A multi-dimensional coupling calculation method for scientific and technological elements, transformation and structure

Through multi-dimensional coupled computing methods, a science-technical knowledge network is built to measure the correlation between science and technology systems, and the uncertainty of correlation measurement in the existing technology is solved, and the comprehensive relationship between science and technology is achieved. , accurately understand.

CN118586790BActive Publication Date: 2024-11-19NANJING UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411076008.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-07
Publication Date
2024-11-19
Estimated Expiration
2044-08-07

AI Technical Summary

Technical Problem

It is difficult for the existing technology to comprehensively and accurately measure the correlation between science and technology systems in the process of knowledge architecture, content semantic correlation, and innovation transformation, resulting in uncertainty in data omissions and single linear relationship calculations in science-technology relationship detection.

Method used

Using multi-dimensional coupled calculation method for scientific and technological elements-transformation-structure, we will build the relationship between the three dimensions of input-output model, innovation transformation efficiency, and knowledge system by integrating multi-source heterogeneous data such as scientific literature, technical patents, and industry statistical yearbooks. , use BERT and SciBERT models to generate word embeddings, build a science-technical knowledge network, calculate the coupling degree of science-technical elements, the coupling degree of conversion efficiency and the coupling degree of knowledge network.

Benefits of technology

A comprehensive and accurate measurement of science and technology systems has been achieved, the one-sidedness and instability of the single measurement relationship has been eliminated, and the complexity of the science-technology relationship and the effects and mechanisms of collaborative innovation have been revealed.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118586790B_ABST
    Figure CN118586790B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for calculating the multidimensional coupling of science and technology element-transformation-structure, comprising the following steps: (1) obtaining input-output statistical data, scientific literature and technology patent literature collection of the target emerging industry field, and performing preprocessing; (2) classifying, extracting and counting data based on the data, scientific literature and patent data sets collected in step (1) with cities as the benchmark; (3) using the BERT model to generate word embeddings of document-level data based on the data, scientific literature and patent data sets collected in step (1); using the SciBERT model to extract the core knowledge entities of science and technology and construct a science and technology knowledge network; (4) constructing an element-transformation-structure multidimensional coupling scheme to obtain a multidimensional coupling degree; (5) constructing a science-technology dynamic coupling type and a time-lag relationship quantification scheme based on the obtained multidimensional coupling degree; the present invention comprehensively and accurately measures the "science-technology relationship".
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the cross-technical field of input-output evaluation methods and computer systems, and in particular to a scientific and technological element-conversion-structure multi-dimensional coupling calculation method. Background Art

[0002] In the current innovation practice process, many industries still face bottlenecks in the transfer and transformation of scientific and technological achievements. Some scientific achievements cannot be transformed and applied to technological practice in a timely manner. Many technical problems cannot be effectively solved due to the lack of new scientific achievements. The connection between scientific basic research and technological application research is not in place, and the path of collaborative innovation is unclear. Although this situation may be caused by many complex reasons such as the institutional mechanism, subject capabilities, and service system in an industry, it is related to the island phenomenon of scientific and technological resources, the poor flow of scientific and technological knowledge, and the low synergy. The layout of strategic emerging industries often requires the deployment of synergy between basic scientific research and technological innovation activities, and the basis of this synergy is reflected in the coupling between scientific systems and technological systems. Science-technology correlation detection is an important way to explore and discover scientific and technological coupling.

[0003] The concrete observation of the interaction between science and technology in quantitative analysis originates from the verification of citation analysis. The prelude to the exploration of the relationship between science and technology represented by papers and patents is thus unveiled. Papers cite patents, patents cite patents, and papers and patents cite each other, forming a network of science-technology citation relationships. The small-world network structure, power-law degree distribution, hierarchical structure, and patents citing unrelated papers make it difficult to use citation analysis to understand the organization of science-technology knowledge. In order to supplement more clues to capture the characteristics of science-technology interaction and the network relationship between scientists and inventors, IPC-ISI (International Patent Classification - Institute for Scientific Information) category mapping is used as a more specific solution for the characteristics of science-technology interaction. However, due to the bias in citation purpose, the sparsity of the author network, and the difficulty in matching IPC-ISI categories, the results obtained by these means in the actual measurement of science-technology interaction are not robust. Furthermore, methods based on keyword co-occurrence, theme, and knowledge network coupling deeply analyze the science-technology system from the knowledge granularity. This type of comprehensive semantic and structural representation method can effectively quantify the interaction between science and technology.

