Methods for identifying problem-solving ideas in knowledge domain questions
By constructing a knowledge graph and similarity algorithm, identifying problem-solving ideas in the field of elementary mathematics, the problem of poor recognition effect in the existing technology is solved, and a fast and accurate recognition effect is achieved.
Patent Information
- Application Number
- CN202311462972.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-06
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2043-11-06
AI Technical Summary
In the prior art, the method of identifying the problem-solving ideas of the question requires a lot of manpower or relying on high-cost labeled data, resulting in poor identification results. Especially when there are fewer data samples, it is difficult to quickly and accurately identify the question-setting intentions in the question and Knowledge corresponding to the intention to set the question.
By establishing a knowledge graph in the knowledge field, using Chinese-named entity recognition and prime indexing technology to extract knowledge entities and relationship knowledge, construct a relationship node matrix, perform labeling and connection judgment, and combine similarity algorithm to identify problem-solving ideas.
It realizes the rapid and accurate identification of problem-solving ideas in the field of knowledge, reduces labor costs and data dependence, and improves the recognition effect.
Smart Images

Figure BDA0004532625610000011 
Figure BDA0004532625610000025 
Figure BDA0004532625610000026
Abstract
Description
Technical Field
[0001] The invention relates to a method for identifying a solution idea of a knowledge domain question, and belongs to the technical field of intelligent education. Background Art
[0002] Knowledge points in various fields of human knowledge (recognized theorems, inferences, properties, axioms, laws or formulas) are mostly presented in the form of questions to test students. Whether the learners can correctly answer the questions is used to judge whether the students have mastered the knowledge points corresponding to the questions. The knowledge points corresponding to these questions represent the solution ideas of the questions. Mastering these knowledge points can answer these questions. Therefore, mastering the solution ideas of the questions becomes the key to answering the questions. In the prior art, there are two ways to identify the solution ideas of questions. One is to manually identify the question text and formula test points to analyze the questions, but it is too labor-intensive and subjective; the other is to analyze the questions through machine learning and deep learning models, but it needs to rely on a large amount of question data with labeled meanings, and the data cost is high. If the data sample is small, the recognition effect will be poor. Summary of the invention
[0003] The technical problem to be solved by the present invention is: how to quickly and accurately identify the question intention in the question and the knowledge corresponding to the question intention.
[0004] The technical solution proposed by the present invention is: a method for identifying solution ideas for knowledge domain questions, comprising the following steps:
[0005] Step 1: Select the knowledge domain that needs to be identified for problem-solving ideas and establish a knowledge graph for the knowledge domain. The specific process of establishing the knowledge graph is as follows:
[0006] Step 1.1: All knowledge entities in the knowledge domain are extracted through the Chinese named entity recognition method to form a knowledge entity set A, as shown in the following formula (1):
[0007] A={A 1 , A 2 , ..., A n} (1),
[0008] In formula (1), A 1 is the first knowledge entity extracted from the knowledge domain, A 2 is the second knowledge entity extracted from the knowledge domain, A n The last knowledge entity extracted from the knowledge domain;
[0009] All entities in the knowledge entity set A are prime-indexed using different prime numbers p in the extraction order to form a prime-indexed knowledge entity set A.p , as shown in the following formula (2),
[0010]
[0011] In formula (2), is a prime number p 1 the first knowledge entity extracted from the knowledge domain to be indexed, is a prime number p 2 a second knowledge entity extracted from the knowledge domain to be indexed, is a prime number p n The last knowledge entity extracted from the knowledge domain to be indexed;
[0012] Step 1.2: Extract all the recognized theorems, inferences, properties, axioms, laws or formulas in the knowledge domain as relational knowledge to form a relational knowledge set B, as shown in the following formula (3):
[0013] B={B 1 , B 2 , ..., B m} (3),
[0014] In formula (3), B 1 is the first relational knowledge extracted from the knowledge domain, B 2 is the second relation extracted from the knowledge domain, B m The last relational knowledge extracted from the knowledge domain;
