A question bank updating method, a storage medium and a device

Abstract

This invention discloses a question bank update method, storage medium, and device. By constructing a two-layer semantic graph of "question-solution," it achieves structured expression and quantitative analysis of differences among multiple solutions, effectively solving the problem of duplicate storage of question stem information caused by the diversity of solutions in the question bank. This invention introduces a solution correctness verification model based on graph neural networks, which can automatically evaluate the logical rigor of new solutions, significantly reducing the risk of dirty data being added to the database. This invention combines students' actual learning behavior data on educational tablets to establish a cognitive adaptation mapping relationship for solutions, thereby enabling personalized recommendations of optimal solution paths based on learning progress. This invention not only improves the accuracy and richness of the question bank content but also enhances its support for students' innovative thinking, promoting the evolution of intelligent education systems from "answer-oriented" to "process-oriented education," and has significant teaching application value and technological advancement.

Description

Technical Field

[0001] This invention relates to the field of intelligent education technology, specifically to a question bank update method, storage medium, and device. Background Technology

[0002] With the deepening development of K-12 education informatization, smart terminals such as educational tablets have been widely used in classroom teaching and students' self-directed learning. Question banks, as core teaching resources, play an important role in knowledge consolidation, ability assessment, and thinking training.

[0003] However, existing question bank systems mostly adopt a static "one question, one solution" management model, meaning each question is associated with only one standard solution path, ignoring the common phenomenon of multiple solutions in subjects like mathematics and physics. Furthermore, due to the lack of fine-grained modeling of solution structures, the same question stem is repeatedly entered due to different solutions, leading to severe redundancy in the question bank. On the other hand, innovative solutions proposed by students during the answering process are difficult to collect and verify effectively, while incorrect or logically flawed solutions easily pollute the question bank data. In addition, traditional methods rely on manual review of solution correctness, which is inefficient and has limited coverage, failing to meet the needs of dynamic updates. Therefore, there is an urgent need for a technical solution that can automatically identify, evaluate, and structurally manage multiple solutions to improve the knowledge completeness and pedagogical adaptability of the question bank. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention aims to provide a question bank updating method, storage medium, and device.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A question bank update method includes the following steps: S1. Obtain the question stem description information and corresponding candidate parsing paths for the candidate questions; S2. Based on the question description information, match whether there is a semantically equivalent benchmark question in the target question bank; S3. If there is a successfully matched benchmark question in the target question bank, then further determine whether the candidate parsing path obtained in step S1 belongs to any standard solution category already included in the benchmark question. S4. If the candidate parsing path obtained in step S1 does not belong to any of the standard solution categories already included in the benchmark problem, then calculate the solution difference degree between the candidate parsing path and each of the solution categories already included in the benchmark problem, and determine whether it is a potential innovative solution based on the solution difference degree; if it is a potential innovative solution, input the candidate parsing path into the solution correctness verification model to obtain the solution confidence score; S5. When the solution confidence score obtained by the solution correctness verification model is greater than or equal to the preset correctness threshold, the candidate parsing path is stored in the target question bank as a new solution category associated with the benchmark question, and a corresponding solution label is generated for the candidate parsing path; the solution label includes solution type, applicable student cognitive level and recommended use scenario.

[0006] Furthermore, the specific process of step S2 is as follows: S2.1 Perform structured analysis on the problem description information of the candidate solution path, and extract the mathematical objects, physical quantities, logical relationships and constraints involved; S2.2 Construct a semantic vector representation of the question stem description information of the candidate solution path, and use a cross-modal encoder with an attention mechanism to map the question stem description information of the candidate solution path into a low-dimensional dense vector; S2.3. Traverse the question stem description information of all existing questions in the target question bank, and calculate the similarity score between the semantic vector representation of the question stem description information of the existing questions and the semantic vector representation of the question stem description information of the candidate questions. S2.4 When the similarity score of the semantic vector representation of the question stem description information of an existing question and a candidate question is greater than the first similarity threshold and the key elements are completely identical, the question is determined to be a semantically equivalent question; the key elements include the core concepts, quantitative relationships and solution objectives involved in the question.

