A keyword-based measure of interdisciplinary

Abstract

The application relates to an interdisciplinary field, in particular to a keyword-based interdisciplinary measurement method. The measurement method mainly analyzes from three dimensions: semantic analysis obtains a matrix and a vector sequence through two ways of processing a corpus, and cosine calculation is performed on the matrix and the vector sequence to obtain an index reflecting the similarity of discipline knowledge; hierarchical structure analysis is based on hierarchical structure analysis to determine the contribution of theme keywords and quantize knowledge distance, and the two are calculated to obtain a cross-disciplinary measurement index; network structure analysis is to obtain the influence index of nodes in the network through the study of keyword network, and the index is applied to the multi-disciplines to calculate the cross-disciplinary strength measurement index based on the network structure; finally, the three dimensions are combined by an entropy weighting method to evaluate the cross-disciplinary interaction strength; the method solves the problems of insufficient granularity and lack of wholeness in the current interdisciplinary detection.

Description

Technical Field

[0001] This invention relates to the field of interdisciplinary studies, and in particular to a keyword-based method for measuring interdisciplinary studies. Background Technology

[0002] Interdisciplinary research, involving the integration of knowledge from multiple disciplines, is an indispensable means of addressing complex contemporary societal challenges and driving scientific innovation. It is also a key factor in creating high-impact knowledge and practical innovations. With the rapid development of science and technology, the solutions and breakthroughs needed to solve today's complex scientific problems often transcend the limitations of a single discipline, requiring interdisciplinary research that integrates theories and technologies from multiple disciplines. According to 2017 statistics, 49.07% of Nobel Prize-winning research was based on interdisciplinary collaboration. Therefore, exploring the nature, forms, and laws of interdisciplinary research is of great significance for promoting scientific development and social progress.

[0003] A survey of domestic and international literature on interdisciplinary measurement reveals that existing research primarily employs two methods: analyzing the academic backgrounds of researchers or the collaborative patterns among authors, or examining publications and their citations. These studies utilize scholar information surveys and mapping of institutional or journal subject categories to derive the specific degree of interdisciplinary integration under different interdisciplinary measurement indicators. For example, Liu et al., in *Science, Technology and Dialectics*, 6, 105-109, revealed the decisive role of interdisciplinary research in original scientific achievements by statistically analyzing the number of Nobel Prize laureates in Physiology or Medicine who obtained degrees outside their own field. Zhang et al., in *Journal of Information Science*, 37(3), 231-242, used the research results of highly cited scholars in the social sciences from 2014 to 2016 as an example to measure the degree of interdisciplinary integration between different disciplinary institutions in scientific collaboration, demonstrating that interdisciplinary institutional collaboration has a certain promoting effect on the output of interdisciplinary scientific achievements. Wang et al., in Scientometrics, 111, 2023-2039, used 60 prolific authors in the fields of applied mathematics, computational science, operations research and management science from the WOS database as examples to measure their interdisciplinary degree based on target literature, references and citing literature.

[0004] However, most current research, including those mentioned above, primarily focuses on interdisciplinary collaboration at the macro level, while research on interdisciplinary knowledge structures at the micro level is relatively limited. At the micro level, keywords in a research paper are considered the most basic carriers of knowledge, directly and simply reflecting the paper's ideas and themes. Compared to subject-level or theme-level methods, keywords offer a more granular perspective when studying the knowledge connections between interdisciplinary fields and their constituent disciplines. A collection of subject-specific keywords encompasses the main research focus of a particular discipline. Therefore, cross-disciplinary keywords from two or more disciplines can reveal the content of interdisciplinary research. This concept forms the basis for this study's choice to explore interdisciplinary collaboration using keywords as a guide, thereby delving into the subtle interactions and knowledge structures that emerge at this granular level. Summary of the Invention

[0005] The purpose of this invention is to overcome the problems of insufficient granularity of research subjects and defects in the comprehensiveness of measurement methods in existing interdisciplinary measurement technologies, and to provide a keyword-based interdisciplinary measurement method.

[0006] In a first aspect, the present invention provides a calculation method for multi-index synthesis, which integrates indices obtained based on the semantic analysis, the hierarchical structure analysis, and the network structure analysis using an entropy weighting method, including the following steps:

[0007] S1: Extract keywords from academic papers;

[0008] S2: By performing semantic analysis on the obtained subject paper keyword dataset, we obtain indicators reflecting the similarity of subject knowledge; by analyzing the hierarchical structure of the obtained subject paper keywords, we obtain indicators measuring the interdisciplinary nature between different subjects; and by analyzing the network structure of the obtained subject paper keywords, we obtain indicators measuring the interdisciplinary strength.