[0015] The first relational knowledge in the relational knowledge set B is divided into entities by Jieba word segmenter, and the results of the division are collected to form the first relational knowledge entity set. As shown in the following formula (4),
[0016]
[0017] In formula (4), a 1 、a 2 , ..., a q are all natural numbers greater than or equal to 1 and less than or equal to n, and a 1 ≠a 2 ≠...≠a q ;
[0018] Then, all the remaining relational knowledge in the relational knowledge set B are divided into entities in turn, and the results of the division are collected in turn to form the corresponding relational knowledge entity set, and all the relational knowledge entity sets are collected to form the relational knowledge entity set matrix B A , as shown in the following formula (5),
[0019]
[0020] Step 1.3: The first relational knowledge B in relational knowledge set B 1 and the first relational knowledge entity set All the relational knowledge entities in the knowledge domain are extracted to form the first relational node C for constructing the knowledge graph of the knowledge domain 1 , Then follow the first relationship node C 1 The construction method constructs the relationship nodes corresponding to the remaining relationship knowledge in the relationship knowledge set B in turn, and collects all the relationship nodes to form a relationship node matrix C, as shown in the following formula (6):
[0021]
[0022] In formula (6), C 2 The second relational knowledge B 2 And the second relation knowledge entity set All the relational knowledge entities in the knowledge domain are extracted to form a second relational node for constructing the knowledge graph of the knowledge domain; C n is the last relational knowledge B n and the final relation knowledge entity set All the relational knowledge entities in the knowledge domain are extracted to form the last relational node for constructing the knowledge graph of the knowledge domain;
[0023] The prime number indexes of all the relational knowledge entities contained in all the relational nodes are multiplied in turn to form the label of the relational node, and all the labeled relational nodes are collected to form the labeled relational node matrix C q , as shown in the following formula (6),
[0024]
[0025] In formula (6), is the first labeled relation node with a label; q 1 It is the C 1 The mark and is the C 1 The result of multiplying the prime indexes of all relational knowledge entities in ; is the second labeled relation node with a label; q 2 It is the C 2 The mark and is the C 2 The result of multiplying the prime indexes of all relational knowledge entities in ; is the last labeled relation node with a tag; q n It is the Cn The mark and is the C n The result of multiplying the prime indexes of all relational knowledge entities in ;
[0026] Step 1.4: The relationship node matrix C q All the labeled relationship nodes in the matrix are connected in pairs to see whether there is an intersection between the relationship knowledge entity sets contained in the two labeled relationship nodes to be judged. If there is an intersection, the two labeled relationship nodes are connected; if there is no intersection, it means that there is no connection between the two labeled nodes and no connection is required; when the relationship node matrix C q After all the labeled relationship nodes in the are connected and judged in pairs in turn, the knowledge graph of the knowledge domain is constructed;
[0027] Step 2: Extract entities from the questions that require problem-solving ideas through the Chinese named entity recognition method, and collect all the extracted recognized knowledge entities to form a recognized knowledge entity set D, as shown in the following formula (7):
[0028] D={D 1 , D 2 ,...,D n} (7),
[0029] In formula (7), D 1 is the first identified knowledge entity extracted from the knowledge domain, D 2 is the second identified knowledge entity extracted from the knowledge domain, D n the last identified knowledge entity extracted from the knowledge domain;
[0030] Match all the knowledge entities in the identified knowledge entity set D with all the knowledge entities in the knowledge entity set A. After the matching is completed, multiple matching pairs of identified knowledge entities and knowledge entities are formed. The prime indexes of the knowledge entities in the matching pairs are matched and marked to the identified knowledge entities in the matching pairs at the same time. Repeat the above steps to match and mark all the identified knowledge entities in all matching pairs that have not been marked with prime indexes to form prime index identified knowledge entities. Collect all the prime index identified knowledge entities that have been marked with matching to form a prime index identified knowledge entity set. As shown in the following formula (8),
[0031]
[0032] In formula (8), is the first prime index identifying the knowledge entity after the matching mark, is the second prime number index identifying the knowledge entity after the matching mark, is the last prime number index identifying the knowledge entity after the matching mark is performed;
