordination, while the "work document writing" category template focuses on data collection, framework building, review and modification. During the matching process, the system first extracts precise semantic intent tags from the semantic analysis results. Then, it searches the scene classification template library using a pre-defined indexing mechanism, such as semantic similarity algorithms and keyword mapping tables. The semantic intent tags are compared with the feature tags of each scene category in the library. When the similarity reaches a preset threshold, the corresponding scene classification template is determined. If the semantic intent is ambiguous, it is further filtered using auxiliary information from the semantic analysis results to ensure that the matched scene classification template highly matches the user's actual needs. This step, through precise matching of semantic intent and scene templates, avoids blind template selection, provides a precise foundation for the targeted generation of subsequent structured prompts, and improves the adaptability of prompts to user needs.

[0032] Step 203: Input the entity information and contextual features from the initial memo data into the scenario classification template to generate structured prompts containing task type, priority requirements, and output format.

[0033] In some embodiments, entity information refers to specific objects extracted from the initial memo data, including core entities (such as "bank," "hospital," "friend") and related items (such as "ID card," "medical examination report," "food ingredients"), which are elements with clear referential meaning. Contextual features are auxiliary information captured during semantic analysis, covering time attributes (such as "urgent," "next Friday," "end of the month"), location attributes (such as "local outlet," "suburban campsite"), participation scale (such as "small gathering of 5 people," "team building of 20 people"), and special constraints (such as "accompanied by the elderly," "fasting requirement"), etc. The scene classification template pre-sets structured placeholders (such as "[core entity]," "[time feature]," "[special constraint]") corresponding to entity information and contextual features, and each placeholder is associated with a clear filling rule. When generating structured prompts, the system first accurately fills the extracted entity information and contextual features into the template according to the correspondence between placeholders and information types. Then, it dynamically adjusts the guidance logic in the template based on the contextual features. For example, when the time feature is "urgent (e.g., tomorrow)," the guidance of "prioritizing core pre-tasks" is strengthened; when there are special entities such as "elderly" or "children," the task decomposition dimension of "adapting to the needs of special groups" is added. The final generated structured prompts must clearly include task type (e.g., "service-related," "event planning," "procurement"), priority requirements (e.g., clearly defining the core judgment criteria for high-priority tasks), and a unified output format (e.g., specifying JSON format and required fields such as "task_name," "priority," and "key hints"). This step, through the accurate filling and logical adaptation of entity information and contextual features, ensures that the structured prompts not only fit the user's specific needs but also have clear guidance and standardized format. This effectively ensures that the extended content output by the large language model is accurate, comprehensive, and easy to parse later, improving the targeting and efficiency of task generation.

[0034] Step 204: Input the structured prompt words into the large language model and receive the extended content output by the large language model engine, which includes multiple subtasks and supplementary information.

