Ontology association visualization method for nuclear power plant layout design based on semantic recognition
Through semantic recognition and TextRank algorithm, the design keywords for nuclear power plant layout are extracted, the relationship is recognized, and the ontology model is established, which solves the problem of multi-source data association in nuclear power plant layout design, realizes comprehensive and accurate visualization of information, and improves design efficiency and accuracy.
Patent Information
- Application Number
- CN202210137068.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-02-15
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2042-02-15
AI Technical Summary
The association and visualization of multi-source unstructured data in the layout design of nuclear power plants is difficult to achieve, resulting in underutilization of knowledge silos and information, affecting design efficiency and accuracy.
The semantic recognition method is adopted to extract the nuclear power plant layout design keywords through the TextRank algorithm and the linear chain conditional random field model, identify the classification and non-classification relationships between keywords, establish the layout design ontology model and visual representation.
It realizes the accurate identification and visual representation of keywords and relationships in the layout design of nuclear power plants, makes full use of multi-data source information, improves design efficiency and accuracy, and has dynamic updates and scalability.
Smart Images

Figure CN114647712B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of nuclear power plant layout design text and is a text processing and visualization method, in particular to a nuclear power plant layout design ontology association visualization method based on semantic recognition. Background Art
[0002] In the nuclear power design process, the layout design of nuclear power plants involves many disciplines, various interfaces, and complex interaction processes. It is the activity that consumes the most man-hours in nuclear power design. At the same time, how to realize the visualization of multi-disciplinary layout knowledge in the design process so that designers can accurately find the required layout design knowledge is also an important aspect. Nuclear power companies have attached great importance to the layout design of nuclear power plants and have conducted relatively extensive research. They have established and developed several information platforms, established a complete and diverse knowledge resource system, and developed a three-dimensional digital design platform for nuclear power plants. However, there are still some problems. How to effectively associate and integrate layout design resources from different platforms and different knowledge bases to obtain more comprehensive and accurate layout full-process information than a single data source, and realize the visualization of this information is of great significance to improving the level of nuclear power plant layout design.
[0003] The nature of nuclear power business and the development of information technology have determined that nuclear power knowledge engineering has been carried out in a "pure manual mode" before. Although several information platforms have been developed in the industry to establish a complete and diverse knowledge resource system, due to different management models, warehouse-style sub-database storage with the help of different information platforms, and overall management by different responsible organizations, knowledge islands have been formed in the industry. Due to the difficulty and high cost of implementing centralized storage, as well as the consideration of the interests of the leading parties of each knowledge resource, nuclear power knowledge engineering mostly adopts "sub-database storage and separate retrieval", and the limitations of the "one label for life" method of knowledge resources are becoming more and more obvious. Similarly, the storage of layout information also faces such problems. So far, the information provided by multiple knowledge bases has not been fully utilized in the business process of nuclear power plant layout, and an effective ontology association visualization method has not been formed for layout design. Summary of the invention
[0004] The purpose of the present invention is to address the above-mentioned problems existing in the prior art. Specifically, the present invention provides a method for visualizing the association of nuclear power plant layout design ontology based on semantic recognition. Aiming at multi-source unstructured data of nuclear power plant layout design, a semantic recognition method is adopted to extract keywords of nuclear power plant layout design, identify the relationship between concepts, establish a layout ontology association model, and realize the visualization representation of layout design ontology.
[0005] In order to achieve the above goals, Figure 1 As shown, the technical solution of the present invention is as follows:
[0006] The present invention establishes a nuclear power plant layout design ontology association model and realizes the visual representation of the nuclear power plant layout design ontology.
[0007] Step 1: Collect text data from various nuclear power layout design platforms as layout design text, pre-process the layout design text, and establish a nuclear power field feature dictionary;
[0008] Step 2: Use the TextRank algorithm to extract nuclear power plant layout design keywords from the nuclear power field feature dictionary and form a nuclear power plant layout design keyword set;
[0009] Step 3: Identify the relationships between the nuclear power plant layout design keywords in the nuclear power plant layout design keyword set, and the identification is divided into classified relationships and non-classified relationships;
[0010] Step 4: Based on the nuclear power plant layout design keywords extracted in the previous step and the relationship recognition results between the nuclear power plant layout design keywords, the nuclear power plant layout design ontology corresponding to the layout design text is associated with modeling, and the layout relationship is visualized using the model obtained by modeling.
