Go question type recommendation system and method based on user ability evaluation

Abstract

The invention discloses a Go question type recommendation system and method based on user ability evaluation, and relates to the technical field of smart learning. The system is composed of a data acquisition module, a data processing module, a learning recommendation module and a feedback evaluation module. The data acquisition module covers quantification of five-force dimensions, actual combat data analysis and training behavior records of users; and the data processing module generates an accurate three-dimensional feature vector through knowledge graph filtering, recent development area principle sorting and feature extraction. The learning recommendation module intelligently matches the most suitable question type library content according to the feature vector of the user, and ensures that the recommended question can not only fit the current capability level of the user, but also meet the specific learning target of the user; and the feedback evaluation module dynamically updates the capability value of the user by using an IRT model, optimizes recommendation algorithm parameters in combination with user feedback and strengthens error exercise to form a closed-loop feedback mechanism. And according to accurate question type recommendation, the learning effect is further enhanced through continuous feedback and optimization.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent learning technology, and specifically relates to a Go question type recommendation system and method based on user ability assessment. Background Art

[0002] As an intellectual sport with a profound cultural heritage, Go's learning process requires extremely high levels of systematicity, logic, and computational skills. Traditional Go teaching relies primarily on one-on-one instruction from teachers, who use their experience to assess students' proficiency and recommend training exercises. This approach has significant limitations: First, it is difficult for teachers to accurately quantify differences in students' abilities across various dimensions, such as layout, midgame, and endgame; second, it is impossible to achieve truly personalized training in large-scale teaching scenarios. With the development of artificial intelligence technology, computer-based Go learning systems are becoming increasingly popular, but existing technologies still have significant shortcomings in terms of the accuracy of problem recommendations, dynamic adaptability, and multi-dimensional ability assessment.

[0003] In view of the above shortcomings, the present invention proposes a Go question type recommendation system and method based on user ability assessment to help users improve their Go skills more effectively. Summary of the Invention

[0004] In order to overcome the shortcomings and deficiencies of the above-mentioned prior art, the first purpose of the present invention is to provide a Go question type recommendation system based on user ability assessment; the second purpose of the present invention is to provide a Go question type recommendation method based on user ability assessment.

[0005] The first object of the present invention adopts the following technical solution:

[0006] A Go question type recommendation system based on user ability assessment, comprising:

[0007] Data Collection Module: This module collects user Go skill levels, learning preferences, and training goals in real time, and determines user basic data for personalized question type recommendations. It includes a skill collection unit, a preference collection unit, and a goal collection unit.

[0008] Data processing module: retrieves, sorts and extracts features from user basic data to generate Go ability feature data, including data retrieval unit, data sorting unit and feature extraction unit;

[0009] Learning recommendation module: matches the question type library according to ability characteristic data and generates a personalized recommendation list, including question type storage unit, question type index unit, comparative analysis unit, question type retrieval unit, and recommendation display unit;

[0010] Feedback and evaluation module: Optimizes recommendation strategies based on user feedback on questions, including user feedback unit, learning evaluation unit, and improvement and optimization unit.

[0011] Preferably, in the data acquisition module,

[0012] Level collection unit: real-time collection of Go ability data, including five-force dimension data, actual combat data and training data;

[0013] Preference collection unit: real-time collection of learning preferences, including question type preferences, learning habits, and interaction preferences;

[0014] Target collection unit: collects training targets in real time, including short-term targets, long-term targets and special targets;

[0015] In the data processing module,

[0016] Data retrieval unit: Filter invalid data based on the five-force model dimension labels, eliminate records with missing key information, and store them by data source;

[0017] Data sorting unit: Sort by the importance and time series of capability dimensions to generate capability change trajectories for each dimension;

[0018] Feature extraction unit: extracts capability feature vectors, generates weak link labels, and calculates learning requirement weights;

[0019] In the learning recommendation module, the question type storage unit stores a Go question bank labeled with the five strength dimensions. Each question includes the question type, difficulty level, knowledge point label, ability relevance, and prerequisite knowledge requirements.