[0033] Step 3: Use prime indexes to identify knowledge entity sets Sequentially and relational knowledge entity set matrix B A The similarity between the two is calculated by the set of all relational knowledge entities in the equation (9):
[0034]
[0035] In formula (9), i is a natural number greater than or equal to 1 and less than or equal to n; is the number of elements in the intersection between the i-th relational knowledge entity set and the prime index identification knowledge entity set; is the number of elements in the i-th relational knowledge entity set; is the number of elements of the set of knowledge entities identified by prime index;
[0036] The calculated similarities are arranged from high to low, and the relational knowledge of the relational knowledge entity set corresponding to the similarities before X in the sorting is output, where X is an empirical value, and the value range is a natural number between 1 and 7. The output relational knowledge is used as a knowledge point to represent the solution idea of the problem that needs to be identified.
[0037] Beneficial effects of the present invention: In a knowledge field that requires identification of problem-solving ideas, the present invention establishes a special knowledge graph by using the recognized theorems, inferences, properties, axioms, laws or formulas in the knowledge field as relationship nodes, and then uses the knowledge graph to identify problem-solving ideas for questions. Through a similarity algorithm, the similarity between the question and the relationship knowledge in the knowledge graph is obtained, and the relationship between the relationship knowledge and the question is obtained through the similarity, thereby obtaining the problem-solving ideas for the question. Specific implementation methods
[0038] The following is a further description of a method for identifying problem-solving ideas for a knowledge domain question in the present invention in conjunction with a specific embodiment.
[0039] The present invention provides a method for identifying a solution idea of a knowledge domain question, comprising the following steps:
[0040] Step 1: Select the knowledge domain that needs to be identified for problem-solving ideas and establish a knowledge graph for the knowledge domain. The specific process of establishing the knowledge graph is as follows:
[0041] Step 1.1: Extract all knowledge entities in the knowledge domain through the Chinese named entity recognition method to form a knowledge entity set A, as shown in the following formula (1):
[0042] A={A 1 , A2 , ..., A n} (1),
[0043] In formula (1), A 1 is the first knowledge entity extracted from the knowledge domain, A 2 is the second knowledge entity extracted from the knowledge domain, A n The last knowledge entity extracted from the knowledge domain;
[0044] All entities in the knowledge entity set A are prime-indexed using different prime numbers p in the extraction order to form a prime-indexed knowledge entity set A. p , as shown in the following formula (2),
[0045]
[0046] In formula (2), is a prime number p 1 The first knowledge entity extracted from the knowledge domain to be indexed, is a prime number p 2 The second knowledge entity extracted from the knowledge domain to be indexed, is a prime number p n The last knowledge entity extracted from the knowledge domain to be indexed;
[0047] Step 1.2: Extract all the recognized theorems, inferences, properties, axioms, laws or formulas in the knowledge domain as relational knowledge to form a relational knowledge set B, as shown in the following formula (3):
[0048] B={B 1 , B 2 , ..., B m} (3),
[0049] In formula (3), B 1 is the first relational knowledge extracted from the knowledge domain, B 2 is the second relation extracted from the knowledge domain, B m The last relational knowledge extracted from the knowledge domain;
[0050] The first relational knowledge in relational knowledge set B is divided into entities through Jieba word segmenter and the results of the division are collected to form the first relational knowledge entity set. As shown in the following formula (4),
[0051]
[0052] In formula (4), a 1 、a 2 , ..., aq are all natural numbers greater than or equal to 1 and less than or equal to n, and a 1 ≠a 2 ≠...≠a q ;
[0053] Then all the remaining relational knowledge in the relational knowledge set B are divided into entities in turn, and the results of the division are collected in turn to form the corresponding relational knowledge entity set, and all the relational knowledge entity sets are collected to form the relational knowledge entity set matrix B A , as shown in the following formula (5),
[0054]
[0055] Step 1.3: The first relational knowledge B in relational knowledge set B 1 and the first relation knowledge entity set All the relational knowledge entities in the knowledge domain are extracted to form the first relational node for constructing the knowledge graph of the knowledge domain Then follow the first relationship node C 1 The construction method constructs the relationship nodes corresponding to the remaining relationship knowledge in the relationship knowledge set B in turn, and collects all the relationship nodes to form a relationship node matrix C, as shown in the following formula (6):
