Semantic model-based power grid dispatching adaptive evaluation question generation method and system
Through the adaptive assessment technology of grid scheduling based on semantic model, combined with the network model of scheduling knowledge and skill points, the adaptive knowledge assessment questions in the field of grid scheduling are generated, and the problems of insufficient diversity and depth in the existing technology are solved, achieving more efficient Knowledge evaluation and learning effect.
Patent Information
- Application Number
- CN202411418797.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-12
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2044-10-12
AI Technical Summary
The existing automatic generation technology of power grid scheduling cannot fully utilize the characteristics of the power grid scheduling field, limiting the diversity and depth of the questions, especially in terms of the description of knowledge points and the adaptability of difficulty.
Adaptive knowledge assessment questions are generated by using a semantic model based on scheduling knowledge and skill point network model and scheduling procedures, accident plans, safety work instructions and other texts. By constructing a scheduling test point network model, relevant knowledge points and their key sentences are extracted Or paragraph, dynamically adjust the difficulty of the question to meet the needs of different students.
提高了题目生成的多样性和深度,能够更精确地评估调度员的知识掌握和复杂关系理解能力,增强了题目的挑战性和教育价值,同时通过个性化生成和动态更新,提升了学习和考核 targeted and timely.
Smart Images

Figure CN118939789A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of power grid simulation intelligent training, and in particular relates to a method and system for generating power grid dispatch adaptive evaluation questions based on a semantic model. Background Art
[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.
[0003] In order to improve the efficiency and effectiveness of training and assessment, automatic question generation technology is applied to the training of power grid dispatchers. It can not only solve many problems existing in traditional methods, but also significantly increase the diversity and flexibility of questions, thereby meeting higher-level assessment needs.
[0004] Existing automatic generation technologies are mainly divided into two categories: text-based automatic generation technology and knowledge graph-based automatic generation technology.
[0005] Text-based question generation technology uses natural language processing and machine learning algorithms to extract text knowledge from electronic documents and corpora to generate questions that meet test requirements. Early methods relied on templates and rules to extract key information from text through grammatical rules. Neural networks, such as long short-term memory networks, convolutional neural networks, Transformer models, and BERT pre-trained language models are also used, but they require specific preprocessing of the input text, such as sentence simplification and semantic disambiguation. The generated questions can only be limited to the text itself, and it is impossible to obtain information associated with the text at the same time, which limits the scope and depth of the examination of the questions.
[0006] The automatic generation technology based on knowledge graphs uses big data and artificial intelligence technology to generate educational test questions using knowledge graphs. It mainly focuses on building small-scale knowledge graphs and generating questions through simple rules. Knowledge graphs intuitively display knowledge concepts or entities (nodes) and their relationships (edges) in a graphical way. However, current technology mainly generates interference items for concept nouns, and there are still limitations in processing sentences describing knowledge points, which limits the difficulty and adaptability of questions. Summary of the invention
[0007] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides a method and system for generating adaptive assessment questions for power grid dispatching based on a semantic model. The method and system can combine the characteristics of the power grid dispatching field, combine the dispatching knowledge and skill point network model with texts such as dispatching procedures, accident plans, safe operation instructions and power grid dispatching professional knowledge, and make full use of the relational features of the knowledge and skill point network model as a structured semantic knowledge base and the rich connotations of the text to generate adaptive knowledge assessment questions in the field of power grid dispatching.
[0008] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: A first aspect of the present invention provides a method for generating adaptive evaluation questions for power grid dispatching based on a semantic model.
[0009] A method for generating questions and answers for adaptive evaluation of power grid dispatching based on a semantic model, comprising: Get the scheduling procedure text data; Construct a network model of dispatching knowledge and skills points, and combine it with the dispatching procedure text data to form a dispatching knowledge base; Preprocessing the scheduling knowledge base to generate a scheduling test point network model with key sentences attached; Based on the scheduling test point network model of the key sentences, relevant knowledge points and their key sentences or key paragraphs are extracted to adaptively generate the test question stem; Generate interference items for the assessment questions according to the knowledge point association relationship of the scheduling test point network model; Conduct an initial difficulty setting analysis on the questions, and then dynamically adjust the difficulty of the questions based on the students' test results.
[0010] A second aspect of the present invention provides a semantic model-based adaptive evaluation question generation system for power grid dispatching.