[0035] In some embodiments, the task type and priority requirements contained in the structured prompts serve as core guidelines for the large language model to generate targeted extended content. The task type clearly defines the scenario to which the initial memo data belongs (e.g., errands, shopping, social commemoration), while the priority requirements clarify the criteria for determining the importance of subtasks (e.g., core pre-task operations should be marked with high priority). Together, they set precise boundaries for the large language model's reasoning process, preventing the generation of irrelevant content that deviates from user needs. When inputting the structured prompts into the large language model, a standardized text transmission method that the model can directly parse is used to ensure that all constraints, guiding logic, and output specifications in the prompts are transmitted completely without information loss or format deviation. As a core component integrating massive amounts of common sense, scenario-based execution rules, and deep logical reasoning capabilities, the large language model, upon receiving prompts, first deeply decomposes the task type, priority requirements, and core needs implicit in the initial memo. Then, it invokes the built-in common sense reasoning mechanism—based on its inherent understanding of the entire task execution process in various scenarios (e.g., administrative tasks need to cover key stages such as preparation, verification, and execution; event planning tasks need to cover core dimensions such as venue, personnel, and materials) and the specific intent of the initial memo, gradually generating multiple logically coherent and independently executable sub-tasks. Simultaneously, the model proactively supplements information such as prerequisites, potential risk warnings, and resource adaptation suggestions for completing each sub-task. For example, for the task type "processing bank transfers," it supplements key content such as "confirming the completeness of the recipient's account information" and "allowing sufficient time for the transfer to arrive." During the generation of subtasks and supplementary information, the large language model strictly adheres to the priority requirements in the prompts. Combining the subtasks' impact on the overall goal and their urgency, it assigns a corresponding priority attribute (e.g., high, medium, and low) to each subtask. Core pre-processing tasks that determine subsequent workflow progression (e.g., preparing ID cards, confirming business type) are marked as high priority, while auxiliary and flexibly adjustable tasks (e.g., searching for restaurants near service points) are marked as medium or low priority, ensuring that priority allocation aligns with actual execution needs. Finally, the large language model integrates all subtasks, supplementary information, and corresponding priority attributes into a unified format of extended content (e.g., JSON strings, structured lists) and outputs it. This step leverages the large language model's common-sense reasoning capabilities and precise constraint response mechanism, ensuring that the generated extended content comprehensively covers key aspects of task execution while possessing clear priority guidance. This effectively solves the problems of disorganized subtasks and lack of execution focus in traditional task generation schemes, providing logically clear and focused foundational data for subsequent structured parsing, thus improving the practicality and efficiency of task planning.

[0036] Step 205: Verify whether the format of the extended content conforms to the predefined structured output specification.

[0037] In some embodiments, extended content refers to text data output by the large language model based on structured prompts, containing multiple subtasks and supplementary information. Its form may be JSON strings, Markdown lists, etc., and it primarily carries the task decomposition results and auxiliary execution information. The predefined structured output specification is a pre-set format validation standard, covering data format type (e.g., specifying JSON as the only valid format), required field requirements (e.g., core fields such as "task_name" and "priority" must be included), field data type constraints (e.g., priority fields only allow "high," "medium," and "low" values), and syntax rules (e.g., JSON format bracket pairing and comma separation specifications). This specification aims to ensure that extended content has uniform parsability. During the validation process, the system first identifies the format type of the extended content and determines whether it is consistent with the preset format; then it checks the completeness of the core fields one by one to confirm that there are no subtasks with missing key information; subsequently, it checks whether the field values ​​conform to the constraint rules and whether there are any syntax errors (e.g., JSON syntax errors, list formatting issues, etc.). For extended content that does not conform to the specification, the system will mark the format as abnormal, but this step only completes the validation judgment and does not involve content correction. This step involves rigorous format validation to filter out valid extended content that meets the parsing requirements, preventing subsequent task object generation failures due to format confusion or missing information, thus ensuring the stability and reliability of the parsing process.

[0038] Step 206: Map the subtasks and supplementary information in the verified extended content to task objects with priority fields, and establish the parent-child task relationship.

[0039] In some embodiments, mapping refers to mapping the scattered subtasks and supplementary information in the validated extended content to specified attribute fields of structured task objects according to preset rules, thereby transforming unstructured data into a standardized data carrier. A task object is a data unit with clearly defined attributes. In addition to the core "task description" field, it also includes a "priority" field, which directly uses the value of "priority" in the extended content or is completed according to preset rules (e.g., "medium" priority is the default if not marked). Supplementary information serves as an auxiliary basis for the execution of subtasks and is mapped to the "auxiliary description" field of the task object. For example, the supplementary information "avoid lunch break" for "querying branch business hours" will be stored in the "auxiliary description" of the task object. The establishment of parent-child task relationships uses the initial memo content as the core of the parent task. Through semantic correlation matching, all subtasks parsed from the extended content are marked as subordinate tasks of the parent task, clarifying the subordinate status of each subtask and forming a hierarchical relationship logic of "parent task - child task," ensuring that the affiliation between tasks is clear and unambiguous. This step transforms the originally scattered task information into structured task objects with complete attributes and clear hierarchy through precise mapping and relationship construction, providing a standardized data foundation for subsequent display and management, and improving the usability and logical coherence of task data.