[0020] Question type indexing unit: Based on the user's ability characteristics, it prioritizes indexing question types within the "zone of proximal development" and filters question types according to the user's training goals;

[0021] Comparative analysis unit: calculates the cosine similarity between question type features and user ability vectors, verifies knowledge prerequisites, and generates a matching score;

[0022] Question type retrieval unit: retrieves candidate question types in descending order of matching scores to control repetition rate;

[0023] Recommended display unit: Displays a recommended list classified by the five forces dimensions, and marks the training value of each question;

[0024] In the feedback evaluation module, the user feedback unit: collects answer data and user subjective feedback;

[0025] Learning Assessment Unit: Use the IRT model to update the capability values ​​of each dimension and generate a recommendation effect report;

[0026] Improved optimization unit: dynamically adjust recommendation algorithm parameters and optimize question bank annotations.

[0027] Preferably, the five-force dimension data collected by the horizontal collection unit includes layout planning ability, attack and defense conversion ability, calculation ability, life and death judgment ability, and endgame handling ability, and the ability value of each dimension is calculated by correcting the standardized test score and the actual combat error rate, wherein the standardized test score formula is:

[0028] Standardized test scores: (i=1,2,3,4,5); where ω i is the dimension weight, which is preset based on the survey data of professional chess players;

[0029] The actual combat error rate correction formula is: The higher the error rate, the greater the ability value discount; E i is the error rate of actual chess records,

[0030] Preferably, the three-dimensional feature vector generated by the data processing module includes:

[0031] Ability level: C i is the capability value of the i-th dimension; ω i is the weight coefficient of the i-th dimension; is the weight normalization factor;

[0032] Weak label generation: If C i <Ability level × set weight and E i >Average error rate, then it is marked as a weak dimension;

[0033] Target weight calculation: Generate weight vector based on user-set target

[0034] Preferably, the question type matching degree calculation of the learning recommendation module includes:

[0035] Cosine similarity:

[0036] Target weighted matching:

[0037] α1 and α2 are weight coefficients, and α1+α1=1.

[0038] Preferably, the question type index unit of the learning recommendation module filters question types based on the zone of proximal development, which is defined as [C 等级 -1,C 等级 +2], which is the difficulty level interval; the question type must meet Q 难度 ∈[C 等级 -1,C 等级 +2] and Priority sorting: sorting value = final matching degree × (1 + 0.1 × target dimension correlation).

[0039] Preferably, the feedback evaluation module uses the IRT model to update the capability value and the IRT model to update the capability value of each dimension. Among them, θ is the user ability value, and the initial value is set to the global average level; a is the question type discrimination; b is the question type difficulty, which is related to the difficulty Q 难度 Corresponding ability update formula: where y i The result of the user's answer (1 = correct, 0 = wrong); i is the predicted correct probability, and m is the number of answered questions.

[0040] Preferably, the improvement and optimization unit of the feedback evaluation module includes:

[0041] A reinforcement mechanism for wrong questions: for questions with a set number of errors or more, an equal number of variant questions with the same knowledge points and difficulty level are associated;

[0042] Periodic evaluation, generating weekly capacity improvement reports and adjusting recommended strategies for the next week;

[0043] Expert calibration, monthly import of professional chess players to analyze question types and update question bank annotations;

[0044] Dynamic adjustments include: for question types whose correct rate is greater than the set threshold for a set number of consecutive times, a priority attenuation coefficient is recommended; for question types whose error rate is greater than the set threshold, the weight of associated variant questions is increased through the associated variant question weight reinforcement coefficient.

[0045] The second object of the present invention adopts the following technical solution:

[0046] A Go question type recommendation method based on user ability assessment is used to implement a Go question type recommendation system based on user ability assessment. The method flow is as follows:

[0047] Step 1: Data collection;

[0048] Level assessment: Quantify the five dimensions of strength, including layout, attack and defense, calculation, life and death, and endgame; analyze actual chess records: mark key mistakes; record training behavior: track error book operations and AI review time;

[0049] Preference capture: Obtain user training focus and active time periods;

[0050] Goal analysis: clarify short-term, long-term and special goals;

[0051] Step 2: Data processing;

[0052] Cleaning and classification: filter invalid data and store them according to actual combat / test / training data;

[0053] Dynamic sorting: adjust dimension weights according to user stages to generate capability trajectories; feature modeling: construct a three-dimensional vector: [capability level, weakness label, target weight];

[0054] Step 3: Question type matching and recommendation;

[0055] Question bank screening: prioritize indexing "Zone of Proximal Development" question types and filter question types by objectives;

[0056] Smart matching: calculate question type-user similarity and verify preconditions;

[0057] Dynamic adjustment: reduce the priority of frequently correct questions and increase the number of related variants of frequently incorrect questions. When the match is insufficient, expand upward, consolidate downward, and connect across dimensions.