[0011] A power grid dispatching adaptive evaluation question generation system based on semantic model, comprising: The semantic model building module is configured to: obtain the scheduling procedure text data; Construct a network model of dispatching knowledge and skills points, and combine it with the dispatching procedure text data to form a dispatching knowledge base; Preprocessing the scheduling knowledge base to generate a scheduling test point network model with key sentences attached; The adaptive test question stem generation module is configured to: extract relevant knowledge points and their attached key sentences or key paragraphs based on the scheduling test point network model of the attached key sentences, and adaptively generate the test question stem; The interference item generation module is configured to: generate interference items for the assessment questions according to the knowledge point association relationship of the scheduling test point network model; The question quality assessment module is configured to: perform an initial difficulty setting analysis on the questions, and then dynamically modify the difficulty of the questions based on the test results of the students.
[0012] The third aspect of the present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in a method as described in the first aspect of the present invention are implemented.
[0013] A fourth aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in a method as described in the first aspect of the present invention.
[0014] One or more of the above technical solutions have the following beneficial effects: The present invention provides a method and system for generating self-adaptive assessment questions for power grid dispatching based on semantic models, which can combine the characteristics of the power grid dispatching field, combine the dispatching knowledge and skill point network model with dispatching regulations, accident plans, safety operation instructions and power grid dispatching professional knowledge and other texts, and make full use of the knowledge and skill point network model as the relationship characteristics of the structured semantic knowledge base and the rich connotation of the text to generate self-adaptive knowledge assessment questions in the field of power grid dispatching. This method can not only give play to the flexibility of text-based technology, but also utilize the comprehensiveness and relevance of knowledge graphs to improve the diversity and depth of question generation.
[0015] In the present invention, a new pseudo-proposition construction method is designed based on the knowledge point association relationship based on the dispatch test point network model, which is specifically aimed at judgment questions and multiple-choice questions, including key number replacement, contrast word replacement, and affirmative and negative word replacement, aiming to more accurately generate interference items that can effectively confuse dispatchers. Through this strategy based on association relationships, we can improve the quality and educational value of the questions, not only testing the dispatcher's basic knowledge mastery, but also evaluating his ability to understand the complex relationships between knowledge points. This method not only makes the questions more challenging, but also helps dispatchers to think and analyze more deeply when solving practical problems.
[0016] In the present invention, it is also possible to combine the learning records and behavior data of students or employees, use intelligent algorithms to generate personalized questions, and improve the pertinence of learning and assessment. By using big data and real-time data analysis technology, the knowledge graph and generation algorithm are dynamically updated to ensure the timeliness and accuracy of question generation.
[0017] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.
[0019] Figure 1 A flowchart of constructing a scheduling test site network model in Embodiment 1 of the present invention; Figure 2 The present invention is a flowchart of a method for automatically generating a fill-in-the-blank question stem in the first embodiment of the present invention. DETAILED DESCRIPTION
[0020] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.
[0021] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.
[0022] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other. Embodiment 1 This embodiment discloses a method for generating questions for adaptive evaluation of power grid dispatching based on a semantic model, and constructs a question generation framework based on a knowledge and skill point network model with key sentences attached, such as Figure 1 As shown, in order to more clearly illustrate this embodiment, a method for generating adaptive evaluation questions for power grid dispatching based on a semantic model can be specifically described as follows: Step 1: Obtain the scheduling procedure text data; The dispatching procedure text data obtained is specifically based on the data information of SCADA / EMS, OMS, PMS, security and other systems, dispatching rules and regulations, technical specifications, management regulations, training specifications, safe operation instructions, power dispatcher professional ability training standard teaching materials, power grid dispatching related basic teaching materials and other related materials. In accordance with the professional skill standards for the scope of knowledge and skills that power dispatchers at all levels need to master, the relevant content of power grid dispatching is sorted out.
[0023] Step 2: Construct a dispatching knowledge and skill point network model, and combine it with the dispatching procedure text data to form a dispatching knowledge base; The specific steps to construct the scheduling knowledge and skill point network model are as follows: Step 2-1: Form key knowledge points and skill points based on the "Grid Dispatching Terminology Specifications", power industry vocabulary standards and professional skill standards.
[0024] Step 2-2: For the structured and semi-structured data of SCADA / EMS, OMS, PMS, security and other systems, combined with step 2-1, extract the knowledge and skill points including equipment knowledge, power grid structure knowledge, power grid business process, etc., and build the association relationship between knowledge and skill points according to the storage structure and relationship in the relational database to form a basic model of the dispatching knowledge and skill point network.
[0025] Step 2-3: For unstructured text data such as dispatching rules and regulations, technical specifications, management regulations, training specifications, and safe operation instructions, based on their text format and content characteristics, hierarchical construction of their sub-knowledge and skill points and related relationships is carried out on the basic network model of dispatching knowledge and skill points.
[0026] Step 2-4: Use professional knowledge and basic knowledge textbooks such as power system analysis, power electronics technology, electrical engineering basics, high voltage technology basics, and electrical drawing basics to extract rich knowledge points such as basic concepts and principles involved in scheduling knowledge points, and construct the relationship between knowledge points based on theoretical logical relationships to complete the construction of the scheduling knowledge skill point network model.