[0056]
[0057] In formula (6), C 2 It is the second relational knowledge B 2 and the second relation knowledge entity set All the relational knowledge entities in the knowledge domain are extracted to form the second relational node for constructing the knowledge graph of the knowledge domain; C n It is the last relation knowledge B n And the final relation knowledge entity set All the relational knowledge entities in the knowledge domain are extracted to form the last relational node for constructing the knowledge graph of the knowledge domain;
[0058] The prime indexes of all relational knowledge entities contained in all relational nodes are multiplied in turn to form the label of the relational node, and all the labeled relational nodes are collected to form the labeled relational node matrix C. q , as shown in the following formula (6),
[0059]
[0060] In formula (6), is the first labeled relation node with a label; q 1 It is C 1 The sign is C 1The result of multiplying the prime indexes of all relational knowledge entities in ; is the second labeled relation node with a label; q 2 It is C 2 The sign is C 2 The result of multiplying the prime indexes of all relational knowledge entities in ; is the last labeled relation node with a tag; q n It is C n The sign is C n The result of multiplying the prime indexes of all relational knowledge entities in ;
[0061] Step 1.4: Transform the relationship node matrix C q All the labeled relationship nodes in the matrix are connected in pairs to see whether there is an intersection between the relationship knowledge entity sets contained in the two labeled relationship nodes to be judged. If there is an intersection, the two labeled relationship nodes are connected; if there is no intersection, it means that there is no connection between the two labeled nodes and they do not need to be connected; when the relationship node matrix C q After all the labeled relationship nodes in the knowledge domain are connected and judged in pairs, the knowledge graph of the knowledge domain is constructed;
[0062] Step 2: Extract entities from the questions that require problem-solving ideas through the Chinese named entity recognition method, and collect all the extracted recognized knowledge entities to form a recognized knowledge entity set D, as shown in the following formula (7):
[0063] D={D 1 , D 2 ,...,D n} (7),
[0064] In formula (7), D 1 is the first identified knowledge entity extracted from the knowledge domain, D 2 It is the second identified knowledge entity extracted from the knowledge domain, D n The last identified knowledge entity extracted by the knowledge domain;
[0065] Match all knowledge entities in the identification knowledge entity set D with all knowledge entities in the knowledge entity set A. After the matching is completed, multiple matching pairs of identification knowledge entities and knowledge entities are formed. The prime indexes of the knowledge entities in the matching pairs are matched and marked to the identification knowledge entities in the matching pairs at the same time. Repeat the above steps to match and mark all identification knowledge entities in all matching pairs that have not been marked with prime indexes to form prime index identification knowledge entities. Collect all the prime index identification knowledge entities that have been marked with matching to form a prime index identification knowledge entity set. As shown in the following formula (8),
[0066]
[0067] In formula (8), It is the first prime index identifying the knowledge entity after matching the mark. It is the second prime index after matching mark to identify knowledge entity. It is the last prime index after matching mark to identify the knowledge entity;
[0068] Step 3: Use prime indexes to identify knowledge entity sets Sequentially and relational knowledge entity set matrix B A The similarity between the two is calculated by the set of all relational knowledge entities in the equation (9):
[0069]
[0070] In formula (9), i is a natural number greater than or equal to 1 and less than or equal to n; is the number of elements in the intersection between the i-th relational knowledge entity set and the prime index identification knowledge entity set; is the number of elements in the i-th relational knowledge entity set; is the number of elements of the set of knowledge entities identified by prime index;
[0071] The calculated similarities are arranged from high to low, and the relational knowledge of the relational knowledge entity set corresponding to the X similarities before the sorting is output. X is an empirical value, and the value range is a natural number between 1 and 7. The output relational knowledge is used as the knowledge point as the solution idea for the problem that needs to be identified.