[0058] Recommended display: display by five forces classification, marking the training value;

[0059] Step 4: Feedback evaluation and optimization;

[0060] Effect evaluation: The IRT model updates the capability value and generates a report including target progress and error repetition rate;

[0061] Dynamic optimization: adjust recommendation parameters; optimize question bank annotations and strengthen training, including wrong question reinforcement, periodic evaluation, and expert calibration.

[0062] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0063] 1. A comprehensive assessment of Go proficiency, from basic skills to practical application, is achieved through a multi-dimensional quantification of the user's Go abilities using the Five Forces model (layout planning, attack-defense conversion, calculation, life-or-death judgment, and endgame handling). This is combined with actual game record analysis (win rate curves, key error predictions), standardized ability testing (50 Five Forces-specific questions), and training behavior logs (error book manipulation, AI review interaction). The data processing module further filters invalid data through the knowledge graph, generating a three-dimensional feature vector containing "ability level, weak label, and target weight," pinpointing the user's weaknesses (e.g., "endgame handling ability is lower than the global rank by 2") and constructing a dynamically updated, personalized profiling of their abilities.

[0064] 2. Based on Vygotsky's "Zone of Proximal Development" theory, the recommendation module prioritizes indexing questions with a difficulty level within ±0.5 of the user's current ability. It calculates the match between the question type and the user's ability vector using cosine similarity, and dynamically adjusts the dimension weights based on the training objectives (e.g., "layout reinforcement goal" increases the layout relevance weight by 40%). When the match is insufficient, a strategy of "upward expansion + downward consolidation + cross-dimensional association" is implemented (e.g., recommending questions 0.5 levels higher with additional AI analysis, inserting 1-2 basic questions, and associating mid-game and closing questions to strengthen weak links in the endgame) to ensure that the recommended questions are always within the optimal training range of "jumping to reach." At the same time, by controlling the repetition rate of questions with the same knowledge point to ≤2 times within 7 days and marking the training value of each question (e.g., "expected improvement in computing power +0.3 levels"), ineffective practice is avoided and training becomes more targeted.

[0065] 3. The feedback and evaluation module uses an IRT model to update ability values ​​for each dimension in real time (difficulty + 1 level → ability value + 0.5), dynamically adjusting recommendation strategies based on the weekly "Ability Improvement Report." For questions answered incorrectly three or more times, three variations on the same knowledge point are automatically linked to form a reinforcement mechanism. Typical question types analyzed by professional chess players are introduced monthly, and the question bank is updated through expert calibration to ensure the accuracy of knowledge point linkage and difficulty assessment. This closed-loop mechanism achieves a continuous optimization cycle of "assessment-recommendation-training-reassessment." BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0067] Figure 1 A module diagram of a Go question type recommendation system based on user ability assessment according to the present invention is shown;

[0068] Figure 2 A flowchart of a Go question type recommendation method based on user ability assessment of the present invention is shown. DETAILED DESCRIPTION

[0069] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0070] In addition, the described features, structures or characteristics can be combined in any suitable manner in one or more example embodiments. In the following description, many specific details are provided to provide a full understanding of the example embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure can be practiced while omitting one or more of the specific details, or other methods, components, steps, etc. can be adopted. In other cases, well-known structures, methods, implementations or operations are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0071] Example 1:

[0072] See Figure 1 As shown, a Go question type recommendation system based on user ability assessment in this embodiment includes a data collection module, a data processing module, a learning recommendation module and a feedback assessment module.

[0073] The data acquisition module includes a level acquisition unit, a preference acquisition unit, and a target acquisition unit; the data processing module includes a data retrieval unit, a data sorting unit, and a feature extraction unit; the learning recommendation module includes a question type storage unit, a question type index unit, a comparative analysis unit, a question type retrieval unit, and a recommendation display unit; the feedback evaluation module includes a user feedback unit, a learning evaluation unit, and an improvement and optimization unit.

[0074] The system achieves accurate recommendations through the following closed loop: users take questions → feedback evaluation module updates ability values ​​→ data processing module generates new features → learning recommendation module dynamically adjusts question type library matching strategy → generates the next round of recommendations.