[0027] On this basis, the nodes and relationships in the knowledge and skill point network model are further clarified. Each knowledge and skill point is a node in the network, which includes node type (knowledge point, skill point), node name, node attribute (knowledge or skill field, such as power grid regulation, power grid abnormality handling, power grid operation, etc.; the occupational level of the node, such as junior worker, intermediate worker, senior worker, technician, senior technician). For example, the "voltage abnormality handling method" node is a knowledge point, the node attribute field is power grid abnormality handling, and the node belongs to the occupational level of senior worker and technician. The "circuit breaker closing operation" node is a skill point, the node attribute field is power grid operation, and the node belongs to the occupational level of junior worker and intermediate worker. To facilitate adaptive learning navigation and recommendation, knowledge and skill points are mainly connected by three types of relationships, including relationships, leading relationships, and related relationships.
[0028] 1) Inclusion relationship: Indicates the subordinate relationship and parent-child relationship between knowledge points, skill points and their child nodes. For example, the "circuit breaker operation" node is a parent node, which contains the child nodes "circuit breaker closing operation" and "circuit breaker opening operation".
[0029] 2) Predecessor relationship: indicates the relationship between the prerequisite knowledge and skill points that must be understood before mastering a certain knowledge and skill point node. For example, the knowledge point "Precautions for handling voltage anomalies" is the predecessor knowledge point of the knowledge point "Methods for handling voltage anomalies".
[0030] 3) Correlation: Indicates that there is a relationship between knowledge and skill points with the same principle or result. For example, the knowledge point "mutual inductor" and the knowledge point "transformer" have the same working principle and are connected by correlation. "Voltage collapse" and "frequency collapse" will both cause a major blackout in the power system or local system, so they are connected by correlation to facilitate dispatchers to expand and associate similar knowledge points.
[0031] Step 3: Preprocess the scheduling knowledge base to generate a scheduling test point network model with key statements attached.
[0032] Among them, the dispatching knowledge base is composed of the dispatching knowledge skill point network model and dispatching procedure text data.
[0033] The specific steps of preprocessing the scheduling knowledge base and generating a scheduling test point network model with key statements are as follows: Step 3-1: The scheduling knowledge and skill point network model is regarded as a list of triples. The preprocessing is to select the knowledge point triples with question relationship to form the test point network model, and delete the unusable triples and skill points.
[0034] In this embodiment, knowledge points with inclusion relationships in the dispatching knowledge and skill point network model and their subordinate "classification" relationship parallel knowledge points or "condition" relationship knowledge points and various aspects of knowledge points contained under the knowledge points (i.e., leaf knowledge points) are retained, and abstract knowledge and skill points that are difficult to quantify by computers are deleted. The leaf knowledge points of the knowledge points are divided into the dispatching test point network model that dispatchers at all levels need to master according to the dispatching occupation standards.
[0035] Step 3-2: Preprocess the dispatching procedure text data and divide the dispatching procedure and other texts into key sentences and key paragraphs according to their semantics. Key sentences are mainly sentences containing facts and knowledge, and key paragraphs are multi-sentence description paragraphs of the reasons, conditions, measures, operating steps, etc. of facts and related knowledge.
[0036] Step 3-3: Perform Jieba word segmentation on key sentences and key paragraph main sentences to separate the hot words.
[0037] Step 3-4: Match the separated hot words with the knowledge points in the scheduling test point network model.
[0038] Step 3-5: Attach key sentences and key paragraphs to knowledge points with high matching degree as extended sentence nodes, and the relationship between them and the knowledge points is "attached".
[0039] Step 3-6: Use expert manual verification technology to check the rationality of the key sentences and key paragraphs.
[0040] Step 3-7: Generate a scheduling test point network model for key statements.
[0041] Step 4: Based on the scheduling test point network model of the attached key sentences, relevant knowledge points and their attached key sentences or key paragraphs are extracted to adaptively generate the test question stem.
[0042] Step 4-1: Adaptive generation of fill-in-the-blank question stems: The types of fill-in-the-blank questions include: blank description type and formula fill-in-the-blank type. This method can effectively evaluate the dispatcher's mastery of specific knowledge points.
[0043] In this embodiment, the specific method of generating a hollowed-out descriptive question stem is: by identifying the key sentences and key paragraphs attached to the knowledge points in the scheduling test point network model, the key matching items (such as knowledge points and key numbers) therein are hollowed out to generate the question stem.