[0009] S3: The semantic analysis index, the hierarchical structure analysis index, and the network structure analysis index are integrated using the entropy weighting method. The greater the dispersion of the resulting index, the smaller the entropy value, and the greater the impact of the index on the comprehensive evaluation.

[0010] Preferably, step S1 includes the following steps:

[0011] S11: Delete meaningless words from the titles and abstracts of academic papers based on the stop word list;

[0012] S12: Use the processed data as input to the Yake Keyword Extractor, a natural language processing tool, and extract noun phrases from it.

[0013] Preferably, the semantic analysis in step S2 further includes the step of calculating the semantic analysis index using cosine similarity.

[0014] Preferably, the semantic analysis involves processing the compiled corpus to obtain a matrix and a vector sequence, and then performing cosine calculation on both to obtain the semantic analysis index.

[0015] More preferably, the semantic analysis includes the following steps:

[0016] Step A: Filter keywords that appear below a specified threshold in the titles and abstracts of papers in the relevant subject area;

[0017] Step B: Use the term frequency-inverse document frequency algorithm to calculate the term frequency-inverse document frequency of each keyword in documents of different disciplines, thereby constructing a discipline-key term matrix; input the corpus into a Word2Vec model for training, and use the CBOW method to convert the discipline keywords in the document data into points in a low-dimensional vector space to generate a keyword vector sequence of a specified dimension;

[0018] Step C: Perform L2 normalization on both the matrix and the vector, and multiply the normalized subject keyword matrix with the keyword vector to obtain the subject keyword vector;

[0019] Step D: Use cosine similarity to calculate the similarity between subject keyword vectors to quantify the semantic relationships between subjects. The calculation result is denoted as P. Semantic , representing the semantic similarity of subject knowledge between subject X and subject Y, that is, reflecting the degree of cross-disciplinary interaction between subject X and subject Y based on document content.

[0020]

[0021] Preferably, the hierarchical structure analysis in step S2 includes: calculating the hierarchical structure analysis index based on the TaxoGen model.

[0022] Preferably, the hierarchical structure analysis is based on the analysis of the hierarchical structure to determine the contribution of the topic keywords and quantify the knowledge distance, and the hierarchical structure analysis index is obtained by calculating the two.

[0023] More preferably, the hierarchical analysis includes the following steps:

[0024] Step A: Construct a hierarchical structure based on the TaxoGen model;

[0025] Step B: By analyzing the representativeness of keywords to each level of themes and the probability of keyword allocation within a theme, a hierarchical theme-keyword contribution can be defined. Combining the subject-theme and corresponding theme-keyword allocation probabilities, the contribution of each subject's keywords to each subject-level theme can be determined. Therefore, the subject contribution Ctk of keyword k under theme t can be derived:

[0026] C tk =D t ×f tk

[0027] In the formula: Dt represents the document frequency of topic t, and ftk represents the frequency of keyword k in the subject corpus under topic t;

[0028] Step C: By matching different topics at different levels based on the same keyword in two subject-specific topic trees, the sum of their level depths represents the path of interdisciplinary knowledge interaction and dissemination. The greater the distance between two concepts, the lower the scope and similarity of the subject knowledge. Knowledge distance can be quantified by the following formula:

[0029]

[0030] In the formula: KD is the quantified value of the knowledge distance, wtk is the keyword k under a certain subject topic t, f is the transfer function of d and l, l is the shortest path length between them, and d is the depth of the included in the hierarchical semantic network, that is, the depth of the lowest common ancestor between two keywords.

[0031] Step D: A metric was constructed based on the contribution of topic keywords and knowledge distance. This hierarchical structure analysis metric can be calculated based on the hierarchical structure, using the following formula:

[0032]

[0033] In the formula: P Hierarchy This refers to the hierarchical structure analysis index, where KD is the quantified value of the knowledge distance, and n(X,Y) is the number of common keywords between discipline X and discipline Y. It is the number of paths that can be formed in the knowledge association process based on common keywords, where T represents the number of keywords in a subject unit's theme.

