A knowledge retrieval method for engineering equipment test cases based on semantic similarity

Abstract

The present invention relates to the technical field of engineering equipment test, and specifically discloses a method for engineering equipment test case knowledge retrieval based on semantic similarity, comprising constructing an engineering equipment test case ontology, performing knowledge retrieval analysis and calculation based on domain ontology semantic similarity, performing a primary retrieval case push, capturing keywords in the pushed engineering equipment test case document, performing secondary knowledge retrieval based on keyword synonym expansion, and performing secondary retrieval case push; by performing knowledge retrieval analysis and calculation based on domain ontology semantic similarity, the explicit and implicit experience knowledge in the engineering equipment test design field is standardizedly represented and retrieved and reused, and engineering equipment test case knowledge retrieval is automatically realized, the retrieval efficiency is low and the retrieval difficulty is low, and the engineering equipment test case knowledge retrieval combining military and ontology is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering equipment testing, and more specifically, to a method for retrieving engineering equipment testing case knowledge based on semantic similarity. Background Art

[0002] Engineering equipment test is a systematic activity to test the combat effectiveness of equipment and various applicability indicators. It can optimize equipment design and facilitate the improvement and modification of equipment. Before conducting engineering equipment test, it is necessary to conduct knowledge retrieval on the outline and design scheme of engineering equipment test. The assessment design scheme and the design scheme involve multidisciplinary knowledge with the characteristics of cross-domain knowledge, multi-professional coverage and multi-level knowledge superposition. The knowledge is highly intensive, and the construction of its evaluation index system is difficult. The design of assessment subjects is challenging. The current knowledge retrieval of engineering equipment test cases relies on the experience accumulation of professionals, supplemented by manuals, books, etc., lacks the standardized representation and retrieval reuse of explicit and implicit experience knowledge in the field of engineering equipment test design, and its retrieval efficiency is low and the retrieval difficulty is great. Ontology is a formal knowledge description established by abstracting concepts, terms and relationships in a specific field. It can realize the standardized representation of knowledge in the field and is widely used in biomedicine, engineering and transportation. However, in the existing knowledge retrieval, there is no knowledge retrieval method for engineering equipment test cases that combines military and ontology. In order to solve the above problems, a technical solution is provided. Summary of the invention

[0003] In order to overcome the above-mentioned defects of the prior art, the present invention provides a knowledge retrieval method for engineering equipment test cases based on semantic similarity. By performing knowledge retrieval analysis and calculation based on domain ontology semantic similarity, the explicit and implicit experience knowledge in the field of engineering equipment test design is standardizedly represented and retrieved for reuse, and the engineering equipment test case knowledge retrieval is automatically realized. The retrieval efficiency is low and the retrieval difficulty is low, thereby realizing the engineering equipment test case knowledge retrieval combining military and ontology.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A method for engineering equipment test case knowledge retrieval based on semantic similarity includes the following steps:

[0006] Step 1: Construct the engineering equipment test case ontology;

[0007] Step 2: perform knowledge retrieval analysis and calculation based on the domain ontology semantic similarity, sort the domain ontology semantic similarity in descending order, and obtain the top 5 cases in terms of domain ontology semantic similarity;

[0008] Step 3: Push a search case;

[0009] Step 4: Capture keywords in the pushed engineering equipment test case documents, perform secondary knowledge retrieval based on keyword synonym expansion, and push secondary retrieval cases;

[0010] In step 2, the knowledge retrieval analysis calculation includes ontology concept semantic distance similarity calculation, case attribute semantic similarity calculation and domain ontology semantic similarity calculation. The ontology concept semantic distance similarity calculation and case attribute semantic similarity calculation are used to obtain the ontology concept semantic distance similarity and case attribute semantic similarity respectively. The domain ontology semantic similarity is obtained by weighting the ontology concept semantic distance similarity and the case attribute semantic similarity. The domain ontology semantic similarity is used to evaluate the similarity between the target case and the candidate case. The formula for domain ontology semantic similarity is:

[0011] S=ω1S g +ω2S p ;

[0012] Where: S is the semantic similarity of domain ontology, S g is the semantic distance similarity of the ontology concepts, S p is the semantic similarity of case attributes, ω1 and ω2 are the weights of the semantic distance similarity of ontology concepts and the semantic similarity of case attributes, respectively.

[0013] As a further solution of the present invention, in step 1, the process of constructing the engineering equipment test case ontology includes:

[0014] Step A1, determine the professional field and scope of the knowledge ontology: based on the engineering equipment test, select the object of its assessment of operation efficiency indicators as engineering equipment, and determine the constructed domain ontology as the engineering equipment type;

[0015] Step A2, reuse existing knowledge ontology: domain experts use the public knowledge base of the engineering corps professional field to obtain historical equipment test plans and collect concept terminology sets in the field of equipment test assessment;

[0016] Step A3, define the hierarchical database system between concepts, determine the attributes of concepts, define the constraints of attributes, and create instances: use the collected equipment test and assessment field concept terminology set, use database technology to divide the levels according to the types and models of engineering equipment, build a hierarchical database system between concepts, obtain the attribute set of each concept, and the attribute set of each concept is {case category, geographical environment, type of equipment, equipment usage scenario, functional component}. The attribute value types of the engineering equipment test design ontology are all entities. According to the constructed ontology model, select a specific class, create an instance of the class, and clarify the attribute value.