[0011] In the step 1, before establishing the dictionary, a linear chain conditional random field model is used to segment text data in the preprocessing process.
[0012] The linear chain conditional random field model is used to segment text data. In the conditional probability P(Y|X), X is the given observation sequence and Y is the output sequence. The definition of the linear chain conditional random field is: Let X = (X1, X2, ...X n ), Y=(Y1,Y2,...Y n ) is a random variable sequence represented by a linear chain, X is the observed variable sequence, and Y is the corresponding labeled sequence. When the observed variable sequence X is known, the conditional probability distribution of the labeled sequence Y satisfies:
[0013] P(Y i |X,Y1,Y2,...Y n )=P(Y i |X,Y i-1 , Y i+1 )
[0014] Where i=1, 2, ..., n, in the present invention, X represents a word in the text data, and Y represents the part of speech of the word. Therefore, under the condition that the word in the text data is x, the conditional probability P(y|x) that the part of speech of the word is y is parameterized as follows:
[0015]
[0016]
[0017] In the above formula, t k is a local feature function, reflecting the characteristics of the current node and the previous node, and the weight coefficient is λ k , k = 1, 2, ... K, K represents the total number of local feature functions defined at the node; s l It is the node feature function, reflecting the current node characteristics, and the weight coefficient is μ l , l = 1, 2, ... L, L represents the total number of node feature functions defined at the node; Z(x) is a normalized function, x represents the current node, and i represents the position of the current node in the sequence X.
[0018] In the step 1, the linear chain conditional random field model is used in the preprocessing to segment the text data, and the feature words of nuclear power field terminology and layout design technical status feature words are extracted during the segmentation process. The feature words of nuclear power field terminology and layout design technical status feature words are merged into a nuclear power field feature dictionary, which can solve the adaptability problem in the field of nuclear power plant layout design.
[0019] In the step 2, specifically:
[0020] S2.1. Use the TextRank algorithm to construct a word graph based on the co-occurrence relationship between words in the nuclear power field feature dictionary. In the word graph, a word is used as a node to form a word node, and the edge between word nodes represents the co-occurrence relationship between two words.
[0021] The co-occurrence relationship refers to the co-occurrence of words within a window of a fixed length, and the window of a fixed length is preset.
[0022] S2.2. In terms of the semantic clustering importance of words, the present invention introduces the semantic clustering importance of node word vectors. The word vector of each word in the nuclear power field feature dictionary processing generator is processed by the BERT method as the word vector of the layout design text, and the AP clustering algorithm is used to perform similarity clustering on the word vectors to obtain each cluster C j ;
[0023] First calculate each word node v according to the following formula j The semantic clustering importance w c (v j ), and then according to the semantic clustering importance w c (v j ) to obtain the probability of semantic clustering importance jump between word nodes As shown below:
[0024]
[0025]
[0026] In the above formula, Represents word node v j The cluster C j The vector of the cluster center is obtained by averaging all the word vectors in the cluster. c (v i ) represents word node v i The semantic clustering importance of Represents word node v j To cluster C j The vector of cluster centers, C j Represents word node v j The cluster where ; Represents word node v j and word node v i The jump probability of semantic clustering importance between them; Out(v j ) represents all word nodes v j The set of word nodes pointed to;
[0027] S2.3. In terms of the statistical importance of words, the present invention introduces the frequency fluctuation degree into the TF-IDF algorithm. First, calculate each word node v according to the following formula j The statistical importance of w s (v j ), and then according to the statistical importance w s (v j ) to obtain the statistical importance jump probability between word nodes As shown below:
[0028] w s (v j )=σ j ×tf jg ×idf j
[0029]
[0030] In the above formula, σ j Represents word node v j The degree of word frequency fluctuation, tf jg Represents word node v j word frequency, idf j Represents word node v j The inverse document frequency of w s (v i ) represents word node v i The semantic clustering importance of Represents word node v j and word node v i The probability of statistical importance jump between time periods;
[0031] S2.4. Calculate the weighted sum of the semantic clustering importance jump probability and the statistical importance jump probability to obtain the word node jump probability:
[0032]
[0033] α1+β1=1
[0034] Among them, p ji Represents word node v j and word node v i The probability of word node jump, α1 and β1 are the influence coefficients of semantic clustering importance and statistical importance respectively;
[0035] S2.5. The word node jump probability p ji Substitute the following formula to iteratively calculate the weight WS(v i ), as shown below:
[0036]
[0037]
[0038] In the above formula, d is the damping coefficient, which represents the pointing probability between word nodes, d∈[0,1], and its value is usually 0.85 to ensure the convergence of iterative calculation; WS(v i ) represents word node v i The weight, w ji is the word node v j and word node v i The weight of the edge between them, w ji is a known quantity, and the edge weight between nodes is generally set to 1, so that the jumps between word nodes are simply set to be uniformly distributed. i ) indicates all nodes v pointing to the word i The word node set, Out(v j ) represents all word nodes v j The set of word nodes pointed to;
[0039] v i Represents any word node, v k Indicated by v i The connected word nodes, v j Indicates pointing to word node v i All word nodes.
