Virtual post-driven multi-type memory scene capability timing evaluation and post-customized resume and interview content generation system and method

Abstract

The application discloses a virtual post-driven multi-type memory scene ability timing evaluation and post-customized resume and interview content generation system and method, and belongs to the technical field of artificial intelligence and human resource management. In view of the problems of static ability evaluation, disconnection between content generation and ability, lack of compliance control and closed-loop feedback in the prior art, the application unifies three types of memory scenes of learning, job hunting and interview through a memory platform module; a virtual post evaluation engine configures a virtual task sequence based on a target post to observe and dynamically update user ability; an ability time prediction module estimates the ability compliance time; a resume and interview generation module generates a post-customized resume and interview question and answer with a risk label based on memory scene matching; a text compliance review module ensures that the content expression is true and compliant; and a feedback optimization module reversely optimizes the system by using real job hunting results to form an evaluation and generation closed loop. The system can provide integrated intelligent job hunting support for users, which integrates virtual training, ability prediction, resume generation and interview preparation.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence, human resource management, and online vocational education, specifically to an intelligent ability assessment and content generation technology solution based on the integration of virtual job task flow and multi-type occupational memory scenarios, and in particular to a system and method for virtual job-driven multi-type memory scenario ability time-series assessment and job-customized resume and interview content generation. Background Technology

[0002] With the development of technologies such as large-scale pre-trained language models, knowledge graphs, and recommendation systems, a large number of intelligent products have emerged focusing on job recruitment, career guidance, and training. Among the publicly available technical solutions, they can be mainly categorized as follows:

[0003] 1. A knowledge graph or rule-based system for matching people to jobs and providing career guidance.

[0004] These systems typically construct occupational knowledge graphs, using keywords and attribute nodes from candidate resumes to match them with job requirement vectors or job knowledge graph nodes, outputting job recommendations or job matching scores. Their technical focus lies in graph structure design and matching strategy optimization, but they suffer from the following limitations: the input side primarily relies on existing candidate resumes or questionnaire results, treating user capabilities as static features and lacking dynamic modeling of subsequent learning behaviors, project practice, and interview processes; the system often uses "current match" as its evaluation objective, making it difficult to answer the question of "under what training path, when can the user reach the target job requirements"—the timeframe for capability development; and the resume content itself, as static text, is only used for matching evaluation rather than systematic generation.

[0005] 2. Deep Learning-Based Resume Structuring and Information Extraction System

[0006] Such systems typically use pre-trained language models, sequence labeling models, or machine reading comprehension models to identify event types, extract time information, and structure elements in natural language resumes, thereby generating structured resumes in a uniform format. However, these technologies generally assume that "the resume text already exists" and do not address "how to automatically construct resume fragments adapted to specific positions from users' long-term learning and practical behaviors," nor do they incorporate virtual job practice data into the resume generation and optimization process.

[0007] 3. Resume matching and non-standard resume processing methods based on artificial intelligence and knowledge graphs

[0008] This approach utilizes text cleaning, keyword extraction, vector representation, and knowledge graphs to standardize and extract deep features from candidate resumes, then performs multi-dimensional matching with job descriptions for candidate-job matching recommendations or screening. Its key characteristics are an emphasis on learning representations from unstructured text and optimizing matching scores and screening effectiveness. However, at the system level, this technology still treats candidate resumes as externally provided input, without involving virtual job task flows, tracking ability evolution, or creating a closed loop between matching results and subsequent candidate training or resume content generation.

[0009] 4. Interview simulation and scoring system based on large models or knowledge graphs

[0010] Related publicly available technologies include methods for generating simulated interview scenarios based on large models, online interview scoring systems based on video / voice / text, and methods for generating interview questions and evaluating answers based on knowledge graphs. These technologies typically achieve the following functions: generating interview questions based on job descriptions and candidate resumes; scoring candidates based on the semantic content, emotional state, and body language of their audio / video streams and text responses; and outputting a comprehensive evaluation index for a single interview.

[0011] Such solutions generally focus on the "current interview stage," with evaluation results primarily serving as a basis for employers' hiring decisions rather than being incorporated into the candidate's long-term professional memory structure. Furthermore, they rarely establish clear algorithmic links between interview performance and resume content generation, or prediction of skill development timelines.

[0012] 5. Process Compliance and General Text Compliance Review System

[0013] Some publicly available technologies propose methods for process compliance review or text content security review based on rule bases and pre-trained models. These methods are applicable to fields such as business process compliance, advertising review, and content security control. The basic logic is as follows: semantic parsing of input process nodes or text content; retrieval of compliance benchmarks in rule databases or vector libraries; identification of potential violations or high-risk nodes; and provision of compliance conclusions.

[0014] However, these technologies are generally geared towards business processes, policy provisions, or general content security scenarios, lacking targeted modeling for the specific issue of "the authenticity and compliance of virtual job descriptions / learning projects in job resumes and interview answers." For example, existing solutions often lack fine-grained structural and semantic constraint mechanisms to identify when "virtual internships," "classroom projects," or "simulated tasks" are mistakenly written as "real employment experience" or "real job responsibilities."

[0015] 6. Virtual Practice and Job Simulation Systems

[0016] Some systems offer exercises or courses similar to real-world jobs through virtual scenarios, simulations, or online projects, recording learning processes and achievements. However, these solutions are mostly used for education or training itself, failing to integrate virtual practice data with job resume generation, interview content generation, and ability time prediction at the system level. Furthermore, virtual task performance is rarely abstracted into a unified "career memory structure," making it impossible to utilize in a closed-loop manner within job-seeking scenarios.