[0040] Step 207: Perform semantic analysis on the refined request command and generate lower-level semantic analysis results. Based on the lower-level semantic analysis results, dynamically instantiate and generate lower-level structured prompt words from the predefined scene classification template library. Input the lower-level structured prompt words into the large language model and obtain the lower-level extended content output by the large language model, which contains multiple sub-tasks and supplementary information. Parse the lower-level extended content into lower-level structured task objects with parent-child relationships and priority attributes.

[0041] In some embodiments, a refinement request instruction is a technical instruction initiated by a user for a displayed target subtask, requesting further breakdown to obtain more specific operational guidance. Its core carriers are the identification information of the target subtask and the user's refinement request signal. The semantic analysis performed on this instruction focuses on the core intent, associated entities, scenario attributes, and potential refinement needs of the target subtask itself. The generated lower-level semantic analysis results need to be more targeted than the initial semantic analysis. For example, when the target subtask is "purchasing meat and vegetables," the lower-level semantic analysis results will clearly specify its scenario as "ingredient procurement," its core intent as "obtaining ingredients needed for barbecue," and associated entities including "barbecue ingredient types" and "freshness requirements." The predefined scenario classification template library still plays a fundamental supporting role in this step. The system will match a subdivided template suitable for the subtask scenario (such as the "barbecue ingredient subdivision" template under the "ingredient procurement" category) based on the lower-level semantic analysis results, rather than reusing the initial scenario template. The process of dynamically instantiating and generating lower-level structured prompts involves filling specific information from the lower-level semantic analysis results into a subdivided template. This explicitly guides the large language model to break down the execution details of the subtask. For example, it generates a prompt stating, "Please refine 'buy meat and vegetables' into specific ingredient types, selection criteria, and supplementary precautions, outputting them in JSON format, including the fields 'sub_task_name', 'priority', and 'supplement'." After inputting this lower-level structured prompt into the large language model, the model outputs lower-level extended content containing more granular operation items and auxiliary information based on the common-sense logic and refinement requirements of the subtask scenario. Examples include "Buy fresh beef (prioritize brisket cuts)" and "Confirm vegetables are free of pesticide residues." Finally, the lower-level extended content undergoes format validation and attribute extraction through a parsing process. It is then transformed into lower-level structured task objects with clear parent-child relationships and priority attributes, with the target subtask serving as the parent task. Priority allocation must align with the urgency of the subtask's execution; for example, "Confirm ingredient freshness" is marked as high priority. This step ensures that lower-level subtasks meet actual execution needs and have a unified structured format by focusing on precise semantic analysis and targeted prompt word generation for subtasks, thus providing standardized and accurate basic data for task hierarchy expansion.

[0042] Step 208: Establish a parent-child relationship between the recursively generated subtasks and the original task object, and update the hierarchical structure of the structured task object.

[0043] In some embodiments, the original task object refers to the structured task object corresponding to the target subtask for which the user initiated the refinement request. This object has already established a parent-child relationship with the initial memo content during the initial task decomposition, possessing clear priority attributes and task description information. The process of establishing the parent-child relationship involves binding the lower-level subtasks generated in step 207 to the original task object through a task identifier matching mechanism. This clarifies the direct subordinate of the lower-level subtasks, with the original task object serving as the parent task and all recursively generated lower-level subtasks as its direct subordinates. This forms a hierarchical relationship logic of "initial memo content - original task object - lower-level subtasks," such as a subordinate relationship chain of "preparing a weekend barbecue - buying meat and vegetables - buying fresh beef." Updating the hierarchical structure of the structured task object involves integrating the newly established parent-child relationship into the overall task system, dynamically expanding the branches of the original task tree, ensuring that the entire task structure retains the logical framework of the initial decomposition while fully presenting the refinement path of the subtasks through the added levels, without disrupting the existing relationships of other task nodes. After the hierarchical structure is updated, the parent-child relationships and priority attributes of each task node remain independent and traceable. Users can clearly identify the subordinate relationships and execution priorities of tasks at each level. This step, through precise association binding and hierarchical expansion, enables task decomposition to form a logically coherent and clearly hierarchical task tree system, solving the problem of traditional task decomposition lacking hierarchical extension. This allows users to intuitively grasp the detailed structure of tasks, significantly improving the organization and ease of execution of task management.