[0040] This is equivalent to repeatedly executing WS(v i ) operation steps, the process of calculating the following quantities from the previous quantities in sequence, is WS(v i ) by WS(v i-1) is calculated. The initial value of the weight is 1 and is updated during the iteration process.
[0041] S2.6, the weight of each word node WS (v i ) are sorted, and after sorting, the first K words are taken as the nuclear power plant layout design keywords of the word nodes to form the nuclear power plant layout design keyword set.
[0042] In the step 2, the semantic relevance of the preprocessed layout design text is clustered, and the semantic clustering importance and statistical importance of the node word vector are introduced to optimize the word node jump probability of TextRank, and the keyword set of the layout design text is obtained by iteration, sorting and induction. Such processing can achieve a more accurate acquisition of the nuclear power plant layout design keyword set, and the obtained nuclear power plant layout design keyword set is more in line with the theme of nuclear power plant layout design.
[0043] In the step three:
[0044] S3.1. For all the nuclear power plant layout design keywords obtained in step 2, perform the following processing:
[0045] The nuclear power plant layout design keyword ranked first in the nuclear power plant layout design keyword set is used as the top nuclear power plant layout design keyword, and the top nuclear power plant layout design keyword is used as the root node. The maximum spanning tree of the nuclear power plant layout design keyword set is obtained based on the Prim algorithm, and then the word node pair set is obtained according to the set of all edges in the obtained maximum spanning tree;
[0046] S3.2. For a word node pair v in the word node pair set ci and v cj , calculate the similarity importance of each word node pair:
[0047]
[0048] in, Represents word node pair v ci and v cj The frequency vector of the terms in the collection;
[0049] S3.3. For the word node pairs in the maximum spanning tree, the search results are specifically the keyword node pairs of nuclear power plant layout design by searching in the nuclear power knowledge resource library and Web resources. The search results are matched with the keywords in the preset classification relationship language template, and then the association importance of the word node pairs is calculated:
[0050]
[0051] In the above formula, val(v ci , v cj) represents the number of matches between word node pairs and classification relationship language template keywords in the search results;
[0052] The classification relationship language template refers to genus-species relationship (A belongs to B, etc.), whole-part relationship (A constitutes / consists of B, etc.), etc., and is pre-set based on the classification relationship in the text data.
[0053] S3.4. According to the similarity importance and association importance of the word node pair, the weighted sum is obtained:
[0054] w cr (v ci , v cj )=α2×w1(v ci , v cj )+β2×w2(v ci , v cj )
[0055] α2+β2=1
[0056] In the above formula, w cr (v ci , v cj ) represents the word node pair v ci and v cj importance; α2 represents the similarity importance coefficient, and β2 represents the association importance coefficient.
[0057] S3.5. Compare and judge the importance of each word node pair with the preset importance threshold:
[0058] If the importance of the word-node pair is greater than the importance threshold, the word-node pair is regarded as a classification relationship;
[0059] If the importance of a word-node pair is not greater than the importance threshold, the word-node pair is treated as a non-classification relationship.
[0060] The classification relationship refers to the conceptual relationship between two words, for example, the main plant area includes the reactor plant.
[0061] In this way, by setting the importance threshold of the classification relationship and removing redundant relationships, the classification relationship can be corrected to obtain a classification relationship set of word node pairs.
[0062] In the step 4, the model obtained by modeling is used to realize visual representation according to the nuclear power plant layout design keywords, classification relationships and non-classification relationships of the keywords determined in the above steps.