[0044] The specific method of generating formula-filling type questions is: by identifying the key sentences and key paragraphs attached to the knowledge points, the variables or parameters in the formulas involved in the key sentences / key paragraphs are hollowed out to generate the question stem.
[0045] In this embodiment, the difficulty of the fill-in-the-blank question lies in the mining of the blanks. Generally, it is possible to consider analyzing the key sentences. First, Jieba word segmentation is used to identify whether it contains the name information of the key knowledge node or whether it contains key numbers for mining. If it cannot be identified, TF-IDF text mining technology is used to find three key sentences with higher frequency as the mining options.
[0046] The specific implementation steps are as follows: 1) First, given a key knowledge point K, all key sentences attached to it are extracted to form a sentence pool based on the scheduling test point network model.
[0047] 2) Randomly select sentence S1 from the sentence pool as the question option.
[0048] 3) Use the Jieba word segmentation method to segment S1.
[0049] 4) Perform word segmentation matching to determine whether the key knowledge point name matches in S1.
[0050] 5) If the word segmentation match is successful, the knowledge point name is left blank and the process goes to step 9).
[0051] 6) If the word segmentation match fails, determine whether S1 contains the key number.
[0052] 7) If it is included, leave the key number as blank and go to step 9).
[0053] 8) If not included, use the TF-IDF algorithm to perform weight analysis on the separated hot words, and take the three words with the highest weight as blanks.
[0054] 9) Remove the blank in S1 and use it as the question stem, and use the blank as the answer. The fill-in-the-blank question is completed successfully.
[0055] Among them, TF-IDF is a common text mining technology. Its full name is Term Frequency-Inverse Document Frequency (TF-IDF). It is a method for evaluating the importance of keywords in a document, and is usually used in fields such as text classification and information retrieval. TF-IDF represents a document as a vector, where each element of the vector corresponds to a word. The value of the element indicates the importance of the word in the document. The importance is calculated based on two indicators: the frequency of the word in the document (TF) and the inverse of the number of documents in which the word appears in the document set (IDF).
[0056] TF (Term Frequency) indicates how often a term appears in a text. This number is usually normalized (usually the term frequency divided by the total number of words in the article) to prevent it from being biased towards long files (the same term may have a higher term frequency in a long file than in a short file, regardless of whether the term is important or not). TF is expressed as follows: ; in, Indicates the term In the documentation The number of times it appears in For entry In the documentation Frequency of occurrence. However, some common words have no role in the test, and some words with less frequent occurrence may express the theme of the regulation text, so it is not appropriate to use TF alone. The design of weights must meet the following requirements: the stronger the ability of a word to predict the theme, the greater the weight, and vice versa. For example, when doing text statistics, for the entire corpus text, some words only appear in a few articles, then such words have a great effect on the theme of the article, and the weights of these words should be designed to be larger. IDF is doing this kind of work.
[0057] IDF (Inverse Document Frequency) indicates the prevalence of keywords. At this time, a corpus is needed to simulate the language usage environment. If the fewer documents contain a certain term, the larger the IDF, it means that the term has a good ability to distinguish categories. The IDF of a specific word can be obtained by dividing the total number of documents by the number of documents containing the term, and then taking the logarithm of the quotient: ; in, Indicates the number of all texts. Indicates that it contains the term The number of documents.
[0058] TF-IDF refers to the high word frequency in a specific text, and the low text frequency of the word in the entire corpus set, which can produce a high-weighted TF-IDF. Therefore, TF-IDF tends to filter out common words and retain important words. Its expression is as follows: ; Step 4-2: Adaptive generation of true or false question stems: True or false questions are mainly used to examine the dispatcher's understanding and cognition of a certain knowledge point. The question stem directly quotes the factual key sentences / key paragraphs attached to the dispatch test point network model.
[0059] At the same time, in order to enrich the number, complexity and confusion of questions, interference technology is used to replace key numbers, contrasting words, and affirmative and negative words to generate false propositions. They can be mainly divided into the following five types: True or false questions: directly judge the truth of a statement or the conformity of a condition.
[0060] Comparative judgment questions: Compare the characteristics or applicable scenarios of two or more knowledge points.
[0061] Causal judgment questions: evaluate the cause-effect relationship between knowledge points.
[0062] Definition judgment question: Determine whether the statement accurately defines a certain knowledge point.
[0063] Operation judgment questions: examine the correctness of the operation steps.
[0064] Step 4-3: Adaptive generation of multiple-choice question stems: Multiple choice questions: divided into single knowledge point analysis type and cross-knowledge point comprehensive comparison type.
[0065] Step 4-3-1: Single knowledge point analysis type multiple choice questions test the understanding and analysis of a single knowledge point. Common question types can be divided into 6 categories according to the label type of the knowledge point in the scheduling knowledge and skill point network model (cause, condition, function, processing method, application scenario, definition, phenomenon, etc.). The options are the number of sentences / words of the type related to this knowledge point, which can generate single-choice or multiple-choice questions.