[0004] However, the coupling of science and technology is not only reflected in the coupling of knowledge systems, but also in the coupling of "input-output" in the innovation production process of the two. In the existing scientific research output performance evaluation system, the consideration of output scale is more than the discussion of the "science-technology" relationship, that is, the emphasis is on the quantitative development of "paper publication-patent application". Although the existing "science-technology relationship" detection methods provide different vantage points to reveal the complex science-technology connection, they focus on the quantitative distribution characteristics and one-dimensional linear relationship of the characteristic items of quantitative scientific and technological literature, and cannot quantify the relationship between the scientific system and the technological system in the knowledge system structure, as well as the "input-output" correlation between the two in the innovation generation process. Therefore, how to integrate multi-source heterogeneous data such as scientific and technological literature and industry statistical yearbooks, and measure the "science-technology relationship" comprehensively and accurately from the correlation between the scientific system and the technological system in the dimensions of knowledge system structure, content semantic association, and innovation transformation process has become an important content for further in-depth research. Summary of the invention

[0005] Purpose of the invention: The purpose of the present invention is to provide a method for calculating the multidimensional coupling of science and technology elements, transformation and structure, which solves the problem of missing data in science and technology association detection, low calculation integrity of coupling relationships caused by single linear relationships, and uncertainty in measurement. By integrating multi-source heterogeneous data such as scientific literature, technical patents, and industry statistical yearbooks, the coupling relationship between science and technology is comprehensively and accurately measured from three dimensions: the composition of input-output models of scientific and technological systems, innovation transformation efficiency, and knowledge system.

[0006] Technical solution: A scientific and technological element-conversion-structure multi-dimensional coupling calculation method described in the present invention comprises the following steps:

[0007] (1) Obtain input-output statistical data, scientific literature, and technology patent literature collections in the target emerging industry field and perform preprocessing;

[0008] (2) Based on the data, scientific literature, and patent datasets collected in step (1), data are classified, extracted, and counted based on cities;

[0009] (3) Based on the data, scientific literature, and patent datasets collected in step (1), the language representation model BERT model is used to generate word embeddings for document-level data; the language representation model SciBERT model trained with paper data is used to extract the core knowledge entities of science and technology and construct a science and technology knowledge network;

[0010] (4) Construct a multi-dimensional coupling scheme of elements, transformation and structure to obtain multi-dimensional coupling degree, including calculation of coupling degree of science-technology elements, calculation of coupling degree of science-technology transformation efficiency and calculation of coupling degree of science-technology knowledge network;

[0011] (5) Based on the obtained multi-dimensional coupling degree, a quantitative scheme for the science-technology dynamic coupling type and time-lag relationship is constructed.

[0012] Furthermore, step (1) preprocessing is specifically as follows: cleaning and preprocessing the collected papers and patent data, and finally obtaining field scientific papers and patent data.

[0013] Furthermore, step (2) includes the following steps:

[0014] (21) Extract scientific input indicators: basic scientific research funding, number of research and experimental development personnel, number of research and development institutions; scientific output indicators: number of published papers;

[0015] (22) Extract technology input indicators: basic applied research funds, number of inventors, number of technical service agencies; technology output indicators: number of valid patents.

[0016] Furthermore, step (3) includes the following steps:

[0017] (31) Use BERT to capture the relationship between words and generate word embedding expressions for document words / phrases; extract word vectors for document words / phrases;

[0018] (32) Use the SciBERT model to extract knowledge entities in each scientific and technical document;

[0019] (33) Use the document retrieval algorithm Contriever to calculate the matching score between the knowledge entity combination and the original document, and determine the best knowledge entity combination;

[0020] (34) Based on the final scientific-computing knowledge entity combination data obtained by matching, a scientific and technological knowledge network is constructed using time slicing and entity co-occurrence; the knowledge entity of each document is used as a knowledge network node, and the co-occurrence edge is used as the node connection edge, and the edge is weighted according to the number of co-occurrences.