[0063] Compared with the prior art, the present invention has the following beneficial effects:
[0064] The method of the present invention aims at the inherent characteristics of the layout relationship in the layout design of a nuclear power plant, realizes the layout design text preprocessing of multi-source unstructured data of the layout relationship in the layout design of a nuclear power plant, the recognition of nuclear power plant layout design keywords, the recognition of classified relationships and non-classified relationships between keywords based on the semantic recognition method, and realizes the visual representation of the layout design ontology.
[0065] By making full and effective use of the information provided by multiple data sources of the nuclear power plant layout design platform and optimizing the jump probability of word nodes, more accurate and suitable nuclear power plant layout design keywords can be obtained. In addition, after processing the word node pairs, not only the comprehensive and accurate identification of keywords and the classified and non-classified relationships between keywords in the nuclear power plant layout design is realized, but also the visualization representation of the nuclear power plant layout design ontology is further realized, which provides important support for the mining and management of the nuclear power plant layout design relationship. At the same time, the visualization expression in the present invention can be dynamically updated, expanded and enriched, has good maintainability and scalability, and is conducive to further services and applications in the later stage. BRIEF DESCRIPTION OF THE DRAWINGS
[0066] Figure 1 Schematic diagram of the method of the present invention.
[0067] Figure 2 The present invention provides a flow chart for extracting keywords for the layout and design of a nuclear power plant.
[0068] Figure 3 This is a schematic diagram of the visualization results of the nuclear power plant layout design ontology implemented by the present invention using Protégé.
[0069] Figure 4 This is a detailed schematic diagram of the visualization results of the nuclear power plant layout design ontology implemented by the present invention using Protégé. DETAILED DESCRIPTION
[0070] The present invention is further described in detail below in conjunction with specific implementation methods.
[0071] like Figure 1 As shown, the embodiments of the present invention and their implementation process are as follows:
[0072] Step 1: Collect text data from various nuclear power layout design platforms as layout design text, pre-process the layout design text based on natural language processing methods, and establish a nuclear power field feature dictionary;
[0073] In the preprocessing, the linear chain conditional random field model is used to segment the text data. During the segmentation process, the characteristic words of nuclear power field terminology and layout design technical status characteristic words are extracted. The characteristic words of nuclear power field terminology and layout design technical status are fused to form a nuclear power field feature dictionary, which can solve the adaptability problem in the field of nuclear power plant layout design.
[0074] The linear chain conditional random field model is used to segment text data. In the conditional probability P(Y|X), X is the given observation sequence and Y is the output sequence. The definition of the linear chain conditional random field is: Let X = (X1, X2, ...X n ), Y=(Y1,Y2,...Y n ) is a random variable sequence represented by a linear chain, X is the observed variable sequence, and Y is the corresponding labeled sequence. When the observed variable sequence X is known, the conditional probability distribution of the labeled sequence Y satisfies:
[0075] P(Y i |X,Y1,Y2,...Y n )=P(Y i |X,Y i-1 , Y i+1 )
[0076] Where i = 1, 2, ..., n, in the present invention, X represents a word in the text data, and Y represents the part of speech of the word. Therefore, under the condition that the word in the text data is x, the conditional probability P(y|x) of the part of speech of the word being y is parameterized as follows: (nodes are words)
[0077]
[0078]
[0079] Step 2: If Figure 2 As shown, the TextRank algorithm is used to extract the keywords of nuclear power plant layout design from the feature dictionary of nuclear power field, and a set of keywords of nuclear power plant layout design is formed;
[0080] S2.1. Use the TextRank algorithm to first construct a word graph G = (V, E) based on the co-occurrence relationship between words in the nuclear power field feature dictionary, where V represents a word node set and E represents an edge set between word nodes; in the word graph, a word node is formed with one word as a node, and the edge between word nodes represents the co-occurrence relationship between two words;
[0081] S2.2. Use the BERT method to process the word vector of each word in the nuclear power field feature dictionary generator as the word vector of the layout design text semantics, and use the AP clustering algorithm to cluster the word vectors by similarity to obtain the clusters C after clustering. j ;
[0082] First calculate each word node v according to the following formula j The semantic clustering importance w c (v j ), and then according to the semantic clustering importance w c (vj ) to obtain the probability of semantic clustering importance jump between word nodes As shown below:
[0083]
[0084]
[0085] S2.3, calculate each word node v according to the following formula j The statistical importance of w s (v j ), and then according to the statistical importance w s (v j ) to obtain the jump probability of the statistical importance between word nodes As shown below:
[0086] w s (v j )=σ j ×tf jg ×idf j
[0087]
[0088] S2.4. Calculate the weighted sum of the semantic clustering importance jump probability and the statistical importance jump probability to obtain the word node jump probability:
[0089]
[0090] α1+β1=1
[0091] Among them, p ji Represents word node v j and word node v i The probability of word node jump, α1 and β1 are the influence coefficients of semantic clustering importance and statistical importance respectively;
[0092] S2.5. Change the word node jump probability p ji Substitute the following formula to calculate the weight WS(v i ), as shown below:
[0093]
[0094]
[0095] S2.6, the weight of each word node WS (v i ) are sorted, and after sorting, the first K words are taken as the nuclear power plant layout design keywords of the word nodes to form the nuclear power plant layout design keyword set.