[0017] In summary, existing technologies reveal several key issues: At the data level, current systems generally lack a unified abstraction of "multi-type professional memory scenarios." Learning behaviors, virtual task performance, job-seeking behaviors, and interview records are fragmented, hindering cross-scenario integration. At the competency assessment level, evaluations are primarily based on current competency matching and current interview scores, lacking continuous observation from the perspective of virtual job task flows. Furthermore, it's difficult to provide interpretable time-series predictions based on individual behavioral history regarding the time required to reach a certain competency level. At the content generation level, while many technical solutions can optimize resumes or generate interview questions based on templates, they remain largely at the "static text processing" stage. They fail to strongly couple project-level experiences with multi-type memory scenarios and competency assessment results, and they don't simultaneously achieve the linkage between "competency assessment—time prediction—resume content—interview questions and answers" within a single system. At the compliance control level, dedicated risk control mechanisms for job application texts driven by virtual positions / internships are still immature, making it difficult to systematically avoid risks such as "learning projects being disguised as real work experience" and "virtual responsibilities being exaggerated into real management responsibilities." At the closed-loop feedback level, most existing systems are one-way or partial closed loops, such as "resume matching → interview scoring". Few systems write back the real job application results (resume passing, interview passing, and hiring status) to the virtual job evaluation system, thereby dynamically adjusting the virtual task arrangement and scenario weights.

[0018] In view of this, there is an urgent need for a new technical solution that takes virtual positions as the driving force, integrates multiple types of professional memory scenarios into a memory platform, performs dynamic modeling of abilities through virtual task performance, and achieves a close integration of ability time prediction, generation of customized resumes and interview content, text compliance review and feedback optimization, forming a system-level closed loop across learning, virtual practice and real job seeking, thereby effectively improving the quality of assessment and generation and building a strong technical barrier. Summary of the Invention

[0019] I. Purpose of the invention.

[0020] The main objective of this invention is:

[0021] 1. Propose a multi-type memory scenario modeling method that can uniformly represent behavioral data from multiple sources such as learning, job seeking, and interviews, and build a memory platform that serves ability assessment and content generation.

[0022] 2. Through the virtual job evaluation engine, the target job is decomposed into multi-level virtual business scenarios and virtual tasks. Based on the performance of virtual tasks, the user's ability is dynamically updated in a fine-grained manner, and on this basis, the ability time sequence is predicted, giving the time range for reaching different ability levels.

[0023] 3. Design a method for generating customized resumes and interview content based on multiple memory scenarios, so that the generated project-level resume fragments, customized resumes, and interview questions and answers are consistent with the user's real learning trajectory, virtual task performance, and interview history.

[0024] 4. In response to the context of virtual positions and learning projects, a dedicated text compliance review mechanism is proposed to prevent risks such as learning projects being mistakenly written as real employment experiences and virtual position responsibilities being exaggerated, thereby ensuring the compliance of generated content in terms of professional ethics and information authenticity.

[0025] 5. Introduce a feedback optimization module to write real resume submissions and interview results back to the memory platform and virtual job evaluation engine. This drives the dynamic adjustment of the importance weight of memory scenarios and task scheduling strategies, thereby building a two-way closed loop between virtual evaluation and real job application results, and continuously improving the overall performance of the system.

[0026] II. Technical Solution

[0027] To achieve the above objectives, the present invention proposes a virtual job-driven multi-type memory scenario ability time-series assessment and job-customized resume and interview content generation system and method, which includes the following key technical components: a memory platform module, a virtual job assessment engine, an ability time prediction module, a resume and interview generation module, a text compliance review module, a feedback optimization module, and an optional industry plug-in module, with each module forming an organic synergy.

[0028] (a) Memory Platform Module

[0029] The memory platform module is the data semantic foundation of this invention. It is used to perform unified structuring and contextual processing on raw interactive data from multiple business domains such as learning, job seeking, and interviewing, forming multiple types of memory scenarios that can be reused by multiple modules.

[0030] 1. Construction of learning and memory units

[0031] The memory platform module structures and stores key user behaviors and events within the learning platform, online courses, virtual projects, and virtual tasks—such as highlighted follow-up questions, code experiments, knowledge chain organization, Feynman paraphrasing, experimental verification, error correction, and bug fix suggestions—into learning memory units. Each learning memory unit includes at least: a unique identifier and a user identifier; a timestamp and event summary; a structured fact array to record the core knowledge points and conclusions generated by this behavior; a prospective information array to record subsequent learning objectives or hypotheses; metadata objects, including project identifiers, concept tag arrays, difficulty coefficients, and session identifiers; and extended fields, including virtual task identifiers, cognitive load scores, and behavioral feature counts.

[0032] 2. Construction of Job Search Memory Units

[0033] The system structures key information from users' job browsing, job collection, filter setting, customized resume generation, resume export and submission, and feedback reception into job search memory units, including: user identifier, company identifier, job identifier; submission channel, submission time, submission status; feedback type (such as "advanced to interview" or "failed to pass the screening") and feedback text summary.

[0034] 3. Construction of Interview Memory Units

[0035] The system stores the question-and-answer process from real and mock interviews as interview memory units, which include: user identifier, company identifier, job identifier, conversation identifier; question text, answer text; scoring results (which may include multi-dimensional scores) and a set of weakness tags.

[0036] 4. Memory Scene Aggregation and Attribute Calculation

[0037] The memory platform module aggregates the above memory units based on project identifiers, company identifiers, and session identifiers to form: learning scenarios (project-dimensional aggregation), reflecting the user's continuous learning and practice records around a certain project or set of virtual tasks; job search scenarios (company-dimensional aggregation), reflecting the user's job search behavior and phased results for a certain company; and interview scenarios (session-dimensional aggregation), reflecting the complete process and results of a single interview.

[0038] For each scenario, the system calculates indicators such as basic importance, average difficulty, time span, and relevance to typical job groups, and uses basic importance as the initial value for subsequent scenario weight updates.

[0039] (ii) Virtual Job Evaluation Engine

[0040] The virtual job evaluation engine constructs virtual job objects around the target job and uses virtual task flows to make repeatable and quantifiable observations of user capabilities.