[0066] In this embodiment, based on the key paragraphs mounted in the scheduling test point model, the question stem can be generated based on the main sentences in the paragraph, and the options use the remaining conditional statements.
[0067] For example, "the conditions of a certain knowledge point include ()", and the options are the remaining conditional statements, which are directly used as the answers to the multiple-choice questions. For the test point network model structure, the question stem can be generated based on the association relationship. For example, "a certain knowledge point includes ()", and the options are the knowledge point set of the next layer of the knowledge point, forming a multiple-choice question.
[0068] Step 4-3-2: Multiple-choice questions that comprehensively analyze multiple knowledge points examine the analysis between multiple knowledge points. The knowledge points involved in the options have certain commonalities, such as "predecessor", "include", and "related" relationships. The question stem describes the relevant content of a certain knowledge point in the options. Therefore, questions that analyze multiple knowledge points require not only four reasonable options, but also a suitable question stem so that only one option meets the description of the question stem.
[0069] (1) Knowledge extraction: Extract relevant knowledge points and their relationships, key sentences / key paragraphs connected with feature descriptions and condition descriptions from the scheduling test point network model.
[0070] (2) Question stem design: Generate question stems based on the extracted information to ensure that the question stems can accurately reflect the knowledge points being tested and their content.
[0071] (3) Option generation: Design options that include correct answers and distractors. Ensure that each option is related to the knowledge point described in the question stem, but only one option is completely correct. Distractors may be incorrect or partially correct. They can be generated by introducing similar but incorrect knowledge points or features to increase the complexity and confusion of the questions. These options can involve relationships such as "predecessor", "include", and "related" to ensure that the candidates' understanding and ability to discern knowledge points are deeply tested.
[0072] (4) Verification and optimization: Verify the generated questions to ensure the rationality and scientificity of the questions and options, and make optimization adjustments if necessary. Ensure that each option is semantically and logically reasonable and has a certain relevance to the knowledge point described in the question stem. At the same time, avoid designing overly obvious interference options to increase the difficulty and challenge of the questions.
[0073] Through these methods, the present invention can automatically generate rich and diverse assessment questions, thereby improving the comprehensiveness and accuracy of knowledge assessment.
[0074] Step 5: Generate interference items for the assessment questions based on the knowledge point association relationships of the scheduling test point network model.
[0075] When evaluating the quality of questions, the role of distractors is crucial, especially in the design of true / false and multiple-choice questions. The quality of distractors directly affects the difficulty and effectiveness of questions. If the distractors are not enough to confuse dispatchers, they may easily identify the correct answer, which will seriously reduce the quality and reliability of the questions.
[0076] The present invention adopts the knowledge point association relationship based on the dispatch test point network model, and designs a new pseudo-proposition construction method, which is specifically aimed at judgment questions and multiple-choice questions, including key number replacement, contrast word replacement, and affirmative and negative word replacement, aiming to more accurately generate interference items that can effectively confuse dispatchers. Through this strategy based on association relationships, we can improve the quality and educational value of the questions, not only testing the dispatcher's mastery of basic knowledge, but also evaluating his ability to understand the complex relationships between knowledge points. This method not only makes the questions more challenging, but also helps dispatchers to think and analyze more deeply when solving practical problems. As shown in Table 1. Table 1 Examples of pseudo-proposition construction methods
[0077] Step 5-1: Replace key figures Key number replacement is a commonly used question generation technique, especially when the numbers in the sentence are the main components. These main components can include subject-predicate structure, verb-object structure and its modifiers. When performing key number replacement, similar numbers are usually used to generate pseudo propositions, thereby increasing the variety and difficulty of the questions.
[0078] In this embodiment, the specific method is: use the jieba word segmentation tool to segment the sentence and extract the key numbers. In order to ensure the effectiveness and logic of the number replacement, it is necessary to consider whether the content of the question contains any hints about the key numbers and exclude them in advance.
[0079] For example, "There are three basic requirements for electrical main wiring design: reliability, flexibility, and economy." The size of the number can be summarized from the information in the second half of the sentence. It does not interfere with the dispatcher and should be identified and eliminated in advance.
[0080] For the choice of replacement numbers, count the frequencies of key sentences and key paragraph numbers attached to all knowledge points above and below the test points. For ordinary numbers, use the results of ±1, * (-1), *2, / 2 to confuse and replace them. For percentages, add ±5%, ±10%, ±20% to the original numbers, or extract numbers with higher frequency from multiple related knowledge points as replacement items or generate other options for multiple-choice questions.