[0075] Data collection module: used to collect the user's Go ability level, learning preferences and training goals in real time, and determine the user's basic data based on personalized question type recommendations, including level collection unit, preference collection unit, and goal collection unit.

[0076] Horizontal acquisition unit: real-time collection of Go ability data, including:

[0077] Five-force dimension data: layout planning ability (correctness of fixed pattern selection, efficiency score of edge removal), attack and defense conversion ability (correctness of mid-game attack and defense decision-making), calculation ability (depth level of life and death problem calculation), life and death judgment ability (common life and death type recognition rate), endgame handling ability (error rate of point calculation); five-force dimension vector The capability values ​​of each dimension are calculated in the following way:

[0078] Standardized test scores: (i=1,2,3,4,5); where ω i is the dimension weight, which is preset based on the survey data of professional chess players.

[0079] Actual game data: The win rate curve and key wrong move positions (such as the number of wrong decisions in ko fights) are obtained through the chess record analysis engine; the error rate of actual chess records: Ability value correction: The higher the error rate, the greater the ability value discount.

[0080] Training data: daily problem-solving records (accuracy, time, error type labels, such as "missed calculation of tricks" and "wrong order of endgame moves").

[0081] When collecting Go ability data in real time, it is obtained through the following methods:

[0082] Standardized ability test: includes 50 questions on the five forces (layout filling-in-the-blank, attack and defense multiple-choice questions, and life and death calculation timed questions);

[0083] Practical chess game analysis: supports importing SGF files from mainstream chess platforms and automatically marks key skill dimension error points;

[0084] Interaction log recording: Track users' operation behaviors in the wrong question book and AI review functions, and quantify learning preferences.

[0085] Preference collection unit: collects learning preferences in real time, including:

[0086] Question type preference: user-selected training focus (e.g., "focus on mid-game offense and defense");

[0087] Study habits: duration of daily training, preferred time period for doing exercises (such as evening special training); interaction preferences: whether to enable AI review and analysis, and frequency of use of the wrong question book.

[0088] Target acquisition unit: collects training targets in real time, including:

[0089] Short-term goal: Improve a certain dimension of ability within a week (e.g., "increase computing power from level 7 to level 8");

[0090] Long-term goal: rank sprint (e.g., "from amateur 4th dan to 5th dan within 3 months");

[0091] Special goal: Overcome weak links (such as "mastering the changes in star-position attack and defense").

[0092] The beneficial effects of this embodiment include: It achieves precise ability profiling through multi-dimensional data collection and quantification using the Five Forces model; dynamically adjusts recommendation strategies based on user preferences and goals, improving question type matching; and a closed-loop feedback mechanism continuously optimizes recommendation effectiveness, shortening the user's ability improvement path and significantly improving the personalization and efficiency of Go training. The data collection module comprehensively assesses ability using the Five Forces model, actual game analysis, and training data. Dynamically adapting recommendations based on user preferences and goals, standardized testing and SGF import improve data accuracy, and real-time feedback optimizes learning paths, significantly enhancing personalized training results.

[0093] Example 2:

[0094] Data processing module: used to retrieve, sort and extract features of user basic data to generate Go ability feature data, including data retrieval unit, data sorting unit and feature extraction unit.

[0095] The data processing module performs the following operations: filtering data based on the Go knowledge graph, for example, excluding question data that is not related to the user's current knowledge points; arranging candidate question types in descending order of difficulty matching according to the "zone of proximal development first" principle; generating a three-dimensional feature vector containing "ability level, weak label, and target weight" as input to the recommendation algorithm.

[0096] Ability level: C i is the capability value of the i-th dimension; ω i is the weight coefficient of the i-th dimension; is the weight normalization factor (to ensure the total weight is 1)

[0097] Weak label generation: If C i <Ability level × 0.8 and E i >Average error rate, then it is marked as a weak dimension;

[0098] Target weight calculation: Generate a weight vector based on the user-set goal (such as "improve life and death ability within 3 months").

[0099] Data retrieval unit: Based on the dimension labels of the Five Forces Model, filter invalid data (such as repeated submissions of answers to the same question); eliminate records with missing key information (such as wrong questions without marked error types); and store data by category (actual data / test data / training data).