[0017] As a further solution of the present invention, in step A3, the case categories include outlines and plans, the geographical environments include plateaus, plains and coastal areas, the types of equipped troops include engineered chemical brigades and combined brigades, and the equipment use scenarios include surface, land and amphibious.

[0018] As a further solution of the present invention, in step 2, in the ontology concept semantic distance similarity calculation, the target case and the historical case ontology concepts in the case library are numbered, and the ontology concept semantic distance similarity is twice the difference between the ontology concept depth and 1, minus the Euclidean distance of the shortest path of the concept value of the target case and the corresponding numbered concept value of the historical case in the case library in the ontology structure hierarchy, and then the ratio of the difference between the ontology concept depth and 1 twice. The formula for the ontology concept semantic distance similarity is:

[0019]

[0020] Where: Dept(Onto) is the depth of the ontology concept, i is the number of edges in the shortest path between the concept node and the root node, and are the target case numbered i and the concepts of the historical case ontology in the case library, For concept and The shortest Euclidean distance in the ontology hierarchy.

[0021] As a further solution of the present invention, in the semantic distance similarity formula of ontology concepts, the semantic path distance and node depth are comprehensively considered, and the depth of the ontology concept is the concept and Concept The sum of the node depths, which is the same as 1 concept The ratio of the sum of the absolute values ​​of the node depth differences. The depth formula of the ontology concept is:

[0022]

[0023] Where: Concept Node depth, concept The node depth of .

[0024] As a further solution of the present invention, in the formula of domain ontology semantic similarity, the weight of ontology concept semantic distance similarity and the weight of case attribute semantic similarity are obtained as follows:

[0025] Step C1, collecting weighted influence data of semantic distance similarity of ontology concepts and weighted influence data of semantic similarity of case attributes: the weighted influence data of semantic distance similarity of ontology concepts includes the average number of edges of ontology distance root nodes, the domain expert requirement setting value, and the application requirement score of ontology concepts; the weighted influence data of semantic similarity of case attributes includes the case attribute importance score, the attribute application scenario score, the user requirement feedback value, and the business rule constraint value;

[0026] Step C2, obtaining the weighted influence coefficient of the semantic distance similarity of the ontology concept and the weighted influence coefficient of the semantic similarity of the case attribute: obtaining the weighted influence coefficient of the semantic distance similarity of the ontology concept through the average number of edges of the ontology distance root node, the domain expert requirement setting value and the application requirement score of the ontology concept, and obtaining the weighted influence coefficient of the semantic similarity of the case attribute through the case attribute importance score, the attribute application scenario score, the user requirement feedback value and the business rule constraint value, wherein the weighted influence coefficient of the semantic distance similarity of the ontology concept is negatively correlated with the average number of edges of the ontology distance root node, positively correlated with the domain expert requirement setting value, and positively correlated with the application requirement score of the ontology concept. The formula of the weighted influence coefficient of the semantic distance similarity of the ontology concept is:

[0027]

[0028] Where: α1 is the weighted influence coefficient of the semantic distance similarity of the ontology concept, is the average number of edges from the entity to the root node, t expert Set the value for domain expert requirements, d x q is the application requirement score of the ontology concept, θ, μ, and ρ are all empirical adjustment coefficients with values ​​between [1,10], obtained based on the numerical analysis of historical data;

[0029] The weighted influence coefficient of case attribute semantic similarity is positively correlated with the case attribute importance score, the attribute application scenario score, the user demand feedback value, and negatively correlated with the business rule constraint value. The formula for the weighted influence coefficient of case attribute semantic similarity is:

[0030]

[0031] Where: α2 is the weight influence coefficient of the semantic similarity of case attributes, d r Score the importance of case attributes, z fk Score the attribute application scenario, Y yw Provide feedback value for user needs. σ and v are both empirical adjustment coefficients with values ​​between [1, 2], obtained based on numerical analysis of historical data;

[0032] Step C3, obtain the semantic distance similarity weight of the ontology concept and the semantic similarity weight of the case attribute: the semantic distance similarity weight of the ontology concept is the ratio of the semantic distance similarity weight influence coefficient of the ontology concept to the semantic similarity weight influence coefficient of the case attribute multiplied by 0.5. The formula of the semantic distance similarity weight of the ontology concept is: The case attribute semantic similarity weight is 1 minus the ontology concept semantic distance similarity weight, and the case attribute semantic similarity weight is ω2=1-ω1.

[0033] As a further solution of the present invention, in step 2, in the case attribute semantic similarity calculation, the weight of each attribute is multiplied by the sum of the similarities of the corresponding numbered attributes of the target case and the historical case as the case attribute semantic similarity. The formula for the case attribute semantic similarity is:

[0034]

[0035] Where: j is the attribute number, n is the total number of attributes in the case, γ j is the weight of the attribute numbered j, where are the j-th attributes of the target case and the historical case, respectively. is the similarity between the target case and the jth attribute of the historical case.

[0036] As a further solution of the present invention, in the formula of the semantic similarity of case attributes, the similarity of the j-th attribute of the target case and the historical case is the ratio of the case attribute depth minus 1 times, the difference between the shortest path Euclidean distance of the j-th attribute of the target case and the historical case in the ontology structure hierarchy, and the case attribute depth minus 1 times, and the similarity formula of the j-th attribute of the target case and the historical case is:

[0037]

[0038] Where: Dept(Onto)′ is the case attribute depth, which is the number of edges contained in the shortest path between the case attribute node and the root node. is the jth attribute of the target case and the jth attribute of the historical case The shortest path Euclidean distance in the ontology hierarchy.