[0007] Furthermore, in step S4, the calculation process for the solution difference between the candidate parsing path and the various solution categories already included in the benchmark problem is as follows: A1. Represent the candidate analytical paths and the solution categories of the benchmark questions as ordered sequences S={s1,s2,...,sn} consisting of several reasoning steps; A2. For each reasoning step si, extract the triple (oi,pi,ci) consisting of the operation type oi, the dependent premise pi, and the output conclusion ci. A3. Define the structural distance Dstruct(A,B) between the candidate parsing path A and the solution category B of the benchmark problem as:

[0008] Where SA and SB are the sets of reasoning steps of A and B, respectively; M is the subset of the maximum co-occurrence steps between A and B found by the optimal matching algorithm. If the cosine similarity of the semantic embeddings of the triples of the two reasoning steps exceeds the second preset threshold, then the two reasoning steps are determined to be a match. A4. If the Dstruct(A,B) of the candidate parsing path is greater than δ for each of the included solution categories of the benchmark problem, where δ is the third preset threshold, then the candidate parsing path is considered to have a significant structural difference from each of the solution categories of the benchmark problem and belongs to a potential innovative solution.

[0009] Furthermore, in step S4, the training process of the solution correctness verification model is as follows: B1. Collect the complete parsing path and its manual annotation results from historical student answer data. The manual annotation results include correct, incorrect, and partially correct. B2. Construct positive and negative sample pairs, where positive samples are correct solutions confirmed by the teacher and negative samples are typical error patterns; B3. The inference chain evaluation module based on graph neural network is adopted to model the solution as a directed inference graph G=(V,E), where the node v∈V represents the intermediate conclusion of each step and the edge e∈E represents the derivation relationship between steps. B4. By examining the directed reasoning graph through a graph neural network, the context information is aggregated using the message passing mechanism to predict the rationality score of each step in the parsing path, and the overall confidence score of the solution is obtained. The output range of the confidence score is [0,1], and the closer it is to 1, the more reliable the solution is.

[0010] Furthermore, in step S5, the process of generating labels applicable to students' cognitive levels is as follows: A cognitive adaptation mapping table for the new solution is established to record the difficulty of understanding the solution for different learning groups. The cognitive adaptation mapping table is constructed by analyzing indicators of students' interactive behavior on the educational tablet, including the completion rate of students watching the video explaining the solution, the success rate of trying to imitate the solution to answer questions, and the frequency of actively collecting or marking "don't understand". The applicable cognitive level of the solution is divided into groups based on the indicators of students' interactive behavior on the educational tablet.

[0011] Furthermore, in step S3, if there is a successfully matched benchmark question in the target question bank, and the candidate parsing path obtained in step S1 belongs to one of the standard solution categories already included in the benchmark question, then the addition of the candidate question is rejected, and only the access statistics of the benchmark question are updated; if there is no successfully matched benchmark question in the target question bank, then the candidate question and the corresponding candidate parsing path are associated and stored in the target question bank, and a corresponding solution tag is generated for the candidate parsing path.

[0012] Furthermore, in step S1, the candidate parsing paths for candidate questions originate from at least one of the following channels: (1) The handwritten answers submitted by students through K12 educational tablets and the structured expressions after OCR recognition and semantic restoration; (2) The extended solutions manually added by teachers during the grading process; (3) The alternative solutions automatically generated by AI models; All candidate parsing paths from all sources must be uniformly formatted and logically segmented to be transformed into comparable standardized reasoning sequences.

[0013] The present invention also provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-described method.

[0014] The present invention also provides a computer device, including a processor and a memory, wherein the memory is used to store a computer program; and the processor is used to execute the computer program to implement the above-described method.

[0015] The beneficial effects of this invention are as follows: 1. This invention constructs a two-layer semantic graph of "question-solution" to realize the structured expression and differential quantification analysis of multiple solutions, effectively solving the problem of duplicate storage of question stem information caused by the diversity of solutions in the question bank; 2. This invention introduces a solution correctness verification model based on graph neural networks, which can automatically evaluate the logical rigor of new solutions and significantly reduce the risk of dirty data being stored in the database. 3. This invention combines students' actual learning behavior data on educational tablets to establish a cognitive adaptation mapping relationship for solutions, thereby enabling personalized recommendations of the optimal solution path based on students' learning progress. 4. This invention not only improves the accuracy and richness of the question bank content, but also enhances its ability to support students' innovative thinking, promoting the evolution of intelligent education systems from "answer-oriented" to "process-oriented education," and has significant teaching application value and technological advancement. Detailed Implementation

[0016] The present invention will be further described below. It should be noted that this embodiment is based on the present technical solution and provides detailed implementation methods and specific operation processes, but the protection scope of the present invention is not limited to this embodiment.