[0034] Preferably, the network structure analysis in step S2 further includes: obtaining the network structure analysis index based on the subject co-word network calculation.

[0035] Preferably, the network structure analysis is performed by studying the keyword network to obtain the influence index of nodes within the network, and then applying this index to multiple disciplines to calculate the network structure analysis index based on the network structure.

[0036] More preferably, the network structure analysis includes the following steps:

[0037] Step A: Extract common keywords between discipline X and discipline Y to construct a set of mediating entities with the same keywords, V. k ;

[0038] V k =V X ∩V Y

[0039] In the formula: V X V represents the set of keywords for subject X. Y This represents the set of keywords for subject Y.

[0040] Step B: Based on the betweenness centrality of the keyword nodes in the computation set obtained in C1, the importance of the nodes is used to calculate the index Bet(v), which is used to determine whether node v has high influence as a keyword in the discipline; the formula for calculating Bet(v) is:

[0041]

[0042] In the formula: Bet(v) is an indicator for judging whether node v has a high influence in the discipline, σ st σ is the total number of shortest paths from node s to node t. st (v) is the number of paths that pass through node v;

[0043] Step C: Combining the characteristics of points and edges as endogenous variables in the interdisciplinary network, the indicator L, used to measure the influence of a node within network A, can be calculated using a formula. A (v), the calculation formula is:

[0044]

[0045] In the formula: L A This represents the metric for measuring the influence of a node within network A. Bet(v) is the metric for determining whether node v has high influence in the subject. Γ(v) represents the set of neighboring nodes of node v. Link(v,p) is the co-occurrence strength of node v and its neighboring node p.

[0046] Step D: Taking into account factors between node pairs and structural information about node embedding within the overall network, calculate the interdisciplinary strength measurement index P based on network structure. network (X,Y), this indicator measures the relative importance of shared keywords as bridge nodes for interdisciplinary interaction in the knowledge networks of disciplines X and Y. The calculation formula is as follows:

[0047]

[0048] In the formula: P Network L is the network structure analysis metric. X (v) represents a metric that measures the node's influence within the X network, L Y (v) represents a metric that measures the node's influence within the Y network. X V Y These represent keyword nodes in subject X and subject Y, respectively.

[0049] Preferably, step S3 specifically includes the following steps:

[0050] S31: Normalization index, which converts the absolute value of the index into a relative value to solve the homogeneity of heterogeneous indicators; considering that the values ​​of positive and negative indicators represent opposite meanings, different algorithms are used for data normalization;

[0051] If the j-th indicator is a positive indicator:

[0052]

[0053] If the j-th indicator is a negative indicator:

[0054]

[0055] In the formula: x' ij These are the semantic analysis indicators, the hierarchical structure analysis indicators, and the network structure analysis indicators that need to be normalized, where i is a natural number used to measure the number of disciplines, and j is a natural number used to measure the number of indicators.

[0056] S32: Calculate the proportion of the i-th subject to the j-th indicator:

[0057]

[0058] In the formula: P ij x' is the proportion of the i-th subject to the j-th indicator. ij These are the semantic analysis indicators, hierarchical structure analysis indicators, and network structure analysis indicators that require indicator normalization, where i is a natural number used to measure the number of disciplines, and j is a natural number used to measure the number of indicators.

[0059] S33: Calculate the entropy of the j-th index, where e j ≥0:

[0060]

[0061] In the formula: e j It is the entropy of the j-th index, P ij It is the proportion of the i-th discipline to the j-th indicator, where i is a natural number used to measure the number of disciplines and j is a natural number used to measure the number of indicators.

[0062] S34: Calculate information entropy redundancy:

[0063] d j =1-e j

[0064] In the formula: d j The information entropy redundancy of the j-th indicator is e. j It is the entropy of the j-th indicator, where j is a natural number used to measure the number of indicators;

[0065] S35: Calculate the weight of each indicator:

[0066]

[0067] In the formula: w j It is the weight of the j-th indicator, d j This is the information entropy redundancy of the j-th indicator, where j is a natural number used to measure the number of indicators;

[0068] S36: The hierarchical structure analysis index is calculated as follows:

[0069] P Interdisciplinarity (X,Y)=α*P Semantic (X,Y)+β*P Hierarchy (X,Y)+(1-α-β)*P Network (X,Y)

[0070] In the formula: P Interdisciplinarity These are the hierarchical structure analysis indicators, where X and Y represent discipline X and discipline Y, respectively, and P... Semantic It is the semantic analysis metric, P Hierarchy It is the hierarchical structure analysis index, P Network The network structure analysis index is α and β, which are probability ratio adjustment factors. The weights calculated by S35 are used for setting, where α = w1 and β = w2.