[0039] As a further solution of the present invention, in the similarity formula of the target case and the j-th attribute of the historical case, the case attribute path distance and the node depth are comprehensively considered, and the case attribute depth is the j-th attribute of the target case. and the jth attribute of the historical case The sum of the node depths of the target case j is the same as 1. The jth attribute of the historical case The ratio of the sum of the absolute values ​​of the differences, the case attribute depth formula is:

[0040]

[0041] Where: are the jth attribute of the target case and the jth attribute of the historical case The node depth of .

[0042] As a further solution of the present invention, in step 4, keywords in the pushed engineering equipment test case document are captured, and secondary knowledge retrieval is performed based on keyword synonym expansion. The process of pushing secondary retrieval cases includes:

[0043] Step B1, generating an extended synonym set of keywords in the engineering equipment test text file by using an engineering equipment test design domain synonym library that takes into account both general knowledge and professional domain knowledge;

[0044] Step B2, perform synonym expansion query based on Lucence Chinese search engine to retrieve cases in a push;

[0045] Step B3, reorder the cases in the first push, output the secondary search results, and push the secondary search cases.

[0046] The technical effects and advantages of the engineering equipment test case knowledge retrieval method based on semantic similarity of the present invention are as follows: the present invention performs knowledge retrieval analysis and calculation based on the semantic similarity of domain ontology, standardizes the representation and retrieval reuse of explicit and implicit experience knowledge in the field of engineering equipment test design, and automatically realizes engineering equipment test case knowledge retrieval with low retrieval efficiency and low retrieval difficulty, realizes engineering equipment test case knowledge retrieval combining military and ontology, and can improve the efficiency and scientificity of test design in the field of engineering equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 Construct a flow chart for the engineering equipment test case ontology;

[0048] Figure 2 Determine the concept attribute diagram for the hierarchical relationship between concepts in the test;

[0049] Figure 3 It is the hierarchical distribution diagram of case attributes;

[0050] Figure 4 Push process diagram for case retrieval;

[0051] Figure 5 It is a flow chart of the retrieval method based on keyword synonym expansion;

[0052] Figure 6This paper presents the workflow of knowledge retrieval method for engineering equipment test cases. DETAILED DESCRIPTION

[0053] The following will be combined with the accompanying drawings in the present invention to clearly and completely describe the technical solution in the present invention. Obviously, the described technical solution is only a part of the present invention but not all of it. Based on the technical solution in the present invention, all other technical solutions obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0054] A method for engineering equipment test case knowledge retrieval based on semantic similarity includes the following steps:

[0055] Step 1: Construct the engineering equipment test case ontology;

[0056] Step 2: perform knowledge retrieval analysis and calculation based on the domain ontology semantic similarity, sort the domain ontology semantic similarity in descending order, and obtain the top 5 cases in terms of domain ontology semantic similarity;

[0057] Step 3: Push a search case;

[0058] Step 4: Capture keywords in the pushed engineering equipment test case documents, perform secondary knowledge retrieval based on keyword synonym expansion, and push secondary retrieval cases;

[0059] In step 2, the knowledge retrieval analysis calculation includes ontology concept semantic distance similarity calculation, case attribute semantic similarity calculation and domain ontology semantic similarity calculation. The ontology concept semantic distance similarity calculation and case attribute semantic similarity calculation are used to obtain the ontology concept semantic distance similarity and case attribute semantic similarity respectively. The domain ontology semantic similarity is obtained by weighting the ontology concept semantic distance similarity and the case attribute semantic similarity. The domain ontology semantic similarity is used to evaluate the similarity between the target case and the candidate case. The formula for domain ontology semantic similarity is:

[0060] S=ω1S g +ω2S p ;

[0061] Where: S is the semantic similarity of domain ontology, S g is the semantic distance similarity of the ontology concepts, S p is the semantic similarity of case attributes, ω1 and ω2 are the weights of the semantic distance similarity of ontology concepts and the semantic similarity of case attributes, respectively.

[0062] The present invention performs knowledge retrieval, analysis and calculation based on the semantic similarity of domain ontology, standardizes the representation and retrieval of explicit and implicit empirical knowledge in the field of engineering equipment test design, and automatically realizes knowledge retrieval of engineering equipment test cases with low retrieval efficiency and difficulty. It realizes knowledge retrieval of engineering equipment test cases combining military and ontology, and can improve the efficiency and scientificity of test design in the field of engineering equipment.

[0063] like Figure 1 As shown, in step 1, the process of constructing the engineering equipment test case ontology includes:

[0064] Step A1, determine the professional field and scope of the knowledge ontology: based on the engineering equipment test, select the object of its assessment of operation efficiency indicators as engineering equipment, and determine the constructed domain ontology as the engineering equipment type;

[0065] Step A2, reuse existing knowledge ontology: domain experts use the public knowledge base of the engineering corps professional field to obtain historical equipment test plans and collect concept terminology sets in the field of equipment test assessment;

[0066] Step A3, define the hierarchical database system between concepts, determine the attributes of concepts, define the constraints of attributes, and create instances: use the collected equipment test and assessment field concept terminology set, use database technology to divide the levels according to the types and models of engineering equipment, build a hierarchical database system between concepts, obtain the attribute set of each concept, and the attribute set of each concept is {case category, geographical environment, type of equipment, equipment usage scenario, functional component}. The attribute value types of the engineering equipment test design ontology are all entities. According to the constructed ontology model, select a specific class, create an instance of the class, and clarify the attribute value.