[0100] Data sorting unit: Sort by the importance of capability dimensions (e.g. beginners prioritize life and death / layout data, while advanced users focus on mid-game / endgame data); sort by time series to generate capability change trajectories for each dimension (e.g., a computing power improvement curve for the past 30 days).

[0101] Feature extraction unit: Extract capability feature vectors: [layout level, attack and defense level, calculation level, life and death level, endgame level, global rank]; generate weak link labels (such as "endgame processing ability - level 2 (lower than global rank)"); calculate learning demand weights (based on target collection unit data, such as "layout strengthening target → layout feature weight + 30%").

[0102] The beneficial effects of this embodiment are as follows: the module filters invalid data through the knowledge graph to improve processing efficiency; dynamically sorts ability dimensions according to user stages to adapt to personalized needs; generates three-dimensional feature vectors to quantify ability shortcomings and target weights, provides accurate input for the recommendation algorithm, and significantly optimizes question type matching and training targeting.

[0103] Example 3:

[0104] Learning recommendation module: matches the question type library according to the ability characteristic data and generates a personalized recommendation list, including question type storage unit, question type index unit, comparative analysis unit, question type retrieval unit, and recommendation display unit.

[0105] Question type storage unit: stores a Go question bank marked with the five force dimensions. Each question contains: {question type (life and death / fixed pattern / endgame, etc.), difficulty level (level 1-9), knowledge point label (such as "corner life and death - club five", "star fixed pattern - point three three variation"), ability correlation (five-dimensional weight value 0-10 points), prerequisite knowledge requirements}.

[0106] Define the question type feature vector

[0107] where Q 关联度i It is the correlation between the question type and the ability of the i-th dimension (0-100).

[0108] Cosine similarity:

[0109] Target weighted matching:

[0110] α1 and α2 are weight coefficients, and α1+α1=1.

[0111] Difficulty adaptation coefficient: Q 难度 -C 等级 |≤2Final matching degree: Final matching degree = matching degree × difficulty coefficient.

[0112] Question type indexing unit: Based on user ability characteristics, priority is given to indexing question types within the "zone of proximal development" (difficulty = current ability ± 0.5 levels); question types are filtered according to user training goals (such as "endgame weakness-strengthening goal → only index endgame question types and related mid-game closing connection questions").

[0113] Definition of the Zone of Proximal Development: 等级 -1,C 等级 +2](difficulty level interval);

[0114] Question type filter: Q 难度 ∈[C 等级 -1,C 等级 +2] and

[0115] Priority sorting: sorting value = final matching degree × (1 + 0.1 × target dimension correlation).

[0116] Comparative analysis unit: Calculate the cosine similarity between question type characteristics and user ability vectors (the formula refers to the three-dimensional matching model of the original solution); verify knowledge prerequisites (for example, if the user has not mastered the "three-three pattern", exclude mid-game offense and defense questions that rely on this pattern); generate a matching score (questions with a similarity ≥ 0.6 and that meet the prerequisites are included in the candidate set).

[0117] The comparative analysis process includes:

[0118] Initial screening of question types: Filter questions that are significantly higher / lower than the current level based on the user's global level (difficulty difference > 1.5 levels);

[0119] Dimension matching: Calculate the weighted sum of the question type ability correlation and the user's goal weight (e.g., if the user sets "layout enhancement 40%", then layout correlation × 0.4 + other dimensions × 0.6);

[0120] Dynamic adjustment: For questions with a correct rate > 80% for three consecutive times, the recommendation priority will be lowered; for questions with an error rate > 60%, recommendations for related similar questions will be added.

[0121] Continuous accuracy processing: If a question type is answered correctly n times in a row, the recommendation priority will be attenuated.

[0122] Error rate enhancement: If the error rate of a question type is greater than 70%, the weight of the related variant questions will be increased:

[0123] Question type retrieval unit: retrieve candidate question types in descending order of matching scores, recommending 3-5 questions for each dimension; control the repetition rate (questions on the same knowledge point cannot be retrieved more than twice within 7 days).

[0124] When the question type matching degree is insufficient, perform the following operations:

[0125] Upward expansion: recommend similar questions with a difficulty level 0.5 higher (with additional AI analysis prompts); downward consolidation: insert 1-2 basic questions on the same knowledge point to ensure that the "basic-advanced" question type ratio is 1:3; cross-dimensional association: recommend derivative question types related to weak dimensions (for example, when the endgame is weak, add mid-game closing connection questions).