[0021] Furthermore, in step (4), the calculation of the coupling degree of science and technology elements includes the following steps:

[0022] (S1) Based on the input-output model of the innovation production function in the scientific system or technological system, obtain the model components and measurement indicators; the elements include: scientific and technological research and development input, knowledge stock, human capital, and scientific and technological output;

[0023] (S2) The obtained components are dimensionless processed by using the range standardization method and quantified based on the grey relational model;

[0024] (S3) The degree of geographical coupling of science and technology is calculated based on the average ratio of the geographical concentration of different components of science and technology in a certain region.

[0025] Furthermore, in step (4), the calculation of the coupling degree of science-technology conversion efficiency includes the following steps:

[0026] (S4) Modeling the process of scientific and technological innovation, assuming that there are multiple inputs and outputs in each period of the scientific or technological system, and defining the production possibility set by its feasible vector;

[0027] (S5) Use the output distance function to calculate the input-output relationship from Towards The change in , we get the MRC index reflecting total factor productivity, which indicates the change in R&D productivity;

[0028] (S6) MRC is decomposed into efficiency change MREC and technology change MRTC;

[0029] (S7) Quantify the conversion efficiency coupling of science-technology innovation production through association analysis.

[0030] Furthermore, in step (4), the calculation of the coupling degree of the science-technology knowledge network includes the following steps:

[0031] (A1) Based on the obtained weighted science-technology knowledge network, extract the science-technology knowledge network coupling node set and edge set of each period as the coupling set;

[0032] (A2) Calculate the node degree distribution and edge weight distribution probability of the coupled nodes and coupled edges in the coupling set in the science and technology knowledge network;

[0033] (A3) The coupling degree of the knowledge architecture is calculated using the inverse of the KL divergence of the coupled node degree distribution and the coupled edge weight distribution probability.

[0034] Furthermore, step (5) includes the following steps:

[0035] (51) Classification and identification of the coupling types of science and technology “element-transformation-structure”; specifically, the following: the science and technology “element-transformation-structure” coupling index distributed in time periods obtained in step (4) is used as three time series; the similarity between the three time series (element, transformation), (element, structure), and (transformation, structure) is used to calculate whether the science and technology coupling belongs to the element-transformation coupling type, element-structure coupling type, or transformation-structure coupling type relationship.

[0036] (52) Identification of the time-lag relationship of the coupling of science and technology "element-transformation-structure"; specifically, as follows: Based on the science and technology "element-transformation-structure" coupling index distributed by time period obtained in step (4), calculate the optimal coupling years for element coupling, transformation efficiency coupling, and knowledge structure coupling; use the time difference between the three groups of optimal coupling years (element, transformation), (element, structure), and (transformation, structure) as the lag time of element coupling, transformation efficiency coupling, and knowledge structure coupling to reveal the time lag effect between different coupling characteristics and other coupling characteristics

[0037] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: designing a computing scheme and tool for the multi-dimensional coupling of "element-transformation-structure" of the science and technology system, integrating multi-source heterogeneous data such as scientific literature, technical patents, and industry statistical yearbooks, and comprehensively and accurately measuring the "science-technology relationship" from the correlation between the scientific system and the technology system in terms of knowledge system structure, content semantic association, and innovation transformation process; transforming the original linear relationship according to the complexity of the scientific system and the technology system, and using the three dimensions of input-output model element composition, innovation transformation efficiency, and knowledge system structure to explore the interactive relationship between science and technology, eliminating the one-sidedness and instability of a single measurement relationship; in response to the current problems of low scientific and technological synergy and unclear paths in strategic emerging industries, the present invention starts from the perspective of science-technology coupling of "parallel observation" to explore the effects and mechanisms of science-technology collaborative innovation in the field of emerging industries. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a schematic diagram of the present invention. DETAILED DESCRIPTION

[0039] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0040] like Figure 1 As shown, an embodiment of the present invention provides a scientific and technological element-conversion-structure multi-dimensional coupling calculation method, comprising the following steps:

