Network academic resource association selection system

Abstract

The invention discloses a network academic resource association selection system, and relates to the technical field of network academic resource management, and the system is characterized in that a semantic vector construction module converts data in academic resources into semantic vectors with similarity comparison through a text preprocessing method, a TF-IDF vectorization method and a potential semantic analysis method; the text similarity module calculates the similarity degree of the resources and the target document by using a cosine similarity algorithm; the network construction module is responsible for analyzing a cooperation relationship between researchers and institutions in academic resources to form an academic cooperation network; the cooperation network association module optimizes resource selection in combination with the academic cooperation network and the academic influence of researchers; the resource selection module is used for sorting and selecting resources by integrating the text similarity and the influence of the academic cooperation network; and the monitoring feedback module tracks dynamic updating and user feedback of academic resources in real time. According to the method, the limitation of a traditional academic resource screening method is effectively solved, and more accurate and efficient academic resource recommendation is provided.

Description

Technical Field

[0001] The present invention relates to the technical field of network academic resource management, and in particular to a network academic resource association selection system. Background Art

[0002] With the rapid development of academic research, the number and types of academic resources are increasing. How to efficiently screen and recommend academic resources related to specific research topics has become a major challenge for researchers. Existing academic resource retrieval systems usually screen documents based on keyword matching or citation analysis. Although these methods can provide relevant resources to a certain extent, they often fail to capture the deep semantic relationship between texts and the comprehensive influence of academic resources. With the increasing academic cooperation, the selection of academic resources not only depends on the similarity of the content of the documents, but also needs to consider the cooperative relationship and academic influence between researchers.

[0003] However, existing academic resource screening technologies have some obvious shortcomings. First, traditional text similarity analysis methods mainly rely on keyword matching and surface text overlap, which cannot effectively capture the potential semantic connections between documents. Second, the existing system's analysis of academic collaboration networks is relatively simple, usually only based on a single citation data to evaluate the influence of academic resources, lacking in-depth mining of complex collaborative relationships between researchers and institutions. In addition, existing technologies lack a comprehensive assessment of the influence of academic resources and researchers, resulting in the screened resources sometimes being unable to fully reflect the latest progress and core influence in the academic field.

[0004] In view of the above problems, the present invention proposes a network academic resource association selection system based on multiple technical means. The system effectively solves the shortcomings of traditional methods in text similarity analysis and cooperation network construction through semantic vector construction, text similarity analysis, academic cooperation network construction and academic influence evaluation methods. By combining text content and cooperation network information, the present invention can more accurately screen out resources that are highly relevant to the target topic and have high academic value, thereby improving the overall performance of the academic resource recommendation system. Summary of the invention

[0005] In view of the above problems, the present invention provides a network academic resource association selection system to solve the deficiencies of the prior art in text similarity analysis and collaborative network construction.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a network academic resource association selection system, comprising: A semantic vector construction module, which uses text preprocessing, TF-IDF vectorization and latent semantic analysis technology to convert the original text data in the academic resources into a semantic vector form that can be used for similarity comparison, and provides the constructed semantic vector to the text similarity module for subsequent calculation; A text similarity module, which compares the semantic vectors of different academic resources and the target document by using cosine similarity calculation and threshold selection to obtain a similarity score, passes the similarity score to the resource selection module for sorting, and feeds back the similarity score to the monitoring feedback module; A network construction module, which is responsible for analyzing the cooperation relationship between researchers and institutions in the academic resources and passing the cooperation relationship information to the cooperation network association module; A collaboration network association module, which analyzes the academic influence of academic resources and researchers, obtains academic influence information through a core node identification method, and provides the academic influence information to the resource selection module; A resource selection module, which performs weighted ranking of academic resources according to the similarity scores of the academic resources and the academic influence of the academic cooperation network, and feeds the selected academic resources back to the monitoring feedback module; Monitoring and feedback module, which is responsible for monitoring the latest status of the system and dynamically updating the academic collaboration network and similarity measurement by continuously tracking academic publications, research collaboration dynamics and user feedback. This module interacts with the text similarity module, collaboration network association module and resource selection module to improve the system's recommendation strategy and weighted ranking through feedback loops.