[0126] Recommended display unit: Displays a list of recommendations classified by the five strength dimensions (e.g., "Calculation power training: 3 life and death questions, 2 killing questions"); marks the training value of each question (e.g., "This question strengthens the calculation of robbery materials, corresponding to an expected improvement in calculation power of +0.3 levels").

[0127] The beneficial effects of this embodiment include: the learning recommendation module accurately adapts to the user's ability shortcomings and goals through multi-dimensional question type matching and dynamic adjustment strategies, thereby improving training efficiency; intelligent initial screening, weighted matching and cross-dimensional recommendation mechanisms avoid ineffective practice, and combine AI analysis and value labeling to enhance learning targeting and significantly optimize personalized learning paths.

[0128] Example 4:

[0129] Feedback and evaluation module: Optimizes recommendation strategies based on user feedback on questions, including user feedback unit, learning evaluation unit, and improvement and optimization unit.

[0130] User feedback unit: collects answer data: accuracy rate, time, and incorrect step annotations (such as "miscalculated breakpoint in the 5th move"); collects user subjective feedback: difficulty evaluation (too easy / moderate / too difficult), and analysis satisfaction.

[0131] Learning Assessment Unit: Using the IRT model to update the ability values ​​of each dimension (Answering the difficulty + 1 level question correctly → ability value +0.5, answering the same type of questions incorrectly continuously → ability value -0.3); where θ is the user's ability value (latent variable), and the initial value is set to the global average level; a is the question type discrimination (preset in the question bank, such as a=1.2 for life and death questions); b is the question type difficulty (related to the difficulty Q 难度 Corresponding). Ability update formula: where y i The result of the user's answer (1 = correct, 0 = wrong); i is the predicted correct probability, and m is the number of questions answered.

[0132] Generate recommendation effect report: target dimension improvement progress, wrong question repetition rate, average time change.

[0133] Improved optimization unit: Dynamically adjust the recommendation algorithm parameters (for example, when the recommendation accuracy of a certain dimension is <60%, expand the range of candidate question types to ±1 level of difficulty); optimize question bank annotation (supplement "easy-to-make mistakes knowledge points" labels based on users' frequently incorrect questions).

[0134] Improvements and optimizations include:

[0135] Wrong question reinforcement mechanism: For questions that have been wrong three or more times, three variations of the same knowledge point and difficulty level will be automatically associated;

[0136] Periodic evaluation: Generate a "Capability Improvement Report" every week, compare the completion of the five-force dimension goals, and dynamically adjust the recommended strategy for the next week;

[0137] Expert calibration: Typical question types analyzed by professional chess players are imported every month to update the knowledge point correlation and difficulty marking of the question bank.

[0138] The beneficial effects of this embodiment include: the feedback evaluation module dynamically optimizes the ability evaluation through the IRT model and adjusts the recommendation strategy in a closed loop; the reinforcement of wrong questions and periodic evaluation accurately overcome weak links; expert calibration ensures the quality of the question bank, forming an efficient "evaluation-feedback-optimization" cycle, significantly improving training effects and users' long-term progress.

[0139] Example 5:

[0140] See Figure 2 As shown, the process of the Go question type recommendation method based on user ability assessment in this embodiment is as follows:

[0141] Step 1: Data collection.

[0142] Level Assessment: Quantify the five key dimensions of strategy, including layout (fixed pattern accuracy / edge-breaking efficiency), offense and defense (ko fight decision accuracy), calculation (life and death problem depth), life and death (common pattern recognition rate), and endgame (point error rate). Analyze actual game records: Annotate key mistakes (e.g., "Ko fight misjudgment win rate decreased by 12%"). Record training behavior: Track error log entries and AI review time.

[0143] Preference capture: Obtain user training focus and active time periods.

[0144] Goal analysis: clarify short-term, long-term and special goals.

[0145] Step 2: Data processing.

[0146] Cleaning and classification: Filter invalid data and store them according to actual combat / test / training data.

[0147] Dynamic ranking: Adjust dimension weights by user stage. Generate capability trajectories. Feature modeling: Construct a three-dimensional vector: [ability level, weakness label, target weight].