[0081] However, when it comes to limited numbers such as voltage levels, such as "500kV terminal tower" and the specific qualifier "three-phase", they need to be identified in advance and replaced according to the corresponding voltage level list and phase sequence list. If you want to increase the difficulty, you can also replace it according to the reference voltage and rated voltage correspondence table. When it comes to numbers related to frequency values, the replacement is based on the results of ±0.2 and ±0.5 of the original numbers. Finally, experts manually eliminate some questions with unreasonable restrictions.
[0082] Step 5-2: Replace contrasting words.
[0083] Contrasting word replacement plays an important role in proposition design. When the main components of the sentence in the test point network model contain contrasting words, you can consider using its antonym contrasting words to generate pseudo propositions. In the power industry, you first need to sort out common contrasting word combinations for subsequent contrasting word replacement work in propositions.
[0084] Table 2 shows some common contrastive words and their antonym contrastive words, which form a contrastive relationship with each other. However, when dealing with propositions with multiple contrastive words, simply replacing one contrastive word may sometimes lead to an unreasonable description of the entire sentence, for example, "Although the effectiveness of backup capacity and emergency response mechanisms is as important as the relationship between equipment life and maintenance frequency, the stability and reliability of the power grid may be different under high load and low load conditions." At this time, simply replacing "high load" with "low load" and "stability" with "reliability" will only make the sentence an erroneous and unreasonable statement. In this case, it is recommended to use natural language analysis to simultaneously handle parallel relationships and contrastive word replacement, or to have experts verify it in the end to ensure the logical and semantic accuracy of the proposition. Table 2 Commonly used comparative terms for power grid dispatching
[0085] Step 5-3, positive / negative word replacement: Affirmative-negative word replacement is a commonly used pseudo-proposition construction scheme. The specific approach is that if the sentence is in a negative form, change it to an affirmative form, and vice versa. First, it is necessary to determine the form of the core structure of the proposition through dependency syntactic analysis. When the proposition changes from an affirmative form to a negative form, it is necessary to determine what kind of negative word to add and how to add the negative word. When the proposition changes from a negative form to an affirmative form, it is necessary to delete all adverbs related to the negative word. The construction results of some affirmative and negative sentences are shown in Table 3.
[0086] Table 3 Partial results of affirmative-negative substitution construction
[0087] Step 6: Analyze the initial difficulty setting of the questions from three aspects: question type, knowledge point, and interference strategy. Then dynamically adjust the difficulty of the questions based on the test results of a large number of students.
[0088] When designing the difficulty of questions for the scheduling professional field, the business characteristics are taken into consideration. For newly generated questions, when no students have answered them yet, the difficulty of the questions needs to be estimated based on the attributes of the questions, and then adjusted based on the feedback from the students' usage information.
[0089] 1) Analysis of factors affecting difficulty The factors that influence the difficulty of questions are complex and diverse. This project conducts an initial difficulty setting analysis on questions from three aspects: question type factors, knowledge point factors, and interference strategy factors.
[0090] The question types are mainly divided into fill-in-the-blank questions, true-or-false questions, and multiple-choice questions. Fill-in-the-blank questions require students to have a high degree of knowledge mastery and be able to directly distinguish the exact content of the question; true-or-false questions only require students to judge whether the description is in line with reality; multiple-choice questions require students to distinguish the relevant description of knowledge points or the covered content or the description of different knowledge points according to their question types. The more descriptions that need to be distinguished, the more difficult the question is. Therefore, usually, from the perspective of question type, true-or-false questions are simpler than multiple-choice questions, single-choice questions are simpler than multiple-choice questions, and fill-in-the-blank questions are the most difficult.
[0091] The knowledge point factor mainly considers the difficulty of the types of knowledge points covered by the questions and the number of knowledge points. There are differences in the difficulty of different knowledge points themselves. For some conceptual knowledge points, they do not require too much thinking, reasoning and deduction, and their difficulty is relatively low. The difficulty of questions that only contain such knowledge points is relatively simple. For some cause-type and measure-type knowledge points, they are relatively complex and require thinking, deduction and reasoning to understand the principles. Such knowledge points are relatively difficult, and the difficulty of questions involving such knowledge points is naturally difficult. Therefore, the difficulty of the question is positively correlated with the difficulty of the knowledge points. Secondly, the number of knowledge points covered by the question will also affect the difficulty of the question. The more knowledge points a question covers, the more it needs to complete the identification and judgment of multiple knowledge points, and the higher the difficulty of the question.