[0067] Engineering equipment test design is an activity that focuses on the performance and other indicators of many different types of engineering equipment under different conditions. The design subject is engineering equipment. Therefore, this paper starts from the assessment object and determines the construction of the domain ontology as the type of engineering equipment. The premise of ontology construction is the sorting and analysis of domain knowledge, and the abstraction and organization of important concepts. Engineering equipment belongs to the engineering corps specialty. This paper sorts out the important concept terminology set in the field of equipment test assessment by investigating relevant materials in the engineering corps specialty field, analyzing historical test plans and outlines, and consulting experts in the field, as shown in Table 1.

[0068] Table 1 List of concepts and terms of engineering equipment test ontology

[0069]

[0070] Since the engineering equipment test is carried out around different types of engineering equipment, the present invention divides the levels according to the types and models of engineering equipment, such as Figure 2 shown.

[0071] The attributes of a concept represent its special properties that are different from other concepts. The present invention sorts out the attribute set according to the characteristics of the test assessment object: {case category, geographical environment, type of equipment, usage scenario, functional component}.

[0072] Case category: Since the engineering equipment test design is divided into test outline and plan, the case category is "outline" or "plan".

[0073] Geographical environment: Different geographical environments will result in different structural characteristics of equipment. Equipment in different geographical environments needs to undergo targeted environmental adaptability tests. According to equipment test requirements, the geographical environment of the test is divided into three categories: plateau, plain, and coastal.

[0074] Type of troops equipped: Engineering equipment is mainly equipped in engineering and chemical brigades and combined brigades, and different types of troops have different requirements for the performance of equipment. Light combined brigades require equipment with good mobility. Therefore, the test and assessment design needs to take into account the type of troops equipped with the equipment.

[0075] Equipment usage scenarios: Different usage scenarios lead to differences in equipment structure and function, including "surface, land, and amphibious".

[0076] Functional components: Different engineering equipment has different functions, and the structural components that need to be evaluated are also different. For structural components with similar functions, even if the equipment types are different, the focus of the component evaluation is often similar, which can be used as a reference. Mine sweepers and impact bridges are two different types of equipment, but when evaluating the mobility of the equipment, both equipment belong to the crawler chassis structure, which can be used as a reference.

[0077] Creating an instance is to select a specific class based on the constructed ontology model, create an instance of the class, and clarify the attribute value. Case A is represented in the ontology as follows: the equipment model is ground blasting equipment, minesweeping equipment, X1 model, the case category is outline, the geographical environment is plateau and plains, the use scenario is land, the type of equipment is medium synthetic brigade, and the functional structure is wheeled.

[0078] It should be noted that in step A3, case categories include outlines and plans, geographical environments include plateaus, plains and coastal areas, types of equipped forces include engineered chemical brigades and combined brigades, and equipment use scenarios include surface, land and amphibious.

[0079] It should be noted that in step 2, in the calculation of the semantic distance similarity of ontology concepts, the ontology concepts of the target case and the historical cases in the case library are numbered. The ontology concept semantic distance similarity is twice the difference between the ontology concept depth and 1, minus the Euclidean distance of the shortest path of the concept value of the target case and the corresponding numbered concept value of the historical case in the case library in the ontology structure hierarchy, and then the ratio of the difference between the ontology concept depth and 1 twice. The formula for the semantic distance similarity of ontology concepts is:

[0080]

[0081] Where: Dept(Onto) is the depth of the ontology concept, i is the number of edges in the shortest path between the concept node and the root node, and are the target case numbered i and the concepts of the historical case ontology in the case library, For concept and The shortest Euclidean distance in the ontology hierarchy.

[0082] It should be noted that in the semantic distance similarity formula of ontology concepts, the semantic path distance and node depth are comprehensively considered, and the depth of the ontology concept is the concept and Concept The sum of the node depths, which is the same as 1 concept The ratio of the sum of the absolute values ​​of the node depth differences. The depth formula of the ontology concept is:

[0083]

[0084] Where: Concept Node depth, concept The node depth of .

[0085] Path distance The smaller it is, the greater the node depth Dept(Onto), and the similarity value S g To facilitate calculation, the present invention adopts a simplified processing method and sets the semantic distance between adjacent concept nodes as unit distance 1. Figure 2 Taking the equipment models X1 and Xn under the self-propelled pontoon bridge in the river crossing equipment as an example, the Euclidean distance between the two is the geometric distance between their corresponding nodes in the network structure under the subclass "self-propelled pontoon bridge" of "river crossing equipment". When the semantic distance is the same, the greater the depth of the two concept nodes, the greater the similarity between the concepts.