[0096] Step 3: Identify the relationships between the nuclear power plant layout design keywords in the nuclear power plant layout design keyword set, and the identification is divided into classified relationships and non-classified relationships in the nuclear power plant layout design keyword set;
[0097] S3.1. For all the nuclear power plant layout design keywords obtained in step 2, perform the following processing:
[0098] The nuclear power plant layout design keyword ranked first in the nuclear power plant layout design keyword set is used as the top nuclear power plant layout design keyword, and the top nuclear power plant layout design keyword is used as the root node. The maximum spanning tree of the nuclear power plant layout design keyword set is obtained based on the Prim algorithm, and then the word node pair set is obtained according to the set of all edges in the obtained maximum spanning tree;
[0099] S3.2. For a word node pair v in the word node pair set ci and v cj , use cosine similarity to calculate the similarity importance of each word node pair:
[0100]
[0101] S3.3, for the word node pairs in the maximum spanning tree, by searching in the known nuclear power knowledge resource library and Web resources, the search results are specifically the search results for the nuclear power plant layout design keyword node pairs, the search results are matched with the keywords in the preset classification relationship language template, the classification relationship language template is shown in Table 1, and then the association importance of the word node pairs is calculated:
[0102]
[0103] Table 1 Classification relationship language template
[0104]
[0105]
[0106] S3.4. According to the similarity importance and association importance of the word node pair, the weighted sum is obtained:
[0107] w cr (v ci , v cj )=α2×w1(v ci , v cj )+β2×w2(v ci , v cj )
[0108] α2+β2=1
[0109] In the above formula, w cr (vci , v cj ) represents the word node pair v ci and v cj The importance of
[0110] S3.5. Compare and judge the importance of each word node pair with the preset importance threshold:
[0111] If the importance of the word node pair is greater than the importance threshold, the word node pair is regarded as a classification relationship; if the importance of the word node pair is not greater than the importance threshold, the word node pair is regarded as a non-classification relationship. Non-classification relationships include attribute relationships such as area and distance, such as the reactor building covers an area of 2045 square meters; it also includes location relationships, such as the switch station is located on the north side of the main plant area.
[0112] Step 4: Based on the nuclear power plant layout design keywords extracted in the previous step and the relationship recognition results between the nuclear power plant layout design keywords, use the Protégé software tool to model the nuclear power plant layout design ontology corresponding to the layout design text, and use the model obtained by modeling to visualize the layout relationship.
[0113] Specifically, the model obtained by modeling is used to achieve visual representation according to the predetermined classification relationship and non-classification relationship of nuclear power plant layout design keywords.
[0114] In step 4, the knowledge of nuclear power plant layout design is collected, the domain and scope of the ontology are determined, the core semantic classes, core semantic class attribute relationships, and constraint relationships of the nuclear power plant layout design are defined, and the specific implementation is based on ontology implementation and ontology evaluation, and the OntoGraf module is used to realize the visualization of the nuclear power plant layout design ontology association modeling, such as Figure 3 , Figure 4 shown.