[0071] Preferably, the interdisciplinary measurement method is based on the three dimensions of semantic analysis, hierarchical structure analysis, and network structure analysis, which solves the problems of insufficient fine-grainedness and lack of comprehensiveness of existing interdisciplinary measurement methods, and also helps to better explore the subtle interaction and deep interaction mechanisms between disciplines.

[0072] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0073] This invention provides a keyword-based interdisciplinary measurement method. Through three-dimensional analysis and calculation, it obtains three levels of indicators for measuring the intensity of interdisciplinary interaction from the perspectives of semantic analysis, hierarchical structure analysis, and network structure analysis. The three indicators are then integrated using an entropy weighting method to obtain a comprehensive measurement index. This method enables more granular and comprehensive research on interdisciplinary measurement, and overcomes the limitations of existing interdisciplinary measurement methods that only study from a macro perspective. It has significant promotional significance and application prospects. Attached Figure Description

[0074] Figure 1 This is a flowchart illustrating the method steps of the interdisciplinary measurement and integration framework in this invention;

[0075] Figure 2 This is a schematic diagram of the hierarchical knowledge distance construction method in the hierarchical structure analysis of this invention;

[0076] Figure 3 This is a schematic diagram of the method for constructing a network graph of shared keywords in network analysis in this invention;

[0077] Figure 4 It is a comprehensive measurement of the intensity of interdisciplinary interaction between the discipline of "Library and Information Science" and six other disciplines in three periods;

[0078] Figure 5 This is a comparison of the ranking results of various indicators from 2005 to 2009;

[0079] Figure 6 This is a comparison of the ranking results of various indicators from 2010 to 2014;

[0080] Figure 7 This is a comparison of the ranking results of various indicators from 2015 to 2019; Detailed Implementation

[0081] The present invention will be further described in detail below with reference to experimental examples and specific embodiments. However, this should not be construed as limiting the scope of the above-mentioned subject matter of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.

[0082] Example 1

[0083] This study takes interdisciplinary research in seven specific disciplines—Library and Information Science, Economics, Computer Science and Information Systems, Education and Educational Research, Geography, Social Sciences and Interdisciplinary Studies, and Management—as an example. A total of 209,062 English-language research articles from 2005 to 2019 were crawled from the Web of Science (WOS) SCI-EXPANDED and SSCI databases. The method focuses on measuring interdisciplinary research within Library and Information Science, examining its interdisciplinary strength with the other six disciplines. To better understand the evolving trends in interdisciplinary research, the dataset is divided into three periods: 2005-2009, 2010-2014, and 2015-2019.

[0084] A keyword-based interdisciplinary measurement method, flowchart as follows: Figure 1 As shown, it includes the following steps:

[0085] S1: Extract keywords from academic papers;

[0086] S2: From the perspective of semantic analysis, the titles and abstracts of papers from various disciplines are integrated into a corpus. The corpus is then processed using algorithms and models to train the corpus, resulting in a subject keyword matrix and keyword vectors. Finally, the semantic analysis index is obtained through cosine similarity calculation. From the perspective of hierarchical structure analysis, a hierarchical structure is constructed to obtain the contribution of topic keywords and knowledge distance, and the hierarchical structure analysis index is built based on these factors. From the perspective of network structure analysis, a subject-specific co-word network is constructed, and network structure analysis is performed by combining the features of points and edges to obtain the network structure analysis index.

[0087] S3: The entropy weighting method is used to integrate the semantic analysis indicators obtained from semantic analysis, the hierarchical structure analysis indicators obtained from hierarchical structure analysis, and the network structure analysis indicators obtained from network structure analysis. The greater the dispersion of the resulting indicators, the smaller the entropy value, and the greater the influence of the indicators on the comprehensive evaluation.

[0088] Step S1 in the measurement method, which processes the acquired abstracts and titles, includes the following steps:

[0089] S11: Delete meaningless words from the titles and abstracts of academic papers based on the stop word list;

[0090] S12: Use the processed data as input to the Yake Keyword Extractor, a natural language processing tool, and extract noun phrases from it.