[0086] It should be noted that in the formula of domain ontology semantic similarity, the weights of ontology concept semantic distance similarity and case attribute semantic similarity are obtained as follows:

[0087] Step C1, collecting weighted influence data of semantic distance similarity of ontology concepts and weighted influence data of semantic similarity of case attributes: the weighted influence data of semantic distance similarity of ontology concepts includes the average number of edges of ontology distance root nodes, the domain expert requirement setting value, and the application requirement score of ontology concepts; the weighted influence data of semantic similarity of case attributes includes the case attribute importance score, the attribute application scenario score, the user requirement feedback value, and the business rule constraint value;

[0088] Step C2, obtaining the weighted influence coefficient of the semantic distance similarity of the ontology concept and the weighted influence coefficient of the semantic similarity of the case attribute: obtaining the weighted influence coefficient of the semantic distance similarity of the ontology concept through the average number of edges of the ontology distance root node, the domain expert requirement setting value and the application requirement score of the ontology concept, and obtaining the weighted influence coefficient of the semantic similarity of the case attribute through the case attribute importance score, the attribute application scenario score, the user requirement feedback value and the business rule constraint value, wherein the weighted influence coefficient of the semantic distance similarity of the ontology concept is negatively correlated with the average number of edges of the ontology distance root node, positively correlated with the domain expert requirement setting value, and positively correlated with the application requirement score of the ontology concept. The formula of the weighted influence coefficient of the semantic distance similarity of the ontology concept is:

[0089]

[0090] Where: α1 is the weighted influence coefficient of the semantic distance similarity of the ontology concept, is the average number of edges from the entity to the root node, t expert Set the value for domain expert requirements, d xq Score the application requirements of the ontology concept. θ, μ, and ρ are all empirical adjustment coefficients with values ​​between [1,10], obtained based on the numerical analysis of historical data;

[0091] The weighted influence coefficient of case attribute semantic similarity is positively correlated with the case attribute importance score, the attribute application scenario score, the user demand feedback value, and negatively correlated with the business rule constraint value. The formula for the weighted influence coefficient of case attribute semantic similarity is:

[0092]

[0093] Where: α2 is the weight influence coefficient of the semantic similarity of case attributes, d r Score the importance of case attributes, z fk Score the attribute application scenario, Y yw Provide feedback value for user needs. σ and ε are both empirical adjustment coefficients with values ​​between [1, 2], obtained based on numerical analysis of historical data;

[0094] Step C3, obtain the semantic distance similarity weight of the ontology concept and the semantic similarity weight of the case attribute: the semantic distance similarity weight of the ontology concept is the ratio of the semantic distance similarity weight influence coefficient of the ontology concept to the semantic similarity weight influence coefficient of the case attribute multiplied by 0.5. The formula of the semantic distance similarity weight of the ontology concept is: The case attribute semantic similarity weight is 1 minus the ontology concept semantic distance similarity weight, and the case attribute semantic similarity weight is ω2=1-ω1.

[0095] This method allows for comprehensive consideration of multiple factors that affect weights, including ontology structure, domain expert needs, application needs, attribute importance, and user feedback, which helps to more comprehensively consider the impact of different factors on weights; by adjusting weights based on actual conditions based on analysis of historical data, this adaptability can cope with different application scenarios and changing needs, making the system more flexible; by setting weights based on actual needs and expert opinions, the accuracy of knowledge retrieval can be improved. Different fields and applications have different requirements for the similarity of different attributes and ontology concepts. This method helps to meet these needs; by adjusting the empirical adjustment coefficient, weights can be set according to specific circumstances. This method is suitable for a variety of different knowledge retrieval tasks because it can be customized according to the characteristics of different tasks; this method can be adjusted over time and as data changes, to adapt to evolving needs and domain knowledge.

[0096] σ and ε are both empirical adjustment coefficients with values ​​between [1, 2], obtained based on numerical analysis of historical data. These coefficients are limited to the range of [1, 2] to ensure that they do not deviate too much from the original weight impact data. This range is selected based on experience to prevent the coefficient values ​​from being biased to extremes and to maintain reasonable relative changes. θ, μ, and ρ are both empirical adjustment coefficients with values ​​between [1, 10], obtained based on numerical analysis of historical data. These coefficients are limited to the range of [1, 10] to ensure that they can balance the impact of various factors on the weights. Different factors have large differences in importance, so these larger ranges allow for greater weight differences. The values ​​of the empirical adjustment coefficients are determined based on the analysis and fitting of historical data. They are obtained by performing statistical analysis on known data to find the values ​​that best fit the data. The values ​​of these coefficients will be tried and adjusted many times in actual applications to evaluate their effects on knowledge retrieval. By continuously optimizing the values ​​of the coefficients, the performance and accuracy of the knowledge retrieval system can be improved. When determining the values ​​of the coefficients, the opinions of domain experts can also be sought to ensure that they are in line with domain knowledge and actual needs. Domain experts will provide suggestions on appropriate coefficient ranges.

[0097] With the help of ontology modeling tools or knowledge engineers, construct the domain ontology structure. The ontology is a hierarchical structure, so it is possible to calculate the distance from each concept to the root node and then calculate the average. Discuss and consult with experts in the field to obtain their opinions and needs. Experts can provide an assessment of the relative importance of ontology concepts to obtain the domain expert demand setting value. The application demand scoring of ontology concepts assigns relative importance scores to ontology concepts by analyzing domain knowledge and actual application needs, which requires the help of domain experts. The case attribute importance scoring assigns scores to attributes based on their importance, which is conducted through expert review meetings or online surveys. The attribute application scenario scoring assigns relative scores to each attribute based on its importance in different application scenarios, which should integrate the opinions of domain experts and actual application situations. User demand feedback values ​​understand user needs and preferences for different attributes through user surveys, feedback or analysis of user behavior, and help determine user demand feedback values. Business rule constraint values ​​understand the restrictions and constraints on different attributes by analyzing the business rules and constraints of the domain, and help determine the business rule constraint values.