[0044] Figure 3 A flowchart illustrating another intelligent memo task generation method provided in this application embodiment includes the following steps: Step 301: Obtain initial memo data in natural language form input by the user, perform semantic analysis on the initial memo data, and generate semantic analysis results.

[0045] Step 302: Based on the semantic analysis results, dynamically instantiate and generate structured prompt words from a predefined scene classification template library.

[0046] Step 303: Input the structured prompt words into the large language model and obtain the extended content output by the large language model, which includes multiple sub-tasks and supplementary information.

[0047] Step 304: Parse the extended content into a structured task object with parent-child relationship and priority attributes, and then display the structured task object.

[0048] For a description of steps 301-304, please refer to the description of steps 101-104 in the above embodiment. This embodiment will not repeat them in detail.

[0049] Step 305: Detect potential relationships between multiple task items in the user's memo, generate cross-task contextual suggestions based on the common sense reasoning ability of the large language model, and add the suggestions to the structured task object in the form of related tasks.

[0050] In some embodiments, multiple task items refer to the structured task objects generated in step 104 and other task data already existing in the user's memo. These task items may belong to different scenarios, but there is an inherent connection in life logic that is not explicitly presented. Potential associations are implicit connections between tasks based on common sense, execution processes, or resource sharing. For example, "going to the supermarket to shop" and "preparing a family dinner" have a food supply association, and "going to the bank to handle business" and "mobile phone plan expiring" have a time utilization association. When detecting potential associations, the system first summarizes all structured task objects and historical task data, extracts key features such as the core intent, entity information, and time / scenario attributes of each task, and constructs a task feature set. Then, relying on the common sense reasoning ability of the large language model, the system performs multi-dimensional association analysis on the tasks in the feature set. By matching the conventional execution logic, resource complementarity, and time connection possibility in life scenarios, the system identifies potential associations between tasks and excludes task combinations that have no practical significance. When generating cross-task scenario-based suggestions, the large language model, based on the identified association logic and the execution details of specific scenarios, outputs actionable guidance. Suggestions must clearly define the associated task, execution method, and associated value. For example, after detecting the association between "going to the supermarket" and "weekend barbecue," the system generates a suggestion to "purchase barbecue charcoal, disposable skewers, and seasonings simultaneously during the shopping trip." Similarly, after detecting the association between "bank services" and "package expiration," the system generates a suggestion to "renew the package through the app while waiting for the service." When adding associated tasks, the system encapsulates the suggestions as task items with association identifiers, marking their associated core tasks and priorities to ensure the priorities align with or adapt to scenario requirements. These are then integrated into the corresponding structured task objects, ensuring clear and traceable cross-scenario connections between tasks. This step breaks down the limitations of isolated tasks, making task planning more holistic through intelligent association and scenario-based suggestions, reducing repetitive operations in multitasking, and further enhancing the intelligence and practicality of the memo app.

[0051] Step 306: Establish a parent-child relationship between the recursively generated subtasks and the original task object, and update the hierarchical structure of the structured task object.

[0052] For a description of step 306, please refer to the description of step 105 in the above embodiment. This embodiment will not repeat the details further.

[0053] Figure 4 This is a schematic diagram of the structure of a memo task intelligent generation device provided in an embodiment of this application, as shown below. Figure 4As shown, it includes: a first generation module 401, a second generation module 402, an acquisition module 403, a parsing module 404, and a response module 405.