[0041] 1. Virtual Job Object Modeling

[0042] For each target position, the system parses the job description text, extracts typical business processes and key responsibilities, and constructs a virtual job object, which includes at least: job identifier and job description text; a capability requirement vector, used to represent the target values ​​of the position in several capability dimensions; a set of virtual business scenarios at levels L1–L3: a multi-level abstraction from the overall business flow and sub-processes to specific operation steps; and a set of virtual tasks obtained by decomposing the L3 level scenarios.

[0043] 2. Virtual Task Modeling

[0044] Each virtual task is defined as the smallest work unit that can be executed online, and has the following attributes: task identifier and the identifier of the virtual business scenario to which it belongs; the set of concept tags or capability tags involved; the basic complexity of the task and the estimated time consumption; task type tags (such as analysis, implementation, communication, review, etc.); and a structured description template of the task result, which is convenient for subsequent writing back to the learning and memory unit.

[0045] 3. Task orchestration based on capability gaps and scenario weights

[0046] The system calculates the capability gap vector D based on the current user capability vector C_u and the job requirement vector C_p; then, it combines this with the final importance I_final of the memory scenario to weight and sort the virtual tasks. After the user provides a time budget, the system selects several tasks from the high-priority tasks to generate a daily or weekly task sequence for the virtual job.

[0047] 4. Writing back task execution results and updating capabilities

[0048] After a user completes a virtual task, the system records the task performance score, cognitive load score, actual time spent, and behavioral characteristics. These records, along with the task identifier, virtual business scenario identifier, project identifier, and concept tag set, are written into an extended learning memory unit. The system calculates the increment of concept mastery based on task performance and updates the ability dimension score through a concept-to-ability dimension mapping matrix. The cognitive load score serves as a learning efficiency adjustment factor in the ability update process, thereby achieving dynamic modeling of individual abilities.

[0049] (III) Capacity Time Prediction Module

[0050] Based on the aforementioned capability updates, the capability time prediction module estimates the time it will take for users to reach the required job capabilities in the future.

[0051] 1. Competency Gap and Growth Rate: The module calculates the competency gap vector D based on the latest competency dimension scores and job requirement vectors, and estimates the competency growth rate v_i from the historical sequence of competency scores within a preset time window, which can be achieved using a weighted moving average or regression model.

[0052] 2. Target Completion Time and Time Interval: Under the premise that D_i > 0 and v_i > 0, the module calculates the target completion time for each dimension, t_i = D_i / max(v_i, ε). If v_i ≤ 0, it is set to a default value greater than the upper limit T_max. Then, based on the maximum value, median, and other indicators of t_i in the core competency dimension set, the overall expected target completion time T for the position is obtained by weighting, and the upper and lower fluctuation intervals [T_low, T_high] are obtained through proportional adjustment, corresponding to the interview level and the recruitment level, respectively.

[0053] This module allows the system to not only explain "how big the gap is between the current situation and the job requirements", but also "when the job requirements can be roughly met under the current learning and virtual task strategy".

[0054] (iv) Resume and Interview Generation Module

[0055] The resume and interview generation module uses multiple memory scenarios as its content foundation, combining ability assessment results with job characteristics to generate customized resumes and various types of interview content for users.

[0056] 1. Project-level resume fragment generation

[0057] The module performs feature analysis on the target job description, obtaining skill and business characteristics. It then conducts a mixed search within learning and job-seeking scenarios, calculating a matching score based on semantic similarity, keyword overlap, and scenario importance weighting to select several highly matched scenarios. For each scenario, a suitable template is selected to generate project-level resume fragments, clearly defining: project title, time frame, main tasks and results; the technology stack, business scenarios, and key metrics involved in the project; in-depth tags related to the project (representing technologies or business points that can be further explored); the level of expression intensity (controlling whether the wording is conservative or relatively prominent); and risk tags (such as learning projects, virtual projects, simulated environments, etc.).

[0058] 2. Customized resume sets for specific job positions

[0059] Based on the comprehensive matching score and expression intensity level of project-level resume fragments, the system selects several fragments to combine into a project experience section, and combines it with modules such as user education experience and skills list to generate a job-customized resume structure, ensuring that the content is highly relevant to the target position and that the source can be traced back to the corresponding memory scenario.

[0060] 3. Interview content generation

[0061] The module further constructs a set of assessment dimensions based on job-specific resumes and related memory scenarios, covering aspects such as project background, technical details, business logic, collaboration and communication, and risk control. It generates multiple types of interview questions for each dimension and generates standard, industry-enhanced, and stress-coping answers based on the facts of the memory scenario, industry knowledge, and expression strategies. These answers are packaged into interview memory cards to support subsequent mock interviews and intensive training.

[0062] (v) Text Compliance Review Module

[0063] The text compliance review module performs specific compliance controls on the project-level resume fragments, job-specific resumes, and interview answer texts generated in this invention. The module employs a combination of rule detection and semantic analysis.

[0064] 1. By using the sensitive word list of employment relationship and the word list of leadership responsibility, identify whether there are expressions such as "real employment relationship" and "senior responsibility" in the text that do not match the source of the learning project or virtual task;

[0065] 2. Check the description content for missing terms such as "learning," "course project," "virtual position," or "simulation task" by using the learning / virtual project declaration template;

[0066] 3. Use semantic analysis models to identify whether there is a risk of exaggerated, fictitious or misleading statements in the overall text.

[0067] The module quantifies rule risk, template risk, and semantic risk into numerical scores, and obtains the total risk score by weighted summation. Based on a set threshold, the text is divided into three levels: low risk, medium risk, and high risk. Low-risk text is allowed to be output directly; medium-risk text requires the user to explicitly confirm that it is learning or virtual content and records the confirmation log; high-risk text is blocked from output and a prompt for modification or weakening of the expression is returned.

[0068] This module effectively constrains the expression boundaries of virtual job positions and learning programs in job-seeking scenarios, reducing the integrity and compliance risks faced by candidates and the platform at the source.