[0148] Step 3: Question type matching and recommendation.

[0149] Question bank screening: Prioritizes "Zone of Proximal Development" questions. Filter questions by objective.

[0150] Smart matching: Calculate question type-user similarity. Verify preconditions.

[0151] Dynamic Adjustment: Lower the priority of frequently correct questions and increase the priority of related variations of frequently incorrect questions. When matching is insufficient, expand upward, consolidate downward, and connect across dimensions.

[0152] Recommended display: Display by five forces classification and mark the training value.

[0153] Step 4: Feedback evaluation and optimization.

[0154] Effectiveness evaluation: The IRT model updates the capability value. Generates a report including target progress and error repetition rate.

[0155] Dynamic Optimization: Adjust recommendation parameters. Optimize question bank annotations. Enhance training, including reinforcement of incorrect questions, periodic evaluation, and expert calibration.

[0156] The beneficial effects of this embodiment are as follows: the method realizes accurate ability profiling through multi-dimensional data collection and dynamic feature modeling; the intelligent matching algorithm dynamically recommends question types in combination with user goals to improve the targeted training; the closed-loop feedback mechanism continuously optimizes the strategy, and the reinforcement of wrong questions and expert calibration ensure long-term progress, significantly shortening the user's ability improvement cycle and improving the efficiency of Go training.

[0157] The weights of this invention are used to measure the degree to which different factors or variables influence a particular outcome or decision. Weights are defined as numerical values ​​assigned to each factor when comparing and evaluating multiple factors, reflecting its importance or priority. These weights can be determined based on specific circumstances and needs, and are typically developed and confirmed by professionals or stakeholders. By properly setting weights, programs or systems can be made to make more accurate decisions or predictions.

[0158] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

[0159] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to specific embodiments. Obviously, many modifications and variations are possible based on the contents of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A Go question type recommendation system based on user ability assessment, characterized by: The system comprises: Data Collection Module: This module collects user Go skill levels, learning preferences, and training goals in real time, and determines user basic data for personalized question type recommendations. It includes a skill collection unit, a preference collection unit, and a goal collection unit. Data processing module: retrieves, sorts and extracts features from user basic data to generate Go ability feature data, including data retrieval unit, data sorting unit and feature extraction unit; Learning recommendation module: matches the question type library according to ability characteristic data and generates a personalized recommendation list, including question type storage unit, question type index unit, comparative analysis unit, question type retrieval unit, and recommendation display unit; Feedback and evaluation module: Optimizes recommendation strategies based on user feedback on questions, including user feedback unit, learning evaluation unit, and improvement and optimization unit.

2. A Go question type recommendation system based on user ability assessment according to claim 1, characterized in that: In the data acquisition module, Level collection unit: real-time collection of Go ability data, including five-force dimension data, actual combat data and training data; Preference collection unit: real-time collection of learning preferences, including question type preferences, learning habits, and interaction preferences; Target collection unit: collects training targets in real time, including short-term targets, long-term targets and special targets; In the data processing module, Data retrieval unit: Filter invalid data based on the five-force model dimension labels, eliminate records with missing key information, and store them by data source; Data sorting unit: Sort by the importance and time series of capability dimensions to generate capability change trajectories for each dimension; Feature extraction unit: extracts capability feature vectors, generates weak link labels, and calculates learning requirement weights; In the learning recommendation module, the question type storage unit stores a Go question bank marked with the five strength dimensions, where each question includes question type, difficulty level, knowledge point label, ability relevance, and prerequisite knowledge requirements. Question type indexing unit: Based on the user's ability characteristics, it prioritizes question types within the zone of proximal development and filters question types according to the user's training goals; Comparative analysis unit: calculates the cosine similarity between question type features and user ability vectors, verifies knowledge prerequisites, and generates a matching score; Question type retrieval unit: retrieves candidate question types in descending order of matching scores to control repetition rate; Recommended display unit: Displays a recommended list classified by the five forces dimensions, and marks the training value of each question; In the feedback evaluation module, the user feedback unit: collects answer data and user subjective feedback; Learning Assessment Unit: Use the IRT model to update the capability values ​​of each dimension and generate a recommendation effect report; Improved optimization unit: dynamically adjust recommendation algorithm parameters and optimize question bank annotations.