[0091] The entropy weighting method described in step S3 is used to integrate the indicators obtained from the three-dimensional analysis and calculate the final indicator for measuring the intensity of disciplinary interaction, including the following steps:

[0092] S31: Normalization index, which converts the absolute value of the index into a relative value to solve the homogeneity of heterogeneous indicators; considering that the values ​​of positive and negative indicators represent opposite meanings, different algorithms are used for data normalization;

[0093] If the j-th indicator is a positive indicator:

[0094]

[0095] If the j-th indicator is a negative indicator:

[0096]

[0097] In the formula: x' ij These are the semantic analysis indicators, the hierarchical structure analysis indicators, and the network structure analysis indicators that need to be normalized, where i is a natural number used to measure the number of disciplines, and j is a natural number used to measure the number of indicators.

[0098] S32: Calculate the proportion of the i-th subject to the j-th indicator:

[0099]

[0100] In the formula: P ij x' is the proportion of the i-th subject to the j-th indicator. ij These are the semantic analysis indicators, hierarchical structure analysis indicators, and network structure analysis indicators that require indicator normalization, where i is a natural number used to measure the number of disciplines, and j is a natural number used to measure the number of indicators.

[0101] S33: Calculate the entropy of the j-th index, where e j ≥0:

[0102]

[0103] In the formula: e j It is the entropy of the j-th index, P ij It is the proportion of the i-th discipline to the j-th indicator, where i is a natural number used to measure the number of disciplines and j is a natural number used to measure the number of indicators.

[0104] S34: Calculate information entropy redundancy:

[0105] d j =1-ej

[0106] In the formula: d j The information entropy redundancy of the j-th indicator is e. j It is the entropy of the j-th indicator, where j is a natural number used to measure the number of indicators;

[0107] S35: Calculate the weight of each indicator:

[0108]

[0109] In the formula: w j It is the weight of the j-th indicator, d j This is the information entropy redundancy of the j-th indicator, where j is a natural number used to measure the number of indicators;

[0110] S36: The hierarchical structure analysis index is calculated as follows:

[0111] P Interdisciplinarity (X,Y)=α*P Semantic (X,Y)+β*P Hierarchy (X,Y)+(1-α-β)*P Network (X,Y)

[0112] In the formula: P Interdisciplinarity These are the hierarchical structure analysis indicators, where X and Y represent discipline X and discipline Y, respectively, and P... Semantic It is the semantic analysis metric, P Hierarchy It is the hierarchical structure analysis index, P Network The network structure analysis index is α and β, which are probability ratio adjustment factors. The weights calculated by S35 are used for setting, where α = w1 and β = w2.

[0113] The comprehensive interdisciplinary interaction intensity scores for each discipline across the three periods, calculated using S36 in the entropy weighting method, are as follows: Figure 4 As shown. The specific calculation formula is:

[0114] Composite score=-0.081+0.215*Semantic_based+0.416*Hierarchy_based+

[0115] 0.378 * Network_based

[0116] Four mainstream interdisciplinary indicators were selected from the perspectives of knowledge network topological characteristics and knowledge integration content characteristics. These were compared with the comprehensive results of the keyword-based measurement method used in this invention. These four indicators are Rao-Stirling, TI value, overlap, and eigenvector centrality. Detailed information is as follows: Figures 5-7 As shown, Figures 5-7 The main focus is on showcasing the value and ranking of various methods in interdisciplinary measurement, thereby demonstrating the strengths and weaknesses of different approaches. A comprehensive comparative analysis shows that the measurement method based on the Rao-Stirling index effectively reveals interdisciplinary differences, but in certain contexts, its accuracy in reflecting actual interdisciplinary interactions is slightly inferior. Measurement methods based on the TI value, overlap, and eigenvector centrality perform poorly in generating interdisciplinary rankings, deviating significantly from the expected evaluation results. Figures 5-7 Judging from the results and rankings, compared with the three measurement indicators used in the currently commonly used measurement methods, the present invention is superior to the current scheme in terms of comprehensive data and ranking. The interdisciplinary measurement method adopted shows higher accuracy in measuring interdisciplinary interaction and has a stronger explanatory power for the actual situation.