[0069] (vi) Feedback Optimization Module

[0070] The feedback optimization module is responsible for writing the actual job application results back into the system, realizing a closed-loop correction and continuous optimization between virtual evaluation and actual performance.

[0071] The module statistically analyzes metrics such as the number of times a memory scenario is used, the number of resumes passed, the number of interviews passed, and the number of hiring events. It combines these metrics with scenario type and time factors to calculate the scenario success rate and feedback increment, and updates the scenario's final importance I_final based on this. Subsequently, the virtual job evaluation engine and the resume and interview generation module introduce the latest scenario weights into task orchestration and retrieval ranking, gradually tilting the system towards scenarios proven effective in real job applications, thus forming a dynamic closed loop of "virtual training—content generation—real job application—effect feedback—strategy update".

[0072] (vii) Industry Plug-in Module

[0073] To adapt to the business logic and expression methods of different industries, this invention introduces optional industry plug-in modules, which provide the virtual job evaluation engine and resume and interview generation module with the following through a unified interface: industry-specific virtual business scenarios and task decomposition rules; industry-specific task scoring and indicator interpretation logic; and industry-specific project descriptions and project story templates.

[0074] Under this mechanism, the system maintains consistency in core algorithms and data structures, while quickly covering multiple industry scenarios, enhancing the versatility and scalability of the overall solution.

[0075] III. Beneficial Effects

[0076] Compared with existing technologies, this invention achieves substantial progress in the following aspects through its overall technical path of "multi-type memory scenario modeling—virtual job-driven competency assessment—competency time-series prediction—job-customized resume and interview content generation—text compliance review—real job application feedback closed loop":

[0077] 1. Through the memory platform module, learning, job seeking, and interview data are uniformly abstracted into multiple types of memory scenarios, providing a traceable and combinable knowledge and experience foundation for ability assessment and content generation.

[0078] 2. Through the virtual job assessment engine and the competency time prediction module, dynamic competency modeling and competency time prediction based on virtual job task flow are realized, breaking through the limitation of static assessment in existing technologies.

[0079] 3. By generating project-level resume fragments, job-specific resumes, and various types of interview memory cards, professional memory is closely linked to the generated content, making the resume and interview content highly personalized and with clear data sources.

[0080] 4. Through the text compliance review module, we can address compliance risks associated with virtual job postings and learning programs in job application documents, ensuring the authenticity of the generated content and compliance with professional ethics.

[0081] 5. By introducing a feedback optimization module, a reverse adjustment mechanism is introduced to the scene weight and task arrangement based on real job search results, so as to realize closed-loop correction and continuous optimization between virtual and real, and gradually improve system performance over time.

[0082] This invention possesses clear and independent technical features and high comprehensive technical barriers in terms of system architecture design, data modeling methods, ability assessment and time prediction methods, job customization content generation mechanisms, and text compliance control for virtual job contexts. It can provide users with an integrated virtual job training and intelligent job search support solution without infringing on existing similar patents. Attached Figure Description

[0083] Figure 1 This is the overall system flowchart of the invention;

[0084] Figure 2 This is a flowchart of the memory platform module in the invention.

[0085] Figure 3 A flowchart for modeling virtual job positions and orchestrating virtual tasks in the invention;

[0086] Figure 4 A flowchart for capability increment updates and time prediction in the invention process;

[0087] Figure 5 Generate flowcharts for project-level resume fragments in the invention;

[0088] Figure 6 Generate flowcharts for customized resumes and interview content for positions in the invention process;

[0089] Figure 7 A flowchart for text compliance review in an invention;

[0090] Figure 8 Flowchart of the feedback optimization module in the invention;

[0091] Figure 9 Flowchart for integrating industry plug-in modules in the invention;

[0092] Figure 10 This is a flowchart of the multi-module collaborative closed-loop process in the invention. Detailed Implementation

[0093] The following embodiments are provided to describe in detail the implementation process and module collaboration relationships of the various functional modules of the present invention without limiting the scope of protection of the present invention. The scope of protection of the present invention is defined by the claims.

[0094] Example 1: Overall System Implementation

[0095] In this embodiment, the virtual job-driven multi-type memory scenario ability time-series assessment and job-customized resume and interview content generation system of the present invention adopts a modular layered architecture, including: a memory platform module; a virtual job assessment engine; an ability time prediction module; a resume and interview generation module; a text compliance review module; a feedback optimization module; and an industry plug-in module.

[0096] The system comprises several modules: a memory platform module responsible for the unified, structured, and contextualized storage of multi-source professional behavior data, serving as the "foundation of professional memory" for the entire system; a virtual job evaluation engine centered on virtual job objects and virtual task sequences to achieve dynamic observation of abilities; an ability time prediction module based on individual time-series ability trajectories to estimate the time required to meet standards; a resume and interview generation module transforming various memory scenarios into structured project-level resume fragments, job-customized resumes, and interview memory cards; a text compliance review module implementing dedicated risk control for job application texts in virtual job and learning project scenarios; a feedback optimization module writing real job application results back to the memory platform and virtual job evaluation engine to build a closed loop combining virtual and real data; and an industry plugin module injecting business knowledge and templates from different industries via interfaces.

[0097] correspond Figure 1 The overall system workflow includes the following main steps:

[0098] 1. System initialization: Load user account information, existing career goal configurations, risk preference parameters, and time budget settings; initialize the user context object in memory.

[0099] 2. The memory platform module receives incremental data streams from the learning platform, job search platform, and interview system. It performs cleaning, field standardization, and outlier filtering on the raw logs from different sources to obtain event records in a unified format.

[0100] 3. The memory platform module maps event records into learning memory units, job search memory units, and interview memory units, and aggregates them according to project identifiers, company identifiers, and conversation identifiers to generate three types of memory scene objects: learning scene, job search scene, and interview scene. At the same time, it calculates the basic importance of the scene.