[0098] Attribute similarity measures the similarity of case attribute values. Considering that the attribute values ​​of engineering equipment test cases are all entity types and difficult to quantify, this paper sorts out and hierarchically divides the attributes listed in the "seven-step method" above, such as Figure 3 Among them, compared with the equipment with the "cold zone" climate environment attribute, the equipment with the "subtropical" climate environment attribute is more similar to the temperate zone equipment, so Figure 3 The "climate environment" attribute is first divided into temperate and cold zones, and then further subdivided to ensure that when calculating the similarity according to the node distance, the distance between subtropical equipment and temperate equipment is smaller than the distance between subtropical equipment and cold zone equipment; similarly, in the attribute "equipped troops", the division into "light and medium combined brigades" and "medium and heavy combined brigades" is first carried out, and then further subdivided.

[0099] It should be noted that in step 2, in the case attribute semantic similarity calculation, the weight of each attribute is multiplied by the sum of the similarities of the corresponding number attributes of the target case and the historical case as the case attribute semantic similarity. The formula for the case attribute semantic similarity is:

[0100]

[0101] Where: j is the attribute number, n is the total number of attributes in the case, γ j is the weight of the attribute numbered j, where are the j-th attributes of the target case and the historical case, respectively. is the similarity between the target case and the jth attribute of the historical case.

[0102] It should be noted that in the formula of case attribute semantic similarity, the similarity between the jth attribute of the target case and the historical case is the ratio of the case attribute depth minus 1 times, the difference in the shortest path Euclidean distance between the jth attribute of the target case and the historical case in the ontology structure hierarchy, and the case attribute depth minus 1 times. The similarity formula between the jth attribute of the target case and the historical case is:

[0103]

[0104] Where: Dept(Onto)′ is the case attribute depth, which is the number of edges contained in the shortest path between the case attribute node and the root node. is the jth attribute of the target case and the jth attribute of the historical case The shortest path Euclidean distance in the ontology hierarchy.

[0105] It should be noted that in the similarity formula between the target case and the jth attribute of the historical case, the case attribute path distance and node depth are comprehensively considered, and the case attribute depth is the jth attribute of the target case. and the jth attribute of the historical case The sum of the node depths of the target case j is the same as 1. The jth attribute of the historical case The ratio of the sum of the absolute values ​​of the differences, the case attribute depth formula is:

[0106]

[0107] Where: are the jth attribute of the target case and the jth attribute of the historical case The node depth of .

[0108] By considering the depth and path distance of attributes, the similarity measure is not only based on the similarity of attribute values, but also takes into account the position of the attributes in the ontology structure, making the similarity measure more comprehensive and better reflecting the relationship between attributes; the calculation of the case attribute depth takes into account the depth of the attribute in the ontology structure, which helps to capture the structural information of the attribute in the ontology. Therefore, the similarity measure can better reflect the position of the attribute in the knowledge system; multiplying the weight of the attribute by the sum of the similarity measures can ensure that the contribution of each attribute to the overall similarity is reasonably considered, which helps to balance the weights between attributes to obtain a more reasonable similarity evaluation; by considering the attribute depth and path distance, the similarity measure can adapt to the differences between different ontology structures and attributes, making the method more universal and applicable to various different fields and knowledge systems.

[0109] When pushing cases, they are sorted in descending order according to the similarity. The entire case retrieval and push process is as follows: Figure 4 shown.

[0110] Before editing the design document content, input the instantiation information of the case ontology to be designed. The system uses a two-layer retrieval algorithm to calculate the conceptual semantic similarity and attribute semantic similarity between the target case and the historical cases in the case library and weights them, sorts the historical cases according to the similarity, and outputs similar cases retrieved once.

[0111] It should be noted that in step 4, the keywords in the pushed engineering equipment test case documents are captured, and secondary knowledge retrieval is performed based on keyword synonym expansion. The process of pushing secondary retrieval cases includes:

[0112] Step B1, generating an extended synonym set of keywords in the engineering equipment test text file by using an engineering equipment test design domain synonym library that takes into account both general knowledge and professional domain knowledge;

[0113] Step B2, perform synonym expansion query based on Lucence Chinese search engine to retrieve cases in a push;

[0114] Step B3, reorder the cases in the first push, output the secondary search results, and push the secondary search cases.

[0115] By using synonym expansion, more documents and cases related to the original keywords can be captured, which helps to improve the coverage of knowledge retrieval and ensure that important information related to the query is not missed; synonym expansion helps correct spelling errors or the use of different terms to describe the same concept, which helps to improve the accuracy of retrieval and ensure that content that better matches user needs is retrieved; in the first search, a large number of relevant documents will appear, some of which are not relevant enough. By re-ranking in the second search, the results are optimized according to a wider range of keywords to reduce information overload and provide more meaningful results; by providing more comprehensive and accurate retrieval results, users can more easily find the information they need, thereby improving user satisfaction and helping to ensure the practical application value of the knowledge retrieval system; this method allows adaptation to knowledge retrieval needs in different fields, because synonym expansion can be generated based on the synonym library of a specific field, which increases the versatility of the method.

[0116] The invention is based on the Lucene Chinese search engine, introduces the idea of ​​keyword synonym expansion query, expands the keywords into phrases that can fully express their semantics, overcomes the problem of keyword heteromorphic synonyms, realizes real-time and accurate push of similar cases, and improves the efficiency and scientificity in the field of engineering equipment test design.