[0115] The specific embodiments are as follows:
[0116] First, according to the set of identified nuclear power plant layout design keywords, the ontology domain and scope are determined. The ontology is obtained from the identified nuclear power plant layout design keywords. The reuse possibility of the existing ontology is examined. Then, according to the identified classification relationship, the layout design ontology of the main equipment of the nuclear power plant is classified:
[0117] ① Nuclear power plants: pressurized water reactor nuclear power plants, heavy water reactor nuclear power plants, high temperature gas-cooled reactor nuclear power plants, etc.;
[0118] ② Area: three repair areas, three waste areas, main plant area, warehouse area, front area, switch station, circulating cooling water system, auxiliary production area, etc.;
[0119] ③ Buildings / Structures: Reactor building, nuclear fuel building, nuclear auxiliary building, steam turbine, generator building, conventional island waste liquid storage tank building, laundry room and laboratory, etc.;
[0120] ④System: main feed water system, main steam system, waste heat removal system, chemical and volume control system, etc.;
[0121] ⑤Function: nuclear reaction, storage, life, work, etc.
[0122] According to the identified non-classified relationships, attribute relationships and constraints are determined. Attribute types mainly include object attributes, that is, the relationship between entities; they also include data attributes, that is, the data characteristics of the entity itself. In the layout design of nuclear power plants, it is found from non-classified relationships that the main relationships of the object attributes of the layout entity are interface connection, east side, west side, south side, north side, inclusion, non-inclusion, and separation, such as the switch station is arranged on the north side of the main plant area; the main relationships of data attributes are area, location, etc., such as the reactor plant covers an area of 2045 square meters.
[0123] The Protégé software tool is used to model the nuclear power plant layout design ontology according to the above rules and realize its visualization. The layout design ontology association visualization results are as follows: Figure 3 At the same time, a simple reasoning arrangement relationship can be obtained from Figure 4 It can be seen that the main feed water system enters the reactor building through the connecting building, and the main feed water pipeline plays a connecting role. Designers can use visual layout relationships and simple reasoning to verify the correctness of the layout design relationship of the nuclear power plant.
[0124] In this way, the information provided by the multiple data sources of the nuclear power layout design platform can be fully and effectively utilized, which not only realizes the comprehensive and accurate identification of the classified and non-classified relationships between keywords in the nuclear power plant layout design, but also realizes the visual representation of the nuclear power plant layout design ontology, which can provide greater convenience for nuclear power plant layout designers and provide important support for the mining and management of nuclear power plant layout design relationships.
Claims
1. A method for visualization of ontology association of nuclear power plant layout design based on semantic recognition, characterized by: The method comprises the following steps in sequence: Step 1: Collect text data from various nuclear power layout design platforms as layout design text, pre-process the layout design text, and establish a nuclear power field feature dictionary; Step 2: Use the TextRank algorithm to extract nuclear power plant layout design keywords from the nuclear power field feature dictionary and form a nuclear power plant layout design keyword set; Step 3: Identify the relationships between the nuclear power plant layout design keywords in the nuclear power plant layout design keyword set, and the identification is divided into classified relationships and non-classified relationships; Step 4: Based on the nuclear power plant layout design keywords extracted in the previous step and the relationship recognition results between the nuclear power plant layout design keywords, the nuclear power plant layout design ontology corresponding to the layout design text is modeled, and the layout relationship is visualized using the model obtained by modeling; In the step 2, specifically: S2.
1. Use the TextRank algorithm to construct a word graph based on the co-occurrence relationship between words in the nuclear power field feature dictionary. In the word graph, a word is used as a node to form a word node, and the edge between word nodes represents the co-occurrence relationship between two words. S2.
2. Use the BERT method to process the word vector of each word in the nuclear power field feature dictionary generator as the word vector of the layout design text, and use the AP clustering algorithm to perform similarity clustering on the word vectors to obtain each cluster C j ; Calculate each word node v according to the following formula j The semantic clustering importance w c (v j ), and then according to the semantic clustering importance w c (v j ) to obtain the probability of semantic clustering importance jump between word nodes As shown below: In the above formula, Represents word node v j The cluster C j The vector of cluster centers; w c (v i ) represents word node v i The semantic clustering importance of Represents word node v j To cluster C j The vector of cluster centers, C j Represents word node v j The cluster where ; Represents word node v j and word node v i The jump probability of semantic clustering importance between them; Out(v j ) represents all word nodes v j The set of word nodes pointed to; S2.3, calculate each word node v according to the following formula j The statistical importance of w s (v j ), and then according to the statistical importance w s (v j ) to obtain the statistical importance jump probability between word nodes As shown below: w s (v j )=s j ×tf jg ×idf j In the above formula, σ j Represents word node v j The degree of word frequency fluctuation, tf jg Represents word node v j word frequency, idf j Represents word node v j The inverse document frequency of w s (v i ) represents word node v i The semantic clustering importance of Represents word node v j and word node v i The probability of statistical importance jump between time periods; S2.