[0101] 4. After the user selects the target position, the virtual position evaluation engine parses the position description text and constructs a virtual position object, including multi-level virtual business scenarios and virtual task sets.

[0102] 5. The virtual job assessment engine estimates the user's current ability vector based on historical task performance and learning records in the memory platform, compares it with the ability requirement vector of the target job to obtain the ability gap vector, and dynamically generates daily / weekly task sequences for virtual jobs by combining the final importance of the memory scenario and the user's time budget.

[0103] 6. The system pushes virtual tasks to the user terminal, collects the user's performance score, cognitive load score, actual time spent and behavioral characteristics during the task execution process, and writes this data into the extended learning and memory unit.

[0104] 7. The virtual job assessment engine updates the concept mastery and ability dimension scores based on the extended learning and memory unit, forming a new ability vector to provide input for the ability time prediction module.

[0105] 8. The capability time prediction module combines the latest capability vectors and historical capability time series data, uses time series modeling methods to estimate the growth rate and the time to reach the target for each capability dimension, and aggregates them on the core capability dimensions to obtain the overall expected target time and time range for the position.

[0106] 9. When a user requests to generate a customized resume and interview content for a job, the resume and interview generation module performs vector retrieval and keyword retrieval in multiple memory scenarios based on the characteristics of the target job. It selects high-matching scenarios to generate project-level resume fragments with expression intensity levels and risk tags, and combines project experience sections accordingly to generate a customized resume for the job. At the same time, it constructs interview assessment dimensions, automatically generates multiple types of interview questions and multiple versions of answers, and encapsulates them into a set of interview memory cards.

[0107] 10. The text compliance review module performs rule detection and semantic risk analysis on project-level resume fragments, job-specific resumes, and interview answer texts, outputting low-risk, medium-risk, or high-risk conclusions and providing decisions: pass directly, require user confirmation, or block output and provide modification suggestions.

[0108] 11. After a user exports a customized resume for a job that has passed the review and completes the actual application, the system listens for events such as changes in application status, interview arrangements, and hiring results. The feedback optimization module writes these results back to the job application memory unit and the interview memory unit.

[0109] 12. The feedback optimization module counts the number of times each memory scenario is used and the number of successful events. It calculates the scenario feedback increment by combining the time decay factor, updates the final importance of the scenario, and provides the latest weight value to the virtual job evaluation engine and the resume and interview generation module, forming a closed loop of continuous iterative optimization.

[0110] Example 2: Detailed Implementation of the Memory Platform Module

[0111] In this embodiment, the memory platform module performs unified modeling on the raw data from the three business domains of learning, job seeking, and interviewing, forming structured memory units and multiple types of memory scenarios, and maintains capability-related statistical information at the scenario level.

[0112] 1. Multi-source data acquisition and standardization

[0113] The memory platform module receives data streams from various business systems through predefined interfaces. These data streams include: learning data such as course start and end events, chapter completion, questions, experiment runs, virtual task submissions, and incorrect question records; job search data such as job browsing, job favorites, job filtering criteria, resume version selection, resume export, application operations, and application feedback; and interview data such as question texts, answer texts, interview scores, and interview results from mock and real interviews.

[0114] The module maps fields from different systems to a unified set of fields through a data source identifier and field mapping table, completes or deletes missing fields, and filters obviously abnormal records to ensure the consistency and stability of subsequent modeling.

[0115] 2. Memory Unit Construction

[0116] For learning data, the module constructs learning memory units. Each learning memory unit includes at least: a unique identifier, a user identifier, a timestamp, an event summary, a structured fact array, a prospective information array, and a metadata object containing an item identifier, a concept tag array, a difficulty coefficient, and a session identifier. When the learning behavior occurs during the execution of a virtual task, the extended fields also include a virtual task identifier, a cognitive load score, and a behavioral feature count.

[0117] For job search data, the module constructs job search memory units, including: unique identifier, user identifier, company identifier, job identifier, application channel, resume version identifier, application time, application status, and feedback information summary.

[0118] For interview data, the module constructs interview memory units, including: unique identifier, user identifier, company identifier, job identifier, conversation identifier, question text, answer text, rating results, and a set of weakness tags.

[0119] 3. Memory Scene Aggregation

[0120] The modules aggregate scenarios according to the following rules:

[0121] Using project identifiers as a dimension, learning memory units and their extended learning memory units related to the same project are aggregated into learning scene objects, and the coverage of concept tags, average difficulty, time span, and number of virtual task executions within the scene are statistically analyzed.

[0122] Using company identification as the dimension, job search memory units under the same company and interview memory units related to the company are aggregated into job search scenario objects, and the number of resumes submitted, the number of interviews entered, and the hiring results are counted.

[0123] Using session identifiers as a dimension, interview memory units under the same session are aggregated into interview scenario objects, and the number of questions, overall score, and concentrated distribution of weaknesses in a single interview are statistically analyzed.

[0124] 4. Calculation of Scene Attributes and Basic Importance

[0125] For each memory scenario, the module calculates a basic importance value based on the average difficulty of the memory units within the scenario, the time span, the diversity of concept tags, and the semantic relevance to typical job groups, and records it in the scenario object for use by the subsequent feedback optimization module and the virtual job evaluation engine.

[0126] Example 3: Detailed Implementation of the Virtual Job Evaluation Engine

[0127] In this embodiment, the virtual job evaluation engine constructs virtual job objects around the target job and performs fine-grained observation and updates of user capabilities through virtual task sequences.

[0128] 1. Virtual Job Object Modeling

[0129] The engine parses the target job description text, extracts core business processes and capability requirements, and constructs virtual job objects, which include at least: job identifier, job description text; capability requirement vector, describing the target level of several capability dimensions; a set of L1–L3 level virtual business scenarios, corresponding to the overall business process, business sub-processes and specific operation scenarios respectively; and a set of virtual tasks, derived from the task decomposition of L3 level virtual business scenarios.