[0017] This embodiment provides a question bank update method, including the following steps: S1. Obtain the question stem description information and corresponding candidate parsing paths for the candidate questions; S2. Based on the question description information, match whether there is a semantically equivalent benchmark question in the target question bank; S3. If there is a successfully matched benchmark question in the target question bank, then further determine whether the candidate parsing path obtained in step S1 belongs to any standard solution category already included in the benchmark question. S4. If the candidate parsing path obtained in step S1 does not belong to any of the standard solution categories already included in the benchmark problem, then calculate the solution difference degree between the candidate parsing path and each of the solution categories already included in the benchmark problem, and determine whether it is a potential innovative solution based on the solution difference degree; if it is a potential innovative solution, input the candidate parsing path into the solution correctness verification model to obtain the solution confidence score; S5. When the solution confidence score obtained by the solution correctness verification model is greater than or equal to the preset correctness threshold, the candidate parsing path is stored in the target question bank as a new solution category associated with the benchmark question, and a corresponding solution label is generated for the candidate parsing path; the solution label includes solution type, applicable student cognitive level and recommended use scenario.

[0018] In this embodiment, the specific process of step S2 is as follows: S2.1 Perform structured analysis on the problem description information of the candidate solution path, and extract the mathematical objects, physical quantities, logical relationships and constraints involved; S2.2 Construct a semantic vector representation of the question stem description information of the candidate solution path, and use a cross-modal encoder with an attention mechanism to map the question stem description information of the candidate solution path into a low-dimensional dense vector; S2.3. Traverse the question stem description information of all existing questions in the target question bank, and calculate the similarity score between the semantic vector representation of the question stem description information of the existing questions and the semantic vector representation of the question stem description information of the candidate questions. S2.4 When the similarity score of the semantic vector representation of the question stem description information of an existing question and a candidate question is greater than the first similarity threshold and the key elements are completely identical, the question is determined to be a semantically equivalent question; the key elements include the core concepts, quantitative relationships and solution objectives involved in the question.

[0019] Step S2 is used to determine whether the candidate questions are essentially the same as the questions already in the target question bank (i.e., "semantic equivalence"). First, the question description is broken down into mathematical objects (such as numbers and graphs), physical quantities (such as speed and mass), logical relationships (such as causal relationships and comparative relationships), and constraints (such as the range of values). For example, the question description "Given triangle ABC, AB=5, BC=12, find the length of AC?" will be parsed as: Mathematical objects: triangle, side lengths AB, BC, AC; Physical quantities: lengths (5, 12); Logical relationships: Pythagorean theorem (if it is a right triangle); Constraints: no special restrictions.

[0020] Then, a cross-modal encoder with an attention-integrated mechanism is used to convert the question stem description text into a string of numbers (vectors), which facilitates the computer's comparison of similarity. For example, if the vectors of question stem description A and question stem description B are very similar, it means that they may be questions of the same type.

[0021] Subsequently, the target question bank is traversed, and the similarity score between the semantic vector representations of the question stem descriptions of candidate questions and existing questions is calculated. The matching condition is that the similarity score exceeds the first similarity threshold (e.g., 90% match) and the key elements are completely consistent, including core concepts (e.g., both are "Pythagorean theorem" problems), quantitative relationships (e.g., both involve the lengths of three sides), and solution objectives (e.g., both are for finding the hypotenuse).