4. Calculate the weighted sum of the semantic clustering importance jump probability and the statistical importance jump probability to obtain the word node jump probability: α1+β1=1 Among them, p ji Represents word node v j and word node v i The probability of word node jump, α1 and β1 are the influence coefficients of semantic clustering importance and statistical importance respectively; S2.
5. The word node jump probability p ji Substitute the following formula to calculate the weight WS(v i ), as shown below: In the above formula, d is the damping coefficient, which represents the pointing probability between word nodes; WS(v i ) represents word node v i The weight, w ji is the word node v j and word node v i The weight of the edge between them, In(v i ) indicates all nodes v pointing to the word i The word node set, Out(v j ) represents all word nodes v j The set of word nodes pointed to; S2.6, the weight of each word node WS (v i ) are sorted, and after sorting, the first K words are taken as the nuclear power plant layout design keywords of the word nodes to form the nuclear power plant layout design keyword set.
2. According to claim 1, a method for ontology association visualization of nuclear power plant layout design based on semantic recognition, characterized in that: In the step 1, before establishing the dictionary, a linear chain conditional random field model is used to segment text data in the preprocessing process.
3. A method for visualization of nuclear power plant layout design ontology association based on semantic recognition according to claim 1 or 2, characterized in that: In the step 1, the linear chain conditional random field model is used in the preprocessing to segment the text data, and the nuclear power field terminology feature words and layout design technical status feature words are extracted during the word segmentation process, and the nuclear power field terminology feature words and layout design technical status feature words are merged into a nuclear power field feature dictionary.
4. The method for ontology association visualization of nuclear power plant layout design based on semantic recognition according to claim 1, characterized in that: In the step three: S3.
1. For all the nuclear power plant layout design keywords obtained in step 2, perform the following processing: The nuclear power plant layout design keyword ranked first in the nuclear power plant layout design keyword set is used as the top nuclear power plant layout design keyword, and the top nuclear power plant layout design keyword is used as the root node. The maximum spanning tree of the nuclear power plant layout design keyword set is obtained based on the Prim algorithm, and then the word node pair set is obtained according to the set of all edges in the obtained maximum spanning tree; S3.
2. For a word node pair v in the word node pair set ci and v cj , calculate the similarity importance of each word node pair: in, Represents word node pair v ci and v cj The frequency vector of the terms in the set; S3.
3. For the word node pairs in the maximum spanning tree, by searching in the nuclear power knowledge resource library and Web resources, the search results are matched with the keywords in the preset classification relationship language template, and then the association importance of the word node pairs is calculated: In the above formula, val(v ci ,v cj ) represents the number of matches between word node pairs and classification relationship language template keywords in the search results; S3.
4. According to the similarity importance and association importance of the word node pair, the weighted sum is obtained: w cr (v ci ,v cj )=α2×w1(v ci ,v crj )+β2×w2(v ci ,v cj ) α2+β2=1 In the above formula, w cr (v ci ,v cj ) represents the word node pair v ci and v cj The importance of; α2 represents the similarity importance coefficient, β2 represents the association importance coefficient; S3.
5. Compare and judge the importance of each word node pair with the preset importance threshold: If the importance of the word-node pair is greater than the importance threshold, the word-node pair is regarded as a classification relationship; If the importance of a word-node pair is not greater than the importance threshold, the word-node pair is treated as a non-classification relationship.
5. The method for ontology association visualization of nuclear power plant layout design based on semantic recognition according to claim 1, characterized in that: In the step 4, the model obtained by modeling is used to realize visual representation according to the nuclear power plant layout design keywords, classification relationships and non-classification relationships of the keywords determined in the above steps.
Citation Information
Patent Citations
System and method for heterogeneous information mining and visual analysis
CN102609512A
Method for identifying emotion key sentence on basis of lexical semantics and syntactic dependency
CN104281645A
Method for constructing correlation networks of keywords of natural language texts
CN104391963A
An intelligent reasoning method of archival data based on semantic ontology
CN109271484A
Method and device for keyphrase extraction
WO2017084267A1