[0072] The following is an identification of problem-solving ideas for questions in the field of elementary mathematics knowledge:
[0073] First, a knowledge graph of elementary mathematics knowledge is established. Then, the problem △ABC, where ∠A=20, ∠B=70, find ∠C. is processed through step 2. The extracted recognition knowledge entity set includes triangle, internal angle and equation. All the extracted recognition knowledge entities are matched with all the knowledge entities in the knowledge graph, and all the recognition knowledge entities are marked for matching. The prime index recognition knowledge entity set after matching marking includes triangle 11 , interior angle 13 and equation 17 , the similarity between the prime index identification knowledge entity set and all the relational knowledge entity sets in the knowledge graph is calculated by formula (9). Some calculation results are shown in Table 1 below.
[0074]
[0075] Table 1
[0076] Based on experience, the relational knowledge represented by the first relational knowledge entity set is selected as the solution to the problem.
Claims
1. A method for identifying solution ideas for knowledge domain questions, characterized by: The following steps are involved: Step 1: Select the knowledge domain that needs to be identified for problem-solving ideas and establish a knowledge graph for the knowledge domain. The specific process of establishing the knowledge graph is as follows: Step 1.1: All knowledge entities in the knowledge domain are extracted through the Chinese named entity recognition method to form a knowledge entity set A, as shown in the following formula (1): A={A1, A2,...,An}(1), In formula (1), A1 is the first knowledge entity extracted from the knowledge domain, A2 is the second knowledge entity extracted from the knowledge domain, and A n The last knowledge entity extracted from the knowledge domain; All entities in the knowledge entity set A are marked with different prime numbers p in the extraction order to complete the prime number indexing, forming a prime number index knowledge entity set A p , as shown in the following formula (2), In formula (2), is the first knowledge entity extracted from the knowledge domain indexed by the prime number p1, is the second knowledge entity extracted from the knowledge domain indexed by prime number p2, is a prime number p n The last knowledge entity extracted from the knowledge domain to be indexed; Step 1.2: Extract all the recognized theorems, inferences, properties, axioms, laws or formulas in the knowledge domain as relational knowledge to form a relational knowledge set B, as shown in the following formula (3): B={B1, B2,...,Bn}(3), In formula (3), B1 is the first relational knowledge extracted from the knowledge domain, B2 is the second relational knowledge extracted from the knowledge domain, and B n It is the last relational knowledge extracted from the knowledge domain; The first relational knowledge in the relational knowledge set B is divided into entities by using the Jieba word segmenter, and the results of the division are collected to form the first relational knowledge entity set. As shown in the following formula (4), In formula (4), a1, a2, ..., a q are all natural numbers greater than or equal to 1 and less than or equal to n, and a1≠a2≠...≠a q ; Then, all the remaining relational knowledge in the relational knowledge set B are divided into entities in turn, and the results of the division are collected in turn to form the corresponding relational knowledge entity set, and all the relational knowledge entity sets are collected to form the relational knowledge entity set matrix B A , as shown in the following formula (5), Step 1.3: Combine the first relational knowledge B1 in the relational knowledge set B and the first relational knowledge entity set All the relational knowledge entities in the knowledge domain are extracted to form the first relational node C1 for constructing the knowledge graph of the knowledge domain. Then, according to the construction method of the first relationship node C1, the relationship nodes corresponding to the remaining relationship knowledge in the relationship knowledge set B are constructed in sequence, and all the relationship nodes are collected to form a relationship node matrix C, as shown in the following formula (6): In formula (6), C2 is the second relational knowledge B2 and the second relational knowledge entity set All the relational knowledge entities in the knowledge domain are extracted to form a