[0054] The first generation module 401 is configured to acquire initial memo data in natural language form input by the user, perform semantic analysis on the initial memo data, and generate semantic analysis results. The second generation module 402 is configured to dynamically instantiate and generate structured prompt words from a predefined scene classification template library based on semantic analysis results; The acquisition module 403 is configured to input structured prompt words into the large language model and acquire the extended content output by the large language model, which includes multiple subtasks and supplementary information. The parsing module 404 is configured to parse the extended content into a structured task object with parent-child relationship and priority attributes, and then display the structured task object. The response module 405 is configured to respond to a user's request for refinement of the target subtask, triggering a recursive decomposition of the target subtask and generating a subtask of the target subtask.

[0055] In some examples of this embodiment, the second generation module 402 is specifically configured to match the corresponding scene classification template from the scene classification template library according to the semantic intent of the initial memo data; input the entity information and context features in the initial memo data into the scene classification template to generate structured prompt words containing task type, priority requirements and output format.

[0056] In some examples of this embodiment, the acquisition module 403 is specifically configured to input structured prompts into the large language model and receive extended content output by the large language model engine, which includes multiple sub-tasks and supplementary information. Based on the task type and priority requirements indicated in the structured prompts, the large language model generates multiple sub-tasks and supplementary information related to the initial memo data based on common sense reasoning, and assigns priority attributes to each sub-task.

[0057] In some examples of this embodiment, the parsing module 404 is specifically configured to verify whether the format of the extended content conforms to the predefined structured output specification; map the subtasks and supplementary information in the verified extended content to task objects with priority fields respectively, and establish the association relationship between parent and child tasks.

[0058] In some examples of this embodiment, the response module 405 is specifically configured to perform semantic analysis on the refined request instruction and generate a lower-level semantic analysis result, and based on the lower-level semantic analysis result, dynamically instantiate and generate a lower-level structured prompt word from a predefined scene classification template library; input the lower-level structured prompt word into a large language model to obtain the lower-level extended content output by the large language model, which includes multiple sub-tasks and supplementary information; parse the lower-level extended content into a lower-level structured task object with parent-child relationship and priority attribute; establish a parent-child relationship between the recursively generated lower-level sub-tasks and the original task object, and update the hierarchical structure of the structured task object.

[0059] It should be noted that other corresponding descriptions of the functional units involved in the intelligent memo task generation device provided in this embodiment can be found in [reference]. Figure 1 , Figure 2 and Figure 3 The corresponding descriptions in [the document] will not be repeated here.

[0060] Based on the above, Figure 1 , Figure 2 and Figure 3 The present embodiment provides a method for intelligently generating memo tasks. Correspondingly, this embodiment also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method. Figure 1 , Figure 2 and Figure 3 This illustrates a method for intelligently generating memo tasks.

[0061] Based on the above, Figure 1 , Figure 2 and Figure 3 The present embodiment provides a method for intelligently generating memo tasks. Correspondingly, this embodiment also provides a computer program product on which a computer program is stored. When executed by a processor, this computer program implements the above-described method. Figure 1 , Figure 2 and Figure 3 This illustrates a method for intelligently generating memo tasks.

[0062] Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer device (such as personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of this application.

[0063] Based on the above, Figure 1 , Figure 2 and Figure 3 A method for intelligently generating memo tasks is shown, and Figure 4To achieve the above objectives, the present application also provides an electronic device, such as a personal computer or a server, in the illustrated virtual device embodiment. This device includes a storage medium and a processor; the storage medium stores a computer program; the processor executes the computer program to implement the above-described virtual device. Figure 1 , Figure 2 and Figure 3 This illustrates a method for intelligently generating memo tasks.

[0064] In some embodiments, the aforementioned physical device may further include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, an input unit such as a keyboard, etc., and optionally, a USB interface, a card reader interface, etc. In some embodiments, the network interface may include a standard wired interface, a wireless interface (such as a Wi-Fi interface), etc.