[0130] 2. Virtual Task Modeling

[0131] Each virtual task is defined as the smallest independently executable task unit, with the following attributes: task identifier, virtual business scenario identifier, set of related concept tags or capability tags, basic task complexity, estimated time consumption, task type tag (such as analysis type, implementation type, communication type, etc.), and a structured description template of the task result.

[0132] The engine maintains dynamic complexity factors for some tasks based on historical task performance, which are used to adapt to the user's current ability level in subsequent task orchestration.

[0133] 3. Task scheduling and push notification

[0134] The virtual job assessment engine retrieves the user's historical learning and virtual task execution records from the memory platform, calculates the current ability vector, and compares it with the job ability requirement vector to obtain an ability gap vector. The engine combines the final importance of the remembered scenario and the task complexity factor to calculate a comprehensive priority for each candidate virtual task. Under the user's given daily or weekly time budget constraints, the engine selects several high-priority tasks to form a daily or weekly task sequence for the virtual job and pushes the task sequence to the user's terminal for execution.

[0135] 4. Task execution data write-back and capability incremental update

[0136] After a user completes a virtual task, the system records the task performance score, cognitive load score, actual time spent, and behavioral characteristics, forming a task execution result data structure. The engine calculates the concept mastery increment based on the task execution result and updates the mastery of the corresponding concept label. Then, it updates the ability dimension score using the concept-to-ability dimension mapping relationship. The cognitive load score serves as a learning efficiency adjustment factor, forming a cognitive load-sensitive ability increment model.

[0137] Example 4: Detailed Implementation of the Capacity Time Prediction Module

[0138] In this embodiment, the capability time prediction module estimates the time required for a user to meet the job capability requirements based on the historical changes in capability dimension scores, and outputs a time interval.

[0139] The module extracts a sequence of capability vectors within a preset time window from the memory platform, combines the latest capability vectors with the job capability requirement vectors to calculate a capability gap vector, estimates the capability growth rate of each capability dimension, and then obtains the time to reach the target for each dimension based on the ratio of the gap value to the growth rate. For the core capability dimension set, the module provides an estimated time range for overall job capability achievement by weighting and combining the estimated achievement times of multiple dimensions and calculating confidence intervals.

[0140] Example 5: Detailed Implementation of Project-Level Resume Fragments and Job-Specific Resume Generation

[0141] In this embodiment, the resume and interview generation module uses multiple types of memory scenarios as sources to automatically generate project-level resume fragments and job-customized resumes.

[0142] The module performs feature analysis on the target job description, extracting sets of hard skill keywords, business keywords, and behavioral ability keywords. It then performs vector and keyword searches in both learning and job-seeking scenarios, weighting the importance of the scenario, the coverage of conceptual tags, and the relevance of abilities to select candidate scenarios. For each highly matched scenario, the module uses a preset template to generate project-level resume fragments. These fragments include a fragment identifier, scenario identifier, project title, time range, array of key project points, array of in-depth tags, expression intensity level, array of risk tags, and industry tags. Subsequently, the module selects several project-level resume fragments based on the comprehensive matching score and expression intensity level to form a project experience section, which is then merged with the education experience and skills list to create a customized resume for the job.

[0143] Example 6: Detailed Implementation of Interview Content Generation and Memory Card Construction

[0144] Based on selected project-level resume segments and their associated learning, job-seeking, and interview scenarios from the customized resumes, the resume and interview generation module extracts assessment dimensions such as project background, key technical points, business logic, collaboration and communication, and risk control. It automatically generates basic, in-depth business, in-depth technical, and stress-based interview questions, and generates standard, industry-enhanced, and stress-response versions of the answers respectively, constructing interview memory card objects. Each interview memory card contains at least the question text, multiple versions of the answer text, a keyword list, structured answer prompts, difficulty level, and learning status fields, stored in a memory platform to support subsequent practice and mock interviews.

[0145] Example 7: Detailed Implementation of the Text Compliance Review Module

[0146] The text compliance review module performs rule-based detection and semantic analysis on project-level resume fragments, job-specific resumes, and interview answer texts to identify risks such as learning projects being mistakenly listed as real work experience or virtual tasks being exaggerated into actual responsibilities. Based on a sensitive word list regarding employment relationships and learning / virtual project declaration templates, combined with the risk probabilities output by the semantic analysis model, the module classifies the text into low-risk, medium-risk, or high-risk levels and returns corresponding decisions and modification suggestions.

[0147] Example 8: Detailed Implementation of the Feedback Optimization Module and Industry Plugin Module

[0148] The feedback optimization module tracks resume submissions, interview pass rates, and hiring results, and calculates the success rate and feedback increment of the remembered scenarios based on the time decay principle, updating the final importance of the scenarios. The industry plugin module provides virtual scenario configurations, task scoring rules, and project description templates for different industries through a unified interface, expanding the system to multiple industries and job groups without changing the core architecture.