second relational node for constructing the knowledge graph of the knowledge domain; C n is the last relational knowledge B n and the last relational knowledge entity set All the relational knowledge entities in the knowledge domain are extracted to form the last relational node for constructing the knowledge graph of the knowledge domain; The prime number indexes of all the relational knowledge entities contained in all the relational nodes are multiplied in turn to form the label of the relational node, and all the labeled relational nodes are collected to form the labeled relational node matrix C q , as shown in the following formula (7), In formula (7), is the first labeled relation node with a label; q1 is the label of C1 and is the result of multiplying the prime indexes of all relational knowledge entities in C1; is a second labeled relation node with a label; q2 is the label of C2 and is the result of multiplying the prime indexes of all relational knowledge entities in C2; is the last labeled relation node with a tag; q n It is the C n The mark and is the C n The result of multiplying the prime indexes of all relational knowledge entities in ; Step 1.4: The labeled relationship node matrix C q All the labeled relationship nodes in the are connected in pairs one by one to see whether there is an intersection between the relationship knowledge entity sets contained in the two labeled relationship nodes being judged. If there is an intersection, the two labeled relationship nodes are connected; If there is no intersection, it means that there is no connection between the two marked nodes and no connection is required; when the marked relationship node matrix C q After all the labeled relationship nodes in the are connected and judged in pairs in turn, the knowledge graph of the knowledge domain is constructed; Step 2: Extract entities from the questions that require problem-solving ideas through the Chinese named entity recognition method, and collect all the extracted recognized knowledge entities to form a recognized knowledge entity set D, as shown in the following formula (8): D={D1, D2,...,Dn}(8), In formula (7), D1 is the first identified knowledge entity extracted from the knowledge domain, D2 is the second identified knowledge entity extracted from the knowledge domain, and D n the last identified knowledge entity extracted from the knowledge domain; Match all the knowledge entities in the identified knowledge entity set D with all the knowledge entities in the knowledge entity set A. After the matching is completed, multiple matching pairs of identified knowledge entities and knowledge entities are formed. The prime indexes of the knowledge entities in the matching pairs are matched and marked to the identified knowledge entities in the matching pairs at the same time. Repeat the above steps to match and mark all the identified knowledge entities in all matching pairs that have not been marked with prime indexes to form prime index identified knowledge entities. Collect all the prime index identified knowledge entities that have been marked with matching to form a prime index identified knowledge entity set. As shown in the following formula (9), In formula (9), is the first prime index identifying the knowledge entity after the matching mark, is the second prime number index identifying the knowledge entity after the matching mark, is the last prime number index identifying the knowledge entity after the matching mark is performed; Step 3: Use prime indexes to identify knowledge entity sets Sequentially and relational knowledge entity set matrix B A The similarity between the two is calculated by the set of all relational knowledge entities in the equation (10). In formula (10), i is a natural number greater than or equal to 1 and less than or equal to n; is the number of elements in the intersection between the i-th relational knowledge entity set and the prime index identification knowledge entity set; is the number of elements in the i-th relational knowledge entity set; is the number of elements of the set of knowledge entities identified by prime index; The calculated similarities are arranged from high to low, and the relational knowledge of the relational knowledge entity set corresponding to the similarities before X in the sorting is output, where X is an empirical value, and the value range is a natural number between 1 and 7. The output relational knowledge is used as a knowledge point to represent the solution idea of the problem that needs to be identified.