[0065] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical device, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0066] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0067] The above are merely specific embodiments of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to these embodiments, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for intelligent generation of memo tasks, characterized in that, The method comprises the following steps: acquiring initial memo data in the form of natural language input by a user, performing semantic analysis on the initial memo data and generating a semantic analysis result; based on the semantic analysis result, dynamically instantiating a structured prompt word from a pre-defined scene classification template library; inputting the structured prompt word into a large language model, and acquiring extended content output by the large language model, the extended content comprising a plurality of sub-tasks and supplementary information; parsing the extended content into a structured task object having a parent-child relationship and priority attribute, and displaying the structured task object; in response to a refinement request of a target sub-task by the user, triggering recursive decomposition of the target sub-task, and generating a lower-level sub-task of the target sub-task. 2.The method of claim 1, wherein, The step of dynamically instantiating a structured prompt word from a pre-defined scene classification template library based on the semantic analysis result comprises the following steps: matching a corresponding scene classification template from the scene classification template library according to the semantic intent of the initial memo data; inputting entity information and context features in the initial memo data into the scene classification template, and generating a structured prompt word comprising a task type, priority requirement and output format. 3.The method of claim 1, wherein, The step of inputting the structured prompt word into a large language model, and acquiring extended content output by the large language model, the extended content comprising a plurality of sub-tasks and supplementary information, comprises the following steps: inputting the structured prompt word into a large language model, and receiving extended content output by the large language model engine, the extended content comprising a plurality of sub-tasks and supplementary information, the large language model generating a plurality of sub-tasks and supplementary information related to the initial memo data based on common sense reasoning according to the task type and priority requirement indicated in the structured prompt word, and assigning a priority attribute to each sub-task. 4.The method of claim 1, wherein, The step of parsing the extended content into a structured task object having a parent-child relationship and priority attribute, and displaying the structured task object, comprises the following steps: verifying whether the format of the extended content conforms to a pre-defined structured output specification; mapping the sub-tasks and supplementary information in the verified extended content into task objects having a priority field respectively, and establishing an association relationship between parent and child tasks. 5.The method of claim 1, wherein, The step of, in response to a refinement request of a target sub-task by the user, triggering recursive decomposition of the target sub-task, and generating a lower-level sub-task of the target sub-task, comprises the following steps: performing semantic analysis on the refinement request instruction and generating a lower-level semantic analysis result, and based on the lower-level semantic analysis result, dynamically instantiating a lower-level structured prompt word from the pre-defined scene classification template library; inputting the lower-level structured prompt word into the large language model, and acquiring lower-level extended content output by the large language model, the lower-level extended content comprising a plurality of sub-tasks and supplementary information; parsing the lower-level extended content into a lower-level structured task object having a parent-child relationship and priority attribute; establishing a parent-child relationship between the recursively generated lower-level sub-task and the original task object, and updating the hierarchical structure of the structured task object. 6.The method of claim 1, wherein, The method further comprises the following steps: Detecting potential associations between multiple task items in a user's memo, generating cross-task scenario suggestions based on the common sense reasoning capabilities of the large language model, and adding the suggestions to the structured task object in the form of associated tasks.

7. A memo task intelligent generation apparatus characterized by comprising: The method comprises the following steps: A first generation module configured to obtain initial memo data in natural language input by a user, perform semantic analysis on the initial memo data, and generate a semantic analysis result; A second generation module configured to dynamically instantiate a structured prompt word from a pre-defined scene classification template library based on the semantic analysis result; An acquisition module configured to input the structured prompt word into a large language model and acquire expanded content containing multiple sub-tasks and supplementary information output by the large language model; An analysis module configured to parse the expanded content into a structured task object with parent-child relationships and priority attributes, and display the structured task object; A response module configured to respond to a user's request for refinement of a target sub-task, trigger recursive decomposition of the target sub-task, and generate lower-level sub-tasks of the target sub-task.

8. An electronic device, comprising: The method comprises the following steps: At least one processor; and a memory connected in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the memo task intelligent generation method of any one of claims 1-6.

9. A non-transitory computer-readable storage medium having stored thereon computer instructions, wherein, The computer instructions are used to enable the computer to perform the memo task intelligent generation method according to any one of claims 1-6.

10. A computer program product, characterised in that, The computer program, when executed by a processor, implements the memo task intelligent generation method according to any one of claims 1-6.

Images