[0041] S101: Select a typical emerging industry field of science and technology strategy as the target field. The target field is required to have accumulated a long period of basic scientific research and technology application research input-output statistical yearbook data, and to have a large amount of scientific papers and patent literature data. Among them, according to the selected target field, the input-output statistical yearbook data mainly comes from the science and technology statistical yearbooks of various provinces and cities in China. The scientific literature data is collected from the China National Knowledge Infrastructure (CNKI) database, and the search formula is constructed as keywords = ((keyword 1) OR (keyword 2) OR (keyword 2) OR … OR (keyword n)). The academic journals are selected by category, and the publication year is from the creation of the field to the present. Patent data is collected from PATSNAP, and the search formula is constructed according to the international patent classification code (IPC) as IPC = ipc1; ipc2…. The collected paper and patent data are cleaned and preprocessed using the unique identifier ISSN-International Standard Serial Number and patent patnum, and finally the field scientific papers and patent data are obtained.

[0042] S102: Obtain industry-related statistical data from the China Science and Technology Statistical Yearbook, China Statistical Yearbook, and China Regional Innovation Capacity Report for basic data such as industrial development, technology "input-output", social economy, and government support. This includes the following steps:

[0043] S102-1: Based on the selected target areas, extract scientific input indicators: basic scientific research funding (Sfund), number of research and experimental development personnel (Auth), number of research and development institutions (Rorg); scientific output indicators: number of published papers (Pap).

[0044] S102-2: Based on the selected target areas, extract technology input indicators: basic applied research funds (Tfund), number of inventors (Inv), number of technical service agencies (Torg); technology output indicators: number of valid patents (Pat).

[0045] S103: Based on the scientific papers and patent literature data collected and cleaned in S101, the BERT model is used to generate word embeddings for document-level data, and the SciBERT model is used to extract the core knowledge entities of science and technology to build a science and technology knowledge network. The following steps are included:

[0046] S103-1: Use BERT to capture the relationship between words and generate word embedding expressions for document words / phrases.

[0047] S103-2: Extract word vectors for document words / phrases.

[0048] S103-3: Use the SciBERT model to extract knowledge entities in each scientific and technical document.

[0049] S103-4: Use Contriever to calculate the matching score between the query composed of each group of knowledge entities and the original document, and determine the best combination of knowledge entities.

[0050] S103-5: Based on the final science-computing knowledge entity combination data obtained by matching in step S103-4, a scientific knowledge network is constructed using time slices and entity co-occurrence. The knowledge entity of each document is used as a knowledge network node, and the co-occurrence edge is used as the node connection edge. The edge is weighted according to the number of co-occurrences. The scientific knowledge network is represented as ,in Identify the network nodes and edge sets respectively, and the technical network is represented as .

[0051] S104: Based on the processing of step S101, step S102 and step S103, a calculation scheme of the multi-dimensional coupling of science and technology "element-transformation-structure" is constructed.

[0052] S104-1: Calculation of coupling degree of science and technology elements. Details are as follows:

[0053] A: Based on the input-output model of the innovation production function in the scientific system or technological system , and obtain the model's constituent elements, scientific and technological R&D investment ( K )、Human capital( L ), technological level ( A ), scientific and technological output ( Y ) and measurement indicators.

[0054] B: Use the range standardization method The components obtained in the previous step are dimensionless and quantified based on the grey relational model. Indicates the minimum value; Indicates the maximum value.

[0055] C: Based on the geographical concentration of different components of science and technology in a certain area The average ratio is used to calculate the science-technology geographical coupling degree. , They represent the scientific and technological elements of element i, and n represents the type of element.

[0056] S104-2: Calculation of coupling degree of science-technology conversion efficiency. Details are as follows:

[0057] D: Model the process of scientific and technological innovation, assuming that the scientific or technological system will have SInput and M outputs, then the production possibility set can be composed of all feasible vectors Define. Among them, is the input vector, representing the various input resources used in the production process, express It belongs to the non-negative real vector space, with the dimension of input type ; is an output vector, representing the various outputs obtained through the production process, express Belongs to a non-negative real vector space, with dimensions of output types M .

[0058] E: Use output distance function Calculate the input-output relationship from Towards The Malmquist index MRC reflecting the total factor productivity is obtained by calculating the change of ;in, They are Production units at time The input vector and output vector of The input vector and output vector of .