[0022] In this embodiment, the calculation process for the solution difference degree between the candidate parsing path and the various solution categories already included in the baseline problem in step S4 is as follows: A1. Represent the candidate analytical paths and the solution categories of the benchmark questions as ordered sequences S={s1,s2,...,sn} consisting of several reasoning steps; A2. For each reasoning step si, extract the triple (oi,pi,ci) consisting of the operation type oi, the dependent premise pi, and the output conclusion ci. A3. Define the structural distance Dstruct(A,B) between the candidate parsing path A and the solution category B of the benchmark problem as:

[0023] Where SA and SB are the sets of reasoning steps of A and B, respectively; M is the subset of the maximum co-occurrence steps between A and B found by the optimal matching algorithm (such as the Hungarian algorithm). If the cosine similarity of the semantic embeddings of the triples of the two reasoning steps exceeds the second preset threshold, then the two reasoning steps are determined to be matched. A4. If the Dstruct(A,B) of the candidate parsing path is greater than δ for each of the included solution categories of the benchmark problem, where δ is the third preset threshold, then the candidate parsing path is considered to have a significant structural difference from each of the solution categories of the benchmark problem and belongs to a potential innovative solution.

[0024] The solution degree difference calculation is used to quantify the degree of difference between the candidate parsing path and the various solution categories already included in the benchmark problem, and to determine whether the candidate parsing path of the candidate problem belongs to an innovative solution.

[0025] `Dstruct(A,B)` represents the structural difference between solution A and solution B, with a value ranging from 0 to 1. The closer to 1, the greater the difference. For example, solution A: {drawing a graph → applying the Pythagorean theorem → calculating} (|SA|=3), solution B: {drawing auxiliary lines → applying trigonometric functions → calculating} (|SB|=3). `M` is the largest subset of matching steps common to both solutions. For example, if only the "calculation" step is successfully matched, then |M|=1.

[0026] This embodiment further explains the calculation of the solution difference in step S4 by distance.

[0027] Problem: Find the length of the hypotenuse of a right triangle. Solution A (General Solution): Draw a right triangle (Operation: drawing; Prerequisite: two legs are known; Conclusion: graphical visualization) Write down the Pythagorean theorem formula (Operation: Apply the formula, Prerequisite: Correct graph, Conclusion: a² + b² = c²) Substitute the values ​​into the formula (Operation: Arithmetic operation, Prerequisite: Formula exists, Conclusion: c=13) Solution B (candidate solution): Construct similar triangles by drawing auxiliary lines (Operation: drawing; Prerequisite: known angles; Conclusion: new shape). List the trigonometric function relationships (operation: formula application, premise: similarity ratio holds, conclusion: sinθ = opposite side / hypotenuse) Substitute the values ​​into the formula (Operation: Arithmetic operation, Prerequisite: Formula exists, Conclusion: c=13) Calculation process: Only the "substitution calculation" step is a perfect match (|M|=1). Dstruct(A,B)=1-2×1 / (3+3)≈0.67 If δ = 0.6, then solution B is considered an innovative solution; if δ = 0.7, then it is not considered an innovative solution.

[0028] In this embodiment, the training process of the solution correctness verification model in step S4 is as follows: B1. Collect the complete parsing path and its manual annotation results (correct / incorrect / partially correct) from historical student answer data. Specifically, complete problem-solving steps can be collected from students' past assignments, along with the teacher's feedback (correct / incorrect / partially correct).

[0029] B2. Construct positive and negative sample pairs. Positive samples are the correct solutions confirmed by the teacher, and negative samples are typical error patterns (such as omission of units, misuse of formulas, and circular reasoning). B3. The inference chain evaluation module based on graph neural network is adopted to model the solution as a directed inference graph G=(V,E), where the node v∈V represents the intermediate conclusion of each step and the edge e∈E represents the derivation relationship between steps. B4. By examining the directed reasoning graph through a graph neural network, the context information is aggregated using the message passing mechanism to predict the rationality score of each step in the parsing path, and the overall confidence score of the solution is obtained. The output range of the confidence score is [0,1], and the closer it is to 1, the more reliable the solution is.

[0030] Specifically, the graph neural network first simulates "message passing," allowing adjacent steps to verify each other's rationality (e.g., whether the conclusion of the previous step supports the next step), then scores each step (0-1 points), and finally combines all steps to obtain the overall credibility. For example: 0.9 points: The solution is rigorous and likely correct. 0.3 points: There are obvious logical loopholes or errors.

[0031] The following are further examples: For the problem "Calculate the hypotenuse of a right triangle (with legs of 5 and 12)," one student's solution included the following steps: Draw a diagram and label the side lengths (Reasonable, score 0.9).

[0032] Write the Pythagorean theorem "a² + b² = c²" (Reasonable, score 1.0).