[0117] Engineering equipment test design involves general terms and professional terms. Therefore, the present invention constructs a synonym thesaurus in the field of engineering equipment test design that takes into account general knowledge and professional knowledge. Among them, the general knowledge synonym dictionary uses "Synonym Dictionary (Extended Edition)". "Synonym Dictionary (Extended Edition)" was compiled and improved by the Information Retrieval Laboratory of Harbin Institute of Technology on the basis of "Synonym Dictionary". With reference to many electronic dictionary resources, it contains a total of 77,343 words, divided into 1,400 subcategories, and can perform synonymous expansion on most general words.

[0118] Regarding the construction of a synonym dictionary in a professional field, considering that there is no relevant synonym database in the field of engineering equipment test design, this paper mines the terms related to engineering equipment test design in the "Military Language of the Chinese People's Liberation Army", collects professional terms in the field, consults experts in the field of engineering equipment test design, and consults materials to organize guiding semantically similar words. This is the first time to establish a synonym dictionary in the field of engineering equipment test design. Some synonyms are shown in Table 2.

[0119] Table 2 Some synonyms in the field of engineering equipment test design

[0120]

[0121]

[0122] The secondary search process based on keyword synonym expansion is as follows Figure 5 As shown. When editing a document, capture keywords based on real-time input content. Distinguish between general terms and professional terms, and use general knowledge synonym dictionaries and professional field synonym dictionaries to expand them, respectively, to generate an extended synonym set. Consider the results of a search, select the top N push cases with high similarity in the search scope, and use the Lucence search engine to search for keywords and synonym sets. According to the search results, re-rank the push cases, and push the top M cases or historical cases with similarity greater than a certain threshold.

[0123] The workflow of the engineering equipment test design knowledge retrieval algorithm is as follows: Figure 6 Before case design, input the concept of the case ontology to be designed, use the knowledge retrieval algorithm based on the semantic similarity of the domain ontology to calculate the similarity, complete the primary search and push of historical cases, and assist the experimental designers to sort out the design ideas; in the actual design case document editing process, according to the specific input content, capture keywords, expand the keyword synonym group, and realize the secondary search and accurate push of the case based on the primary search results.

[0124] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0125] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for engineering equipment test case knowledge retrieval based on semantic similarity, comprising the following steps: Step 1: Construct the engineering equipment test case ontology; Step 2: perform knowledge retrieval analysis and calculation based on the domain ontology semantic similarity, sort the domain ontology semantic similarity in descending order, and obtain the top 5 cases in terms of domain ontology semantic similarity; Step 3: Push a search case; Step 4: Capture keywords in the pushed engineering equipment test case documents, perform secondary knowledge retrieval based on keyword synonym expansion, and push secondary retrieval cases; It is characterized in that In step 2, the knowledge retrieval analysis calculation includes ontology concept semantic distance similarity calculation, case attribute semantic similarity calculation and domain ontology semantic similarity calculation. The ontology concept semantic distance similarity calculation and case attribute semantic similarity calculation are used to obtain the ontology concept semantic distance similarity and case attribute semantic similarity respectively. The domain ontology semantic similarity is obtained by weighting the ontology concept semantic distance similarity and the case attribute semantic similarity. The domain ontology semantic similarity is used to evaluate the similarity between the target case and the candidate case. The formula for domain ontology semantic similarity is: ； Where: is the domain ontology semantic similarity, is the semantic distance similarity of the ontology concepts, is the semantic similarity of case attributes, and are the weights of the semantic distance similarity of ontology concepts and the semantic similarity of case attributes, respectively; In step 2, in the ontology concept semantic distance similarity calculation, the ontology concepts of the target case and the historical cases in the case library are numbered. The ontology concept semantic distance similarity is twice the difference between the ontology concept depth and 1, minus the Euclidean distance of the shortest path between the concept value of the target case and the corresponding numbered concept value of the historical case in the case library in the ontology structure hierarchy, and then the ratio of the difference between the ontology concept depth and 1 twice. The formula for ontology concept semantic distance similarity is: ； Where: is the depth of the ontology concept, is the number of edges contained in the shortest path between the concept node and the root node, is the number of the ontology concept, , They are numbered The concept of the target case and the historical case ontology in the case library, For concept and The shortest Euclidean distance in the ontological structure hierarchy; In the formula of domain ontology semantic similarity, the weights of ontology concept semantic distance similarity and case attribute semantic similarity are obtained as follows: Step C1, collecting weighted influence data of semantic distance similarity of ontology concepts and weighted influence data of semantic similarity of case attributes: the weighted influence data of semantic distance similarity of ontology concepts includes the average number of edges of ontology distance root nodes, the domain expert requirement setting value, and the application requirement score of ontology concepts; the weighted influence data of semantic similarity of case attributes includes the case attribute importance score, the attribute application scenario score, the user requirement feedback value, and the business rule constraint value; Step C2, obtaining the weighted influence coefficient of the semantic distance similarity of the ontology concept and the weighted influence coefficient of the semantic similarity of the case attribute: obtaining the weighted influence coefficient of the semantic distance similarity of the ontology concept through the average number of edges of the ontology distance root node, the domain expert requirement setting value and the application requirement score of the ontology concept, and obtaining the weighted influence coefficient of the semantic similarity of the case attribute through the case attribute importance score, the attribute application scenario score, the user requirement feedback value and the business rule constraint value, wherein the weighted influence coefficient of the semantic distance similarity of the ontology concept is negatively correlated with the average number of edges of the ontology distance root node, positively correlated with the domain expert requirement setting value, and positively correlated with the application requirement score of the ontology concept. The formula of the weighted influence coefficient of the semantic distance similarity of the ontology concept is: ； Where: is the weight influence coefficient of the semantic distance similarity of the ontology concept, is the average number of edges from the entity to the root node, Set values ​​for domain expert requirements, Score the application needs of the ontology concepts, , , All are empirical adjustment coefficients with values ​​between [1,10], obtained based on numerical analysis of historical data; The weighted influence coefficient of case attribute semantic similarity is positively correlated with the case attribute importance score, the attribute application scenario score, the user demand feedback value, and negatively correlated with the business rule constraint value. The formula for the weighted influence coefficient of case attribute semantic similarity is: ； Where: is the weight influence coefficient of case attribute semantic similarity, Score the importance of case attributes, Score the attribute application scenarios, Provide feedback value for user needs. , , All are empirical adjustment coefficients with values ​​between [1, 2], obtained based on numerical analysis of historical data; Step C3, obtain the semantic distance similarity weight of the ontology concept and the semantic similarity weight of the case attribute: the semantic distance similarity weight of the ontology concept is the ratio of the semantic distance similarity weight influence coefficient of the ontology concept to the semantic similarity weight influence coefficient of the case attribute multiplied by 0.