[0092] Interference item strategies mainly include key number replacement, contrast word replacement and affirmative and negative word replacement. Generally, contrast word replacement and affirmative and negative word replacement mainly test students' ability to distinguish knowledge descriptions, while key number replacement is more disruptive and requires students to clarify the actual action time, correct frequency range, correct voltage range, or correct voltage level in specific concepts. It may be more difficult than the other two strategies.
[0093] 2) Difficulty Quantification Evaluation Design According to the difficulty influencing factors summarized above, they are divided into different levels according to their different degrees of influence and quantified. This project uses the grade assignment method for quantification. For example, for the influencing factor of question type, "True or false questions", "Single knowledge analysis single choice questions", "Single knowledge analysis multiple choice questions", "Multiple knowledge point analysis single choice questions", and "Fill in the blank questions" are assigned values of 1, 3, 5, 3, and 5 respectively. The quantification of other influencing factors is shown in Table 4.
[0094] Table 4 Quantitative table of influencing factors
[0095] The initial total difficulty of the question can be measured by the weighted average of the difficulty influencing factors. The formula is as follows: ; Among them, d represents the initial difficulty value of the question, d i represents the difficulty value of the i-th difficulty influencing factor, and n represents the total number of difficulty influencing factors.
[0096] Combined with the values assigned in the quantitative table of influencing factors, 1 is easy, 3 is medium, and 5 is difficult. 2 and 4 represent adjacent judgments, which are intermediate values between 1-3 and 3-5, representing medium to easy and medium to difficult. When d is a fraction, consider taking the file set represented by the nearest integer.
[0097] After the initial difficulty assessment of the questions, it is necessary to dynamically modify the difficulty of the questions based on the test results of the students. Considering the use of classic test theory, the answers of all students N to the questions are counted, and the number of correct answers is R. The probability of correct answers is calculated by the formula: ; The difficulty of the questions is updated according to the students' actual learning and answering situations. Assuming that the current difficulty of the question is preset to d, when the probability of the student answering the question correctly is greater than 75%, the difficulty of the question d is lowered by one level to d-1. When d is 1, the difficulty of the question remains unchanged. When the probability of the student answering the question correctly is less than 25%, the difficulty of the question d is raised by one level to d+1. When d is 5, which is the most difficult, the difficulty of the question remains unchanged. However, when the probability of the student answering the question correctly is (25%, 75%), it is considered that the preset difficulty of the question is more reasonable and will not be adjusted for the time being.
[0098] The difficulty of a question is an indicator that measures the difficulty of a question in a standardized test system according to the teaching objectives. It reflects the characteristics of the question itself (such as question type, question content, etc.) and is also a key parameter in the intelligent teaching system. Determining the difficulty level of a question can reasonably recommend adaptive questions to dispatchers with different levels of knowledge and skills. Question difficulty design can effectively support multiple intelligent functions including automatic question generation and personalized question recommendation.
[0099] Embodiment 2 This embodiment discloses a system for generating self-adaptive evaluation questions for power grid dispatching based on a semantic model, including: The semantic model building module is configured to: obtain the scheduling procedure text data; Construct a network model of dispatching knowledge and skills points, and combine it with the dispatching procedure text data to form a dispatching knowledge base; Preprocessing the scheduling knowledge base to generate a scheduling test point network model with key sentences attached; The adaptive test question stem generation module is configured to: extract relevant knowledge points and their attached key sentences or key paragraphs based on the scheduling test point network model of the attached key sentences, and adaptively generate the test question stem; The interference item generation module is configured to: generate interference items for the assessment questions according to the knowledge point association relationship of the scheduling test point network model; The question quality assessment module is configured to: perform an initial difficulty setting analysis on the questions, and then dynamically modify the difficulty of the questions based on the test results of the students.
[0100] Embodiment 3 The purpose of this embodiment is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the program.
[0101] Embodiment 4 The purpose of this embodiment is to provide a computer-readable storage medium.
[0102] A computer-readable storage medium stores a computer program, which executes the steps of the above method when executed by a processor.
[0103] The steps involved in the apparatus of the above embodiment correspond to the method embodiment 1, and the specific implementation method can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.
[0104] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.
[0105] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.
Claims
1. A method for generating adaptive evaluation questions for power grid dispatching based on semantic model, characterized in that: include: Get the scheduling procedure text data; Construct a network model of dispatching knowledge and skills points, and combine it with the dispatching procedure text data to form a dispatching knowledge base; Preprocessing the scheduling knowledge base to generate a scheduling test point network model with key sentences attached; Based on the scheduling test point network model of the key sentences, relevant knowledge points and their key sentences or key paragraphs are extracted to adaptively generate the test question stem; Generate interference items for the assessment questions according to the knowledge point association relationship of the scheduling test point network model; Conduct an initial difficulty setting analysis on the questions, and then dynamically adjust the difficulty of the questions based on the students' test results.