Claims

1. A virtual job-driven system for multi-type memory scenario ability temporal assessment and customized resume and interview content generation, characterized in that, include: The memory platform module is used to collect and store user interaction data during the learning, job search and interview process, generate learning memory units, job search memory units and interview memory units, and aggregate the memory units based on project identifiers, company identifiers and session identifiers to construct three types of memory scenario objects: learning scenario, job search scenario and interview scenario. The virtual job assessment engine is used to construct virtual job objects for target jobs. The virtual job objects include a multi-level set of virtual business scenarios and a set of virtual tasks obtained by decomposing the multi-level virtual business scenarios. Based on the ability gap vector between the user's current ability vector and the ability requirement vector of the target job, the importance weight of each memory scenario, and the user's set time budget, a sequence of virtual job tasks is generated. The virtual tasks are pushed to the user for execution at a daily or weekly pace. The engine records the user's decision results, task performance scores, cognitive load scores, actual time consumption, and behavioral characteristics in each virtual task. The recorded information is written into an extended learning memory unit, and the user's ability dimension score is updated based on the mapping relationship between concept mastery and ability dimension. The capability time prediction module is used to estimate the expected time for each core capability dimension to meet the target job requirements after obtaining the updated capability dimension scores, based on the capability gap vector and historical capability change data within a preset time window, and aggregate them to obtain the overall expected time to meet the job requirements and the time interval. The resume and interview generation module is used to perform mixed retrieval and sorting based on the learning scenario, job search scenario and interview scenario after receiving information about the target company and target position. It selects the memory scenario with a high degree of matching with the target position, generates project-level resume fragments with expression intensity level and risk label, generates a job-customized resume according to a predetermined combination rule, and automatically generates multiple types of interview questions and corresponding answers based on the job-customized resume and the memory scenario it references. The questions and answers are encapsulated into interview memory card objects. The text compliance review module is used to perform rule detection based on the employment relationship sensitive word list and learning or virtual project declaration template, as well as risk assessment based on semantic analysis model, on the project-level resume fragments, job-customized resumes and interview answer texts before they are displayed, exported or used for mock interviews. It outputs the final risk level as low risk, medium risk or high risk, and returns the review result to the resume and interview generation module, indicating whether the text passes, prompts for confirmation or blocks the output. The feedback optimization module is used to write the export behavior of customized resumes, resume submission status, interview scores and hiring results into the job search memory unit and the interview memory unit, update the corresponding job search scenario and interview scenario statistics, and dynamically adjust the importance weight of each memory scenario in combination with the virtual task performance results output by the virtual job evaluation engine, so that the virtual job evaluation and the real job search results form a closed loop optimization.

2. The system according to claim 1, characterized in that, The memory platform module is specifically used for: The user's highlighted questions, code experiments, knowledge chain organization, Feynman paraphrasing, experimental verification, and suggestions for filling gaps during the learning process are structured and generated into learning memory units. Each learning memory unit includes at least a unique identifier, user identifier, timestamp, event summary, structured fact array, prospective information array, and metadata object containing item identifier, concept tag array, difficulty coefficient, and session identifier. Virtual task identifier, cognitive load score, and behavioral feature fields are added to the extended learning memory unit. The job screening process, customized resume versions, resume export records, channel submission information, and submission feedback results are structured and processed to generate job search memory units. Each job search memory unit includes at least a unique identifier, user identifier, company identifier, job identifier, submission channel, submission time, submission status, feedback type, and feedback text summary. The question text, answer text, scoring results, and weakness tag set from real and simulated interviews are structured and processed to generate interview memory units. Each interview memory unit includes at least a unique identifier, user identifier, company identifier, job identifier, conversation identifier, question text, answer text, scoring results, and weakness tag set. Using project identifier as the aggregation dimension, learning memory units under the same project identifier and their corresponding extended learning memory units are aggregated into learning scenario objects. Using company identifier as the aggregation dimension, job search memory units under the same company identifier and interview memory units related to that company are aggregated into job search scenario objects. Using conversation identifier as the aggregation dimension, interview memory units under the same conversation identifier are aggregated into interview scenario objects.

3. The system according to claim 1, characterized in that, The virtual job evaluation engine includes: The virtual job modeling submodule is used to parse the job description text and competency configuration for each target job and construct a virtual job object. The virtual job object includes at least a job identifier, a job competency vector, a multi-level virtual business scenario set, and a set of virtual tasks obtained from the multi-level virtual business scenarios by task splitting rules. The virtual task modeling submodule is used to configure the virtual business scenario identifier, the set of concept tags or capability tags involved, the basic complexity of the task, the estimated time, the task type tag, and the structured description format of the task result for each virtual task. The task sequence arrangement submodule is used to sort virtual tasks under the virtual job object according to the user's current ability gap vector, the final importance weight of each memory scenario and the user's input time budget, and generate daily or weekly task sequences for virtual jobs.

4. The system according to claim 1, characterized in that, The virtual job evaluation engine, when executing virtual tasks, is further used for: For each virtual task, the user's task performance score, cognitive load score, actual time spent, and behavioral characteristics including the number of edits, retries, and viewing prompts are recorded. The recorded data, along with the virtual task identifier, the virtual business scenario identifier, the associated project identifier, and the set of related concept tags or ability tags, are written into the extended learning and memory unit. Based on the virtual task performance recorded in the extended learning memory unit, the concept mastery is incrementally updated. The incremental update is based on at least the task performance score, the deviation between the actual time and the expected time, the cognitive load score, and the behavioral characteristics to construct a weighted function to obtain the incremental value of concept mastery. The incremental value of concept mastery is then mapped to the incremental score of each ability dimension according to the preset concept-to-ability dimension weight matrix, thereby updating the user's ability dimension score. During the ability dimension score update process, the cognitive load score is used as a learning efficiency adjustment factor. When the cognitive load score is within the preset appropriate load range, the corresponding ability dimension score increment is amplified. When the cognitive load score is lower or higher than the range, the corresponding ability dimension score increment is attenuated.

5. The system according to claim 1, characterized in that, The capacity time prediction module is specifically used for: Based on the latest capability dimension scores and pre-configured concept-to-capability dimension weights, a user's current capability vector is generated. The current capability vector is then compared with the target job capability requirement vector to obtain capability gap values ​​in each capability dimension, forming a capability gap vector. Within a preset time window, time series statistics are performed on the scores of each capability dimension, and the capability growth rate of each capability dimension is calculated using weighted moving average or regression estimation. When the capability gap value of any capability dimension is positive and the corresponding capability growth rate is greater than zero, the expected attainment time of that capability dimension is calculated as the ratio of the capability gap value to the capability growth rate. When the capability growth rate is less than or equal to zero, the expected attainment time of that capability dimension is set to a value greater than the preset upper limit. Based on the preset set of core competency dimensions, the maximum, average or median of the expected time to reach the target for each competency dimension is weighted and combined to obtain the overall expected time to reach the target for the position. The corresponding lower and upper time limits are generated according to the preset fluctuation ratio to represent the time interval for reaching different competency levels.