3. A Go question type recommendation system based on user ability assessment according to claim 1, characterized in that: The five-dimensional data collected by the horizontal collection unit include layout planning ability, attack and defense conversion ability, calculation ability, life and death judgment ability, and endgame handling ability. The ability value of each dimension is calculated by correcting the standardized test score and the actual combat error rate. The standardized test score formula is: Standardized test scores: Among them, ω i is the dimension weight, which is preset based on the survey data of professional chess players; The actual combat error rate correction formula is: The higher the error rate, the greater the ability value discount; E i The error rate of actual chess records, 4. A Go question type recommendation system based on user ability assessment according to claim 1, characterized in that: The three-dimensional feature vector generated by the data processing module includes: Ability level: C i is the capability value of the i-th dimension; ω i is the weight coefficient of the i-th dimension; is the weight normalization factor; Weak label generation: If C i <Ability level × set weight and E i >Average error rate, then it is marked as a weak dimension; Target weight calculation: Generate weight vector based on user-set target 5. A Go question type recommendation system based on user ability assessment according to claim 1, characterized in that: The question type matching degree calculation of the learning recommendation module includes: Cosine similarity: Target weighted matching: α1 and α2 are weight coefficients, and α1+α1=1.

6. A Go question type recommendation system based on user ability assessment according to claim 1, characterized in that: The question type index unit of the learning recommendation module filters question types based on the zone of proximal development, which is defined as [C 等级 -1,C 等级 +2], which is the difficulty level interval; the question type must meet Q 难度 ∈[C 等级 -1,C 等级 +2] and Priority sorting: sorting value = final matching degree × (1 + 0.1 × target dimension correlation).

7. A Go question type recommendation system based on user ability assessment according to claim 1, characterized in that: The feedback evaluation module uses the IRT model to update the capability value and the IRT model to update the capability value of each dimension. Among them, θ is the user ability value, and the initial value is set to the global average level; a is the question type discrimination; b is the question type difficulty, which is related to the difficulty Q 难度 Corresponding ability update formula: where y i The result of the user's answer (1 = correct, 0 = wrong); i is the predicted correct probability, and m is the number of answered questions.

8. A Go question type recommendation system based on user ability assessment according to claim 2, characterized in that: The improvement and optimization unit of the feedback evaluation module includes: A reinforcement mechanism for wrong questions: for questions with a set number of errors or more, an equal number of variant questions with the same knowledge points and difficulty level are associated; Periodic evaluation, generating weekly capacity improvement reports and adjusting recommended strategies for the next week; Expert calibration, monthly import of professional chess players to analyze question types and update question bank annotations; Dynamic adjustments include: for question types whose correct rate is greater than the set threshold for a set number of consecutive times, a priority attenuation coefficient is recommended; for question types whose error rate is greater than the set threshold, the weight of associated variant questions is increased through the associated variant question weight reinforcement coefficient.

9. A method for recommending Go problem types based on user ability assessment, for implementing a Go problem type recommendation system based on user ability assessment as claimed in any one of claims 1 to 8, characterized in that: The method flow is as follows: Step 1: Data collection; Level assessment: Quantify the five dimensions of strength, including layout, attack and defense, calculation, life and death, and endgame; analyze actual chess records: mark key mistakes; record training behavior: track error book operations and AI review time; Preference capture: Obtain user training focus and active time periods; Goal analysis: clarify short-term, long-term and special goals; Step 2: Data processing; Cleaning and classification: filter invalid data and store them according to actual combat / test / training data; Dynamic sorting: adjust dimension weights according to user stages to generate capability trajectories; Feature modeling: construct a three-dimensional vector: [ability level, weakness label, target weight]; Step 3: Question type matching and recommendation; Question bank screening: prioritize the questions in the zone of proximal development and filter the questions by objectives; Smart matching: calculate question type-user similarity and verify preconditions; Dynamic adjustment: reduce the priority of frequently correct questions and increase the number of related variants of frequently incorrect questions. When the match is insufficient, expand upward, consolidate downward, and connect across dimensions. Recommended display: display by five forces classification, marking the training value; Step 4: Feedback evaluation and optimization; Effect evaluation: The IRT model updates the capability value and generates a report including target progress and error repetition rate; Dynamic optimization: adjust recommendation parameters; optimize question bank annotations and strengthen training, including wrong question reinforcement, periodic evaluation, and expert calibration.

Images