[0033] The square was omitted during the calculation; 5+12=17 was written directly (error, score 0.1).

[0034] The final output score is 0.67 (this may be partially correct and requires manual review in step 3).

[0035] The solution correctness verification model is an AI model. After training, it can automatically determine the correctness of candidate solution paths (e.g., in math or physics problems). By analyzing the rationality of each reasoning step, it ultimately provides a score (0-1, with higher scores indicating greater reliability) for the overall solution.

[0036] In this embodiment, the process of generating labels applicable to students' cognitive levels in step S5 is as follows: A cognitive adaptation mapping table for the new solution method is established to record the difficulty of understanding the solution method for different learning groups. This cognitive adaptation mapping table is constructed by analyzing students' interactive behaviors on the educational tablet, including the completion rate of watching the video explanation of the solution method, the success rate of attempting to imitate the solution method in answering questions, and the frequency of actively saving or marking "don't understand". Based on the above interactive behavior indicators, the applicable cognitive levels of students for this solution method are clustered, generating labels such as "suitable for basic level" and "suitable for thinking expansion class".

[0037] Through the above process, each solution category for each question in the target question bank can be tagged with its student-friendliness. By analyzing actual student usage data (such as video completion rate, imitation success rate, etc.), a solution can be automatically marked as suitable for "basic students" or "advanced students". When encountering similar students in the future, solutions that match their abilities can be recommended first. For example, "basic students" can be recommended solutions with detailed steps that are easy to imitate, while "advanced students" can be recommended solutions that are concise but require reasoning, thereby achieving personalized recommendations.

[0038] In this embodiment, in step S3, if there is a successfully matched benchmark question in the target question bank, and the candidate parsing path obtained in step S1 belongs to one of the standard solution categories already included in the benchmark question, then the addition of the candidate question is rejected, and only the access statistics of the benchmark question are updated; if there is no successfully matched benchmark question in the target question bank, then the candidate question and the corresponding candidate parsing path are associated and stored in the target question bank, and a corresponding solution tag is generated for the candidate parsing path.

[0039] In this embodiment, in step S1, the candidate parsing paths for candidate questions come from at least one of the following channels: (1) The structured expression of the handwritten answers submitted by students through K12 educational tablets after OCR recognition and semantic restoration; (2) The extended solutions manually added by teachers during the grading process; (3) Alternative solutions automatically generated by AI models (such as those generated by a combination of large language model and symbolic reasoning engine). All candidate parsing paths from all sources must undergo unified formatting and logical sentence segmentation to be transformed into comparable standardized reasoning sequences.

[0040] For those skilled in the art, various corresponding changes and modifications can be made based on the above technical solutions and concepts, and all such changes and modifications should be included within the protection scope of the claims of this invention.

Claims

1. A question bank updating method, characterized in that, Includes the following steps: S1. Obtain the question stem description information and corresponding candidate parsing paths for the candidate questions; S2. Based on the question description information, match whether there is a semantically equivalent benchmark question in the target question bank; S3. If there is a successfully matched benchmark question in the target question bank, then further determine whether the candidate parsing path obtained in step S1 belongs to any standard solution category already included in the benchmark question. S4. If the candidate parsing path obtained in step S1 does not belong to any of the standard solution categories already included in the benchmark problem, then calculate the solution difference degree between the candidate parsing path and each of the solution categories already included in the benchmark problem, and determine whether it is a potential innovative solution based on the solution difference degree. If it is a potential innovative solution, the candidate parsing path is input into the solution correctness verification model to obtain a solution confidence score; S5. When the solution confidence score obtained by the solution correctness verification model is greater than or equal to the preset correctness threshold, the candidate parsing path is stored in the target question bank as a new solution category associated with the benchmark question, and a corresponding solution label is generated for the candidate parsing path; the solution label includes solution type, applicable student cognitive level and recommended use scenario.

2. The method according to claim 1, characterized in that, The specific process of step S2 is as follows: S2.1 Perform structured analysis on the problem description information of the candidate solution path, and extract the mathematical objects, physical quantities, logical relationships and constraints involved; S2.2 Construct a semantic vector representation of the question stem description information of the candidate solution path, and use a cross-modal encoder with an attention mechanism to map the question stem description information of the candidate solution path into a low-dimensional dense vector; S2.