6. The system according to claim 1, characterized in that, The resume and interview generation module is specifically used to generate project-level resume fragments and job-specific resumes: The job description text of the target position is parsed, and a set of keywords for hard skills, business domain, and behavioral ability is extracted to construct a job feature summary and corresponding feature vector. Using job feature summaries and feature vectors as query conditions, dense vector retrieval and keyword retrieval are performed in learning scenarios and job search scenarios related to the target company, respectively. The search results are scored according to semantic similarity, keyword overlap and scenario type, and then combined with the importance weight of the memory scenario for fusion and ranking to obtain a set of candidate memory scenarios. In the candidate memory scene set, a comprehensive matching score is calculated based on skill coverage, business scenario matching degree, industry consistency and user task performance statistics in the scenario. Several memory scenes with high comprehensive matching scores are selected as source scenes for generating project-level resume fragments. For each source scenario, a preset project-level resume fragment template is selected based on the scenario type. The project title, time range, task content, technology stack, business indicators, and result summary extracted from the scenario are filled into the template to generate a project-level resume fragment. For each project-level resume fragment, an expression intensity level, a deep mining tag array, a risk tag array, and industry tags are set. The expression intensity level is used to control the strength of the words and the degree of achievement description. The risk tag array is used to mark whether it involves learning projects, virtual projects, or simulated environments. The generated project-level resume fragments are combined with the user's basic information. The fragments are sorted and filtered according to their comprehensive matching score and expression intensity level to construct a project experience section. This section is then combined with education experience and skills information to generate a job-customized resume.

7. The system according to claim 1, characterized in that, The resume and interview generation module, when generating interview content, is further used for: Using selected project-level resume segments from job-specific resumes and their associated learning, job-seeking, and interview scenarios as input, we extract key project points, delve into tags, risk tags, and weakness tags, and construct a multi-dimensional set of assessment dimensions covering project background, business logic, technical details, collaboration and communication, and risk control. For each assessment dimension, basic interview questions, in-depth business interview questions, in-depth technical interview questions, and stress interview questions are generated, resulting in a set of multiple types of interview questions; Standard answer text is generated based on factual information recorded in project-level resume fragments and structured fact arrays in corresponding memory scenarios. Industry-enhanced answer text is generated by semantically enhancing the standard answer text based on the industry terminology and case library provided by the industry plugin module. Stress-coping answer text is also generated based on the preset stress coping strategy template. Each interview question and its corresponding standard, industry-enhanced, and stress-coping answers are encapsulated into an interview memory card object. The interview memory card object includes at least the question text, a keyword list, multiple versions of the answer text, structured answer prompts, difficulty level, learning status field, and tag array, and is written into the memory platform module for use in simulated interviews and review.

8. The system according to claim 1, characterized in that, The text compliance review module is specifically used for: The rule engine loads a list of sensitive words for employment relationships, a list of words for leadership responsibilities, and a template for a declaration of learning or virtual projects. Lexical analysis and pattern matching are performed on the output text. When it is detected that the text involving learning or virtual projects is missing words indicating learning or virtual attributes, the template risk is set to medium level. When it is detected that the words used to describe real employment relationships or senior management responsibilities are inconsistent with the source of the memory scenario, the rule risk is set to high level. The semantic analysis model is used to understand the semantics of the whole text, calculate the probability distribution of the text belonging to low-risk, medium-risk and high-risk categories, and determine the semantic risk level. The rule risk level, template risk level, and semantic risk level are converted into numerical scores and weighted and summed to obtain the total risk score. The final risk level is determined by comparing the total risk score with multiple preset risk thresholds. Texts judged as low risk are allowed to be output directly. Texts judged as medium risk require users to explicitly confirm that they are learning projects or virtual projects and a confirmation log is recorded. Texts judged as high risk block output and return modification suggestions and expressive elements that need to be weakened.

9. The system according to claim 1, characterized in that, When updating the weights of the memory scenes, the feedback optimization module is specifically used for: Within a preset statistical period, the number of times each memory scenario is used in project-level resume fragment generation, job-customized resume combination, and virtual job task scheduling is counted, and the number of successful events such as resume passing, interview passing, and job offer obtained associated with that memory scenario is also counted. Based on the number of times the memory scene is used, the number of successful events, and the scene type, the scene success rate is calculated, and the corresponding baseline success rate is obtained from the global statistics of the same position and the same scene type. Based on the difference between the scenario success rate and the baseline success rate, the scenario feedback increment is calculated in combination with the time decay factor, and the scenario feedback increment is weighted with the basic importance to synthesize the final scenario importance value; When performing memory scene retrieval, virtual job task arrangement, and project-level resume fragment generation, the final importance value of the scene is used as a ranking weighting factor, so that memory scenes with higher final importance are given priority under the same matching degree.

10. The system according to claim 1, characterized in that, The system also includes an industry plugin module, which implements a uniformly defined set of interfaces, including at least: The virtual scenario configuration interface is used to define multi-level virtual business scenarios and their task splitting rules for specific industries based on industry knowledge and business processes, providing loadable industry-specific virtual scenario and virtual task configurations for the virtual job evaluation engine; The task scoring and indicator conversion interface is used to weight and score the technical indicators in the virtual task execution results according to industry evaluation standards, and convert the technical indicators into indicator descriptions or business contribution tags that are easy for business personnel to understand. The project description document and project story template rendering interface is used to render factual data from learning scenarios and virtual tasks into industry-style project description documents and project stories in accordance with industry-specific project description templates and project story templates, providing industry context enhancement for project-level resume fragments and interview answers; The industry plug-in module is registered and loaded according to industry and job identifiers, so that the virtual job evaluation engine and resume and interview generation module can adapt to the business context and indicator system of different industries and positions while maintaining the consistency of ability modeling and generation logic.

Images