[0117] 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, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A keyword-based interdisciplinary measurement method, characterized in that, Includes the following steps: S1: Extract keywords from subject-specific papers and construct a subject-specific paper keyword dataset; S2: By performing semantic analysis on the subject paper keyword dataset, an index reflecting the similarity of subject knowledge is obtained, and it is defined as a semantic analysis index; by performing hierarchical structure analysis on the subject paper keyword dataset, an index measuring the interdisciplinary nature between different disciplines is obtained, and it is defined as a hierarchical structure analysis index; by performing network structure analysis on the subject paper keyword dataset, an interdisciplinary strength measurement index is obtained, and it is defined as a network structure analysis index. S3: The semantic analysis index, the hierarchical structure analysis index, and the network structure analysis index are integrated using the entropy weighting method, and the output is a cross-disciplinary interaction intensity measurement result; The semantic analysis involves processing the subject-specific paper keyword dataset to obtain a matrix and a vector sequence, and then calculating the cosine similarity between the two to obtain the semantic analysis index. The semantic analysis includes the following steps: Step A: Filter keywords that appear below a specified threshold in the titles and abstracts of papers in the relevant discipline; Step B: Use the term frequency-inverse document frequency algorithm to calculate the term frequency-inverse document frequency of each keyword in documents of different disciplines, thereby constructing a discipline-keyword matrix; input the keywords into a Word2Vec model for training, and use the CBOW method to convert the discipline keywords in the document data into points in a low-dimensional vector space to generate a keyword vector sequence of a specified dimension; Step C: Perform L2 normalization on both the matrix and the vector, and multiply the normalized subject keyword matrix with the keyword vector to obtain the subject keyword vector; Step D: Use cosine similarity to calculate the similarity between subject keyword vectors to quantify the semantic relationships between subjects. The calculation result is denoted as P. Semantic This represents the semantic similarity of subject knowledge between subject X and subject Y, reflecting the degree of interdisciplinary interaction between subject X and subject Y based on document content. In the formula: P Semantic (X,Y) is the semantic analysis index, where X and Y represent different disciplines, i is a natural number used to measure the number of disciplines, and n is the total number of disciplines to be calculated, which is a natural number. The hierarchical structure analysis is based on analyzing the hierarchical structure to determine the contribution of topic keywords and quantify knowledge distance, and the hierarchical structure analysis index is calculated from both. The hierarchical structure analysis includes the following steps: Step A: Construct a hierarchical structure based on the TaxoGen model; Step B: By analyzing the representativeness of keywords to each level of the topic and the probability of keyword allocation within the topic, a hierarchical topic-keyword contribution can be defined. Combining the subject-topic and corresponding topic-keyword allocation probabilities, the contribution of each subject's keywords to each subject-level topic can be determined. Therefore, the subject contribution Ctk of keyword k under topic t can be derived: In the formula: D t f represents the document frequency of topic t. tk This indicates the frequency of keyword k in the subject corpus under topic t; Step C: By matching different topics at different levels based on the same keyword, the sum of the hierarchical depths of the matching results represents the path of interdisciplinary knowledge interaction and dissemination. The greater the distance between two concepts, the lower the scope and similarity of the disciplinary knowledge. Knowledge distance can be quantified by the following formula: In the formula: KD is the quantified value of the knowledge distance. Let k be a keyword under a subject topic t, i be a natural number that measures the number of subject topics t, and f be the sum of d and d. The transfer function, is the shortest path length between them, and d is the depth of the constituents in the hierarchical semantic network, that is, the depth of the lowest common ancestor between the two keywords; Step D: A metric was constructed based on the contribution of topic keywords and knowledge distance. This hierarchical structure analysis metric can be calculated based on the hierarchical structure, using the following formula: In the formula: P Hierarchy (X,Y) are the hierarchical structure analysis indicators. This represents the subject contribution of keyword k under subject topic t, where i is a natural number measuring the number of subject topics t, KD is the quantified value of the knowledge distance, and n(X,Y) is the number of common keywords between subject X and subject Y. It is the number of possible paths that can be formed in the knowledge association process based on common keywords, where T represents the number of keywords in a subject unit's theme; The network structure analysis involves studying keyword networks to obtain the influence index of nodes within the network, applying this influence index to multiple disciplines, and calculating the network structure analysis index based on the network structure. The network structure analysis includes the following steps: Step A: Extract common keywords between discipline X and discipline Y to construct a set of mediating entities with the same keywords, V. k ; In the formula: V X V represents the set of keywords for subject X. Y This represents the set of keywords for subject Y. Step B: Calculate the betweenness centrality of the keyword nodes in the set within the network as the node's importance. The resulting index, Bet(v), is used to determine whether node v is a highly influential keyword in the subject. The formula for Bet(v) is: In the formula: Bet(v) is an indicator for judging whether node v has a high influence in the discipline, σ st From node s The total number of shortest paths to node t, σ st (v) is the number of paths that pass through node v; Step C: Combining the characteristics of points and edges as endogenous variables in the interdisciplinary network, the indicator L, used to measure the influence of a node within network A, can be calculated using a formula. A (v), the calculation formula is: In the formula, L A This represents the metric for measuring the influence of a node within network A. Bet(v) is the metric for determining whether node v has high influence in the subject. Γ(v) represents the set of neighboring nodes of node v. Link(v,p) is the co-occurrence strength of node v and its neighboring node p. Step D: Taking into account factors between node pairs and the structural information of nodes embedded in the overall network, calculate the network structure analysis index based on the network structure. This index measures the relative importance of shared keywords as bridge nodes for interdisciplinary interaction in the knowledge networks of disciplines X and Y. The calculation formula is as follows: In the formula: P Network L is the network structure analysis metric. X (v) represents a metric that measures the node's influence within the X network, L Y (v) represents a metric that measures the node's influence within the Y network. X V Y These represent keyword nodes in subject X and subject Y, respectively.