3. Traverse the question stem description information of all existing questions in the target question bank, and calculate the similarity score between the semantic vector representation of the question stem description information of the existing questions and the semantic vector representation of the question stem description information of the candidate questions. S2.4 When the similarity score of the semantic vector representation of the question stem description information of an existing question and a candidate question is greater than the first similarity threshold and the key elements are completely identical, the question is determined to be a semantically equivalent question; the key elements include the core concepts, quantitative relationships and solution objectives involved in the question.

3. The method according to claim 1, characterized in that, In step S4, the calculation process for the solution difference between the candidate parsing path and the various solution categories already included in the benchmark problem is as follows: A1. Represent the candidate analytical paths and the solution categories of the benchmark questions as ordered sequences S={s1,s2,...,sn} consisting of several reasoning steps; A2. For each reasoning step si, extract the triple (oi,pi,ci) consisting of the operation type oi, the dependent premise pi, and the output conclusion ci. A3. Define the structural distance Dstruct(A,B) between the candidate parsing path A and the solution category B of the benchmark problem as: Where SA and SB are the sets of reasoning steps of A and B, respectively; M is the subset of the maximum co-occurrence steps between A and B found by the optimal matching algorithm. If the cosine similarity of the semantic embeddings of the triples of the two reasoning steps exceeds the second preset threshold, then the two reasoning steps are determined to be a match. A4. If the Dstruct(A,B) of the candidate parsing path is greater than δ for each of the included solution categories of the benchmark problem, where δ is the third preset threshold, then the candidate parsing path is considered to have a significant structural difference from each of the solution categories of the benchmark problem and belongs to a potential innovative solution.

4. The method according to claim 1, characterized in that, In step S4, the training process of the solution correctness verification model is as follows: B1. Collect the complete parsing path and its manual annotation results from historical student answer data. The manual annotation results include correct, incorrect, and partially correct. B2. Construct positive and negative sample pairs, where positive samples are correct solutions confirmed by the teacher and negative samples are typical error patterns; B3. The inference chain evaluation module based on graph neural network is adopted to model the solution as a directed inference graph G=(V,E), where the node v∈V represents the intermediate conclusion of each step and the edge e∈E represents the derivation relationship between steps. B4. By examining the directed reasoning graph through a graph neural network, the context information is aggregated using the message passing mechanism to predict the rationality score of each step in the parsing path, and the overall confidence score of the solution is obtained. The output range of the confidence score is [0,1], and the closer it is to 1, the more reliable the solution is.

5. The method according to claim 1, characterized in that, In step S5, the process of generating labels applicable to students' cognitive levels is as follows: A cognitive adaptation mapping table for the new solution is established to record the difficulty of understanding the solution for different learning groups. The cognitive adaptation mapping table is constructed by analyzing indicators of students' interactive behavior on the educational tablet, including the completion rate of students watching the video explaining the solution, the success rate of trying to imitate the solution to answer questions, and the frequency of actively collecting or marking "don't understand". The applicable cognitive level of the solution is divided into groups based on the indicators of students' interactive behavior on the educational tablet.

6. The method according to claim 1, characterized in that, In step S3, if there is a successfully matched benchmark question in the target question bank, and the candidate parsing path obtained in step S1 belongs to one of the standard solution categories already included in the benchmark question, then the addition of the candidate question is rejected, and only the access statistics of the benchmark question are updated. If there is no matching baseline question in the target question bank, the candidate question and the corresponding candidate parsing path are associated and stored in the target question bank, and a corresponding solution tag is generated for the candidate parsing path.

7. The method according to claim 1, characterized in that, In step S1, the candidate parsing paths for candidate questions come from at least one of the following channels: (1) The handwritten answers submitted by students through K12 educational tablets and the structured expressions after OCR recognition and semantic restoration; (2) The extended solutions manually added by teachers during the grading process; (3) The alternative solutions automatically generated by AI models; All candidate parsing paths from all sources must be uniformly formatted and logically segmented to be transformed into comparable standardized reasoning sequences.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method described in any one of claims 1-7.

9. A computer device, characterized in that, It includes a processor and a memory, the memory being used to store a computer program; the processor being used to execute the computer program to implement the method of any one of claims 1-7.

Images