[0059] F: MRC decomposed into efficiency change (MREC) and technological change (MRTC).

[0060] S104-3: Calculation of coupling degree of science-technology knowledge network.

[0061] G: Extract the scientific knowledge network of each period based on the weighted scientific-technical knowledge network obtained in step S103-4 Technical knowledge network The coupled node set and edge sets as a coupled set.

[0062] H: Calculate the node degree distribution of coupled nodes and coupled edges in the science and technology knowledge network in the coupled set , .in , Represent the nodes in the scientific and technological knowledge network The node degree of . The edge weight distribution , .in , Represents nodes in the scientific and technological knowledge network and nodes The weight of the edge.

[0063] I: Coupling of scientific and technical knowledge architecture The inverse of the KL divergence (Kullback-Leibler Divergence) of the coupled node degree distribution and coupled edge weight distribution probability is used for calculation.

[0064] ;

[0065] ;

[0066] ;

[0067] in, Represents a network system of scientific and technological knowledge; Represents the node degree distribution of scientific knowledge network and the node degree distribution of technical knowledge network The KL divergence between Represents the edge weight distribution of scientific knowledge network and edge weight distribution of technical knowledge network The KL divergence between Indicates the number of public nodes, Indicates the number of common edges.

[0068] S105: Based on the calculation results of step S104, a quantification scheme of the science-technology dynamic coupling type and time-lag relationship is constructed. The details are as follows:

[0069] S105-1: Classification and identification of the coupling type of science-technology "element-transformation-structure". Specifically: Based on the science-technology "element-transformation-structure" coupling index distributed by time period obtained in steps S104-1, S104-2, and S104-3, the year is used As the horizontal axis, three coupling time series of element coupling, conversion efficiency coupling and knowledge structure coupling are constructed and recorded as , , ;in, Indicates Geographical concentration of components in the year, Indicates The conversion efficiency coupling degree of the year, Indicates The knowledge structure coupling degree of the year. The three time series of (element, transformation), (element, structure), and (transformation, structure) calculated have the same year, that is, the same length. Therefore, the classic Euclidean distance measurement is used to calculate the difference in science-technology coupling relationship to identify whether it belongs to the element-transformation coupling type, element-structure coupling type, or transformation-structure coupling type. The calculation formula is as follows.

[0070] ;

[0071] ;

[0072] ;

[0073] in, Indicates the largest year , They represent the three coupling time series of science and technology element coupling, conversion efficiency coupling and knowledge structure coupling respectively; It indicates the difference in the relationship between the coupling of science and technology elements and the coupling of conversion efficiency; It indicates the difference in the relationship between coupling of science and technology elements and coupling of knowledge structures; It represents the difference in the relationship between the coupling of science and technology transformation efficiency and the coupling of knowledge structure.

[0074] S105-2: Identification of the time-lag relationship of the coupling of science and technology "element-transformation-structure". Specifically: Based on the science and technology "element-transformation-structure" coupling index distributed by time period obtained in steps S104-1, S104-2, and S104-3, calculate the best coupling year for element coupling, conversion efficiency coupling, and knowledge structure coupling, and record Year ; Calculate the time difference between the three groups of best coupling years (element, transformation), (element, structure), (transformation, structure) ( ), ), ) as the lag time of factor coupling, conversion efficiency coupling and knowledge structure coupling, revealing the time lag effect between different coupling characteristics and other coupling characteristics.