2. A method for generating self-adaptive evaluation questions for power grid dispatching based on semantic model according to claim 1, characterized in that: Preprocessing the dispatching knowledge base is performed, specifically: retaining knowledge points with inclusion relationships in the dispatching knowledge and skill point network model, and their subordinate "classification" relationship parallel knowledge points, or "condition" relationship knowledge points and leaf knowledge points contained under the knowledge points, and deleting abstract knowledge and skill points that are difficult to quantify by computer, and dividing the leaf knowledge points of the knowledge points into the dispatching test point network model that dispatchers at all levels need to master according to the dispatching occupation standard; Divide the scheduling procedure text into key sentences and key paragraphs according to its semantics.
3. The method for generating self-adaptive evaluation questions for power grid dispatching based on semantic model according to claim 1, characterized in that: The specific steps of generating a scheduling test point network model for attaching key sentences are as follows: Preprocess the scheduling knowledge and skill point network model to obtain the scheduling test point network model; Divide the dispatching procedure text into key sentences and key paragraphs according to its semantics; Perform Jieba segmentation on key sentences and key paragraphs to separate the hot words; Match the separated hot words with the knowledge points in the scheduling test point network model; Attach key sentences and key paragraphs to knowledge points with high matching degree as extended sentence nodes; Expert manual verification technology is used to check the rationality of key sentences and key paragraphs; Generate a scheduling test site network model for hooking key statements.
4. The method for generating self-adaptive evaluation questions for power grid dispatching based on semantic model according to claim 1, characterized in that: Based on the scheduling test site network model of the key sentences, Jieba word segmentation is used to analyze the key sentences to identify whether they contain the name information of the key knowledge nodes or whether they contain key numbers for filling in the blanks; or Using TF-IDF text mining technology, we find three key sentences with high frequency as the elimination options.
5. The method for generating self-adaptive evaluation questions for power grid dispatching based on semantic model according to claim 1, characterized in that: The interference items of the assessment questions mainly include key number replacement, contrast word replacement, and affirmative and negative word replacement; Among them, for key number replacement, first use the jieba word segmentation tool to segment the sentence and extract the key numbers; for the selection of replacement numbers, count the frequency of occurrence of key sentences and key paragraph numbers attached to all knowledge points above and below the test knowledge points, and extract numbers with high frequency from multiple related knowledge points as replacement items or generate other options for multiple-choice questions.
6. A method for generating self-adaptive evaluation questions for power grid dispatching based on semantic model according to claim 1, characterized in that: The initial difficulty setting analysis of the questions specifically adopts a grade assignment method to quantify the difficulty influencing factors, and the initial total difficulty of the questions is measured by a weighted average of the difficulty influencing factors.
7. The method for generating self-adaptive evaluation questions for power grid dispatching based on semantic model according to claim 1, characterized in that: The difficulty of the questions will be dynamically modified based on the test results of a large number of students. Consider adopting classic testing theory and updating the difficulty of the questions based on the students' actual learning and answering situation.
8. A power grid dispatch adaptive evaluation question generation system based on semantic model, characterized in that: include: The semantic model building module is configured to: obtain the scheduling procedure text data; Construct a network model of dispatching knowledge and skills points, and combine it with the dispatching procedure text data to form a dispatching knowledge base; Preprocessing the scheduling knowledge base to generate a scheduling test point network model with key sentences attached; The adaptive test question stem generation module is configured to: extract relevant knowledge points and their attached key sentences or key paragraphs based on the scheduling test point network model of the attached key sentences, and adaptively generate the test question stem; The interference item generation module is configured to: generate interference items for the assessment questions according to the knowledge point association relationship of the scheduling test point network model; The question quality assessment module is configured to: perform an initial difficulty setting analysis on the questions, and then dynamically modify the difficulty of the questions based on the test results of the students.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method described in any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 7 are performed.
Citation Information
Patent Citations
Ontology-based power grid dispatching regulation stipulated knowledge modeling method and system
CN110489562A
Question and answer method and system based on knowledge graph, electronic equipment and medium
CN112395403A
Chinese choice question interference item generation method based on free text
CN112686025A
Intelligent alarm processing method and device for power monitoring system based on knowledge base
CN113516565A
Automatic test paper generation and examination method and device, electronic equipment and medium
CN115730038A
Cited By
Large language model capability assessment method and system based on dynamic data assessment
CN119988914A
A large language model capability evaluation method and system based on dynamic data evaluation
CN119988914B
AI-based large model evaluation question bank automatic generation method and system
CN120179810A
An AI-based method and system for automatically generating a large model evaluation question bank
CN120179810B
Synthetic method for cognitive ability test data of mine large model industry
CN120596664A