5. The formula of the semantic distance similarity weight of the ontology concept is: , the case attribute semantic similarity weight is 1 minus the ontology concept semantic distance similarity weight, and the case attribute semantic similarity weight is ; In step 2, in the case attribute semantic similarity calculation, the weight of each attribute is multiplied by the sum of the similarities of the corresponding number attributes of the target case and the historical case as the case attribute semantic similarity. The formula for the case attribute semantic similarity is: ； Where: is the attribute number, is the total number of attributes in the case, For the number The weight of the attribute, where , , The target case and the historical case are properties, For target cases and historical cases The similarity of the attributes.

2. The method for engineering equipment test case knowledge retrieval based on semantic similarity according to claim 1, characterized in that: In step 1, the process of building the engineering equipment test case ontology includes: Step A1, determine the professional field and scope of the knowledge ontology: based on the engineering equipment test, select the object of its assessment of operation efficiency indicators as engineering equipment, and determine the constructed domain ontology as the engineering equipment type; Step A2, reuse existing knowledge ontology: domain experts use the public knowledge base of the engineering corps professional field to obtain historical equipment test plans and collect concept terminology sets in the field of equipment test assessment; Step A3, define the hierarchical database system between concepts, determine the attributes of concepts, define the constraints of attributes, and create instances: use the collected equipment test and assessment field concept terminology set, use database technology to divide the levels according to the types and models of engineering equipment, build a hierarchical database system between concepts, obtain the attribute set of each concept, and the attribute set of each concept is {case category, geographical environment, type of equipment, equipment usage scenario, functional component}. The attribute value types of the engineering equipment test design ontology are all entities. According to the constructed ontology model, select a specific class, create an instance of the class, and clarify the attribute value.

3. The method for engineering equipment test case knowledge retrieval based on semantic similarity according to claim 2, characterized in that: In step A3, case categories include outlines and plans, geographical environments include plateaus, plains, and coastal areas, types of troops include engineered chemical brigades and combined brigades, and equipment usage scenarios include surface, land, and amphibious.

4. The method for engineering equipment test case knowledge retrieval based on semantic similarity according to claim 1, characterized in that: In the semantic distance similarity formula of ontology concepts, the semantic path distance and node depth are comprehensively considered, and the depth of the ontology concept is the concept and Concept The sum of the node depths, which is the same as 1 ,concept The ratio of the sum of the absolute values ​​of the node depth differences. The depth formula of the ontology concept is: ； Where: , Concept Node depth, concept The node depth of .

5. The method for engineering equipment test case knowledge retrieval based on semantic similarity according to claim 1, characterized in that: In the formula of case attribute semantic similarity, the target case and the historical case The similarity of the attribute is twice the depth of the case attribute minus 1. The difference in the shortest path Euclidean distance of each attribute in the ontology structure hierarchy is then divided into the ratio of the case attribute depth minus 1 times the target case and the historical case. The similarity formula for each attribute is: ； Where: is the case attribute depth, is the number of edges contained in the shortest path between the case attribute node and the root node, For the target case Attributes and historical case Attributes The shortest path Euclidean distance in the ontology hierarchy.

6. The method for engineering equipment test case knowledge retrieval based on semantic similarity according to claim 5, characterized in that: In the target case and historical case In the similarity formula of each attribute, the case attribute path distance and node depth are comprehensively considered. The case attribute depth is the target case Attributes and historical case Attributes The sum of the node depths of the same target case as 1 Attributes 、Historical Case No. Attributes The ratio of the sum of the absolute values ​​of the differences, the case attribute depth formula is: ； Where: , The target cases are Attributes and historical case Attributes The node depth of .

7. The method for engineering equipment test case knowledge retrieval based on semantic similarity according to claim 1, characterized in that: In step 4, the keywords in the pushed engineering equipment test case documents are captured, and secondary knowledge retrieval is performed based on keyword synonym expansion. The process of pushing secondary retrieval cases includes: Step B1, generating an extended synonym set of keywords in the engineering equipment test text file through a synonym library in the engineering equipment test design field that takes into account both general knowledge and professional field knowledge; Step B2, perform synonym expansion query based on Lucence Chinese search engine to retrieve cases in a push; Step B3, reorder the cases in the first push, output the secondary search results, and push the secondary search cases.

Images