Claims

1. A scientific and technological element-conversion-structure multi-dimensional coupling calculation method, characterized in that: The following steps are involved: (1) Obtain input-output statistical data, scientific literature, and technology patent literature collections in the target emerging industry field and perform preprocessing; (2) Based on the data, scientific literature, and patent datasets collected in step (1), data are classified, extracted, and counted based on cities; (3) Based on the data collected in step (1), scientific literature, and patent datasets, use the language representation model BERT model to generate word embeddings for document-level data; Use the SciBERT model, a language representation model trained with paper data, to extract the core knowledge entities of science and technology and build a science and technology knowledge network; (4) Constructing a multi-dimensional coupling scheme of elements, transformation and structure to obtain a multi-dimensional coupling degree, including calculation of the coupling degree of science-technology elements, calculation of the coupling degree of science-technology transformation efficiency, and calculation of the coupling degree of science-technology knowledge network. The calculation of the coupling degree of science-technology elements includes the following steps: (S1) Based on the input-output model of the innovation production function in the scientific system or technological system, obtain the model components and measurement indicators; the elements include: scientific and technological research and development input, knowledge stock, human capital, and scientific and technological output; (S2) The obtained components are dimensionless processed by using the range standardization method and quantified based on the grey relational model; (S3) Calculate the science-technology geographical coupling degree based on the average ratio of the geographical concentration of different components of science and technology in a certain region; The calculation of the coupling degree of science-technology conversion efficiency includes the following steps: (S4) Modeling the process of scientific and technological innovation, assuming that there are multiple inputs and outputs in each period of the scientific or technological system, and defining the production possibility set by its feasible vector; (S5) Use the output distance function to calculate the input-output relationship from Towards The change in , we get the MRC index reflecting total factor productivity, which indicates the change in R&D productivity; (S6) MRC is decomposed into efficiency change MREC and technology change MRTC; (S7) Quantify the conversion efficiency coupling of science-technology innovation production through association analysis; The calculation of the coupling degree of science-technology knowledge network includes the following steps: (A1) Based on the obtained weighted science-technology knowledge network, extract the science-technology knowledge network coupling node set and edge set of each period as the coupling set; (A2) Calculate the node degree distribution and edge weight distribution probability of the coupled nodes and coupled edges in the coupling set in the science and technology knowledge network; (A3) The coupling degree of the knowledge architecture is calculated using the inverse of the KL divergence of the coupling node degree distribution and the coupling edge weight distribution probability; (5) Based on the obtained multidimensional coupling degree, a quantitative scheme for the science-technology dynamic coupling type and time-lag relationship is constructed; the following steps are included: (51) Classification and identification of the coupling type of science-technology "element-transformation-structure"; specifically, the following: the science-technology "element-transformation-structure" coupling index distributed by time period obtained in step (4) is used as three time series; the similarity between the three time series (element, transformation), (element, structure), (transformation, structure) is used to calculate whether the science-technology coupling belongs to the element-transformation coupling type, element-structure coupling type, or transformation-structure coupling type relationship; (52) Identification of the time-lag relationship of the coupling of science and technology "element-transformation-structure"; specifically, as follows: Based on the science and technology "element-transformation-structure" coupling index distributed by time period obtained in step (4), calculate the optimal coupling years for element coupling, transformation efficiency coupling, and knowledge structure coupling; use the time difference between the three groups of optimal coupling years (element, transformation), (element, structure), and (transformation, structure) as the lag time of element coupling, transformation efficiency coupling, and knowledge structure coupling to reveal the time lag effect between different coupling characteristics and other coupling characteristics.

2. A scientific and technological element-conversion-structure multi-dimensional coupling calculation method according to claim 1, characterized in that: Step (1) preprocessing is specifically: cleaning and preprocessing the collected papers and patent data to finally obtain scientific papers and patent data in the field.

3. A scientific and technological element-conversion-structure multi-dimensional coupling calculation method according to claim 1, characterized in that: Step (2) includes the following steps: (21) Extract scientific input indicators: basic scientific research funding, number of research and experimental development personnel, number of research and development institutions; scientific output indicators: number of published papers; (22) Extract technology input indicators: basic applied research funds, number of inventors, number of technical service agencies; technology output indicators: number of valid patents.

4. A scientific and technological element-conversion-structure multi-dimensional coupling calculation method according to claim 1, characterized in that: Step (3) includes the following steps: (31) Use BERT to capture the relationship between words and generate word embedding expressions for document words / phrases; extract word vectors for document words / phrases; (32) Use the SciBERT model to extract knowledge entities in each scientific and technical document; (33) Use the document retrieval algorithm Contriever to calculate the matching score between the knowledge entity combination and the original document, and determine the best knowledge entity combination; (34) Based on the final scientific-computing knowledge entity combination data obtained by matching, a scientific and technological knowledge network is constructed using time slicing and entity co-occurrence; the knowledge entity of each document is used as a knowledge network node, and the co-occurrence edge is used as the node connection edge, and the edge is weighted according to the number of co-occurrences.