2. The keyword-based interdisciplinary measurement method according to claim 1, characterized in that, Step S1 specifically includes the following steps: S11: When extracting keywords from academic papers, first delete meaningless words in the title and abstract of the academic papers according to the preset stop word list; S12: Use the data processed in S11 as input to the Yake keyword extractor, a natural language processing tool, and extract noun phrases from the data processed in S11.

3. The measurement method according to claim 1, characterized in that, Step S3 specifically includes the following steps: S31: Indicator normalization, converting the absolute values ​​of the semantic analysis indicators, the hierarchical structure analysis indicators, and the network structure analysis indicators into relative values ​​to address the homogeneity of heterogeneous indicators; considering that the values ​​of positive and negative indicators represent opposite meanings, different algorithms are used for indicator normalization; If the j-th indicator is a positive indicator: If the j-th indicator is a negative indicator: In the formula: These are the semantic analysis indicators, the hierarchical structure analysis indicators, and the network structure analysis indicators that need to be normalized, where i is a natural number used to measure the number of disciplines, and j is a natural number used to measure the number of indicators. S32: Calculate the proportion of the i-th subject to the j-th indicator: In the formula: P ij It is the proportion of the i-th subject to the j-th indicator. These are the semantic analysis indicators, hierarchical structure analysis indicators, and network structure analysis indicators that require indicator normalization, where i is a natural number used to measure the number of disciplines, and j is a natural number used to measure the number of indicators. S33: Calculate the entropy of the j-th index, where e j ≥0; In the formula: e j It is the entropy of the j-th index. P ij It is the proportion of the i-th discipline to the j-th indicator, where i is a natural number used to measure the number of disciplines and j is a natural number used to measure the number of indicators. S34: Calculate information entropy redundancy: d j =1-e j In the formula: d j The information entropy redundancy of the j-th indicator is e. j It is the entropy of the j-th indicator, where j is a natural number used to measure the number of indicators; S35: Calculate the weight of each indicator: In the formula: w j It is the weight of the j-th indicator, d j This is the information entropy redundancy of the j-th indicator, where j is a natural number used to measure the number of indicators; S36: The results of the interdisciplinary interaction intensity measurement are calculated as follows: In the formula: P Interdisciplinarity This refers to the measurement results of the interdisciplinary interaction strength, where X and Y represent discipline X and discipline Y, respectively, and P... Semantic It is the semantic analysis metric, P Hierarchy It is the hierarchical structure analysis index, P Network The network structure analysis index is α and β, which are probability ratio adjustment factors. The weights calculated by S35 are used for setting, where α=w1 and β=w2.

4. A measurement method, characterized in that, It includes at least one integration method and three analysis methods connected to the at least one integration method; the integration method is based on the three analysis methods, which are processed by the at least one integration method to enable the at least one integration method to perform the method of any one of claims 1 to 3.

Images