[0007] The functions of the semantic vector construction module include: text preprocessing, TF-IDF vectorization, and latent semantic analysis; The text preprocessing includes text tokenization, removal of stop words, and standardization of content. By uniformly preprocessing all documents, it is ensured that common words and formats do not distort the analysis, wherein the standardization content includes converting uppercase and lowercase letters and removing special symbols. The TF-IDF vectorization converts each document into a high-dimensional vector using the TF-IDF weighting scheme, thereby creating a document-term matrix in which each entry reflects the importance of the word in the document relative to the entire corpus. Words that appear frequently in the document but are rare in the entire corpus will receive higher weights, thereby strengthening the expression of unique terms. The latent semantic analysis uses singular value decomposition technology to decompose the original high-dimensional TF-IDF matrix into three matrices, namely U, Σ, and V, where U is a left singular vector, that is, each column is a feature vector of a text; Σ is a diagonal matrix containing all potential semantic concepts; V is a right singular vector, that is, each row is a feature vector of a word; In the dimensionality reduction process, the first k largest singular values ​​and their corresponding singular vectors are selected to reduce the dimension and retain the most important semantic information.

[0008] The functions of the text similarity module include: cosine similarity calculation and threshold selection; The cosine similarity calculation calculates the cosine similarity between the vector of a given target resource and the vectors of all other documents in the corpus. The cosine similarity score ranges from 0 to 1, where 1 indicates that the documents are exactly the same in the vector space and 0 indicates that they have almost no similarity in the space. The specific calculation of cosine similarity is shown in the formula: Among them, A and B represent the vector representation of two documents respectively. represents the dot product of vectors, ||A|| and ||B|| represent the respective vector moduli; The threshold selection is performed by setting a similarity threshold. Any document with a similarity score higher than the threshold is considered to be closely related to the target topic and is retained, while documents with a similarity score lower than the threshold are filtered out. The threshold is selected by observing the natural gap in the distribution of similarity scores and identifying the natural dividing point by hierarchical clustering of the similarity scores. A similarity score threshold is set at this dividing point as the boundary for judging the relevance of the document to the target topic.

[0009] The functions of the network construction module include: data collection, collaborative graph construction, and edge weighting; Said data collection,is done by collecting data on authors’ co-authors and affiliations from a corpus of,academic sources; The collaboration graph construction uses the collected data to construct a graphical representation of an academic collaboration network, wherein the academic collaboration network refers to a network constructed by analyzing researchers, institutions and the collaboration relationships between them in academic resources, wherein nodes represent researchers and institutions, and edges represent the collaboration relationships between them; The edge weighting is performed by weighting the edge according to the number of collaborations, the number of co-authored papers, and the number of citation impacts.

[0010] The functions of the cooperation network association module include: core node identification, community analysis, and resource selection; The core node identification analyzes the academic collaboration network, identifies the most important and influential nodes in the collaboration graph, and calculates the centrality of the nodes to evaluate the importance of each node in the academic collaboration network. The centrality includes the following three types: Degree centrality: indicates how many direct collaborators a researcher has collaborative relationships with, measuring the breadth of their collaboration in the academic collaboration network; Betweenness centrality: indicates that the researcher is in the intermediary position of the shortest collaboration path between other researchers, reflecting his / her bridging role in the academic collaboration network; Closeness centrality: represents the average distance between a researcher and other researchers, measuring their closeness in the academic collaboration network; Based on the calculated centrality values, nodes are divided into high centrality and low centrality, with high centrality nodes highlighting the “hub” individuals and institutions that are at the core of the academic collaboration network; The community analysis uses the Louvain method to identify communities and clusters of researchers who frequently collaborate, and optimizes modularity to identify groups of nodes that collaborate closely in the academic collaboration network, revealing closely connected research teams and expert groups in sub-fields. In this way, the structure of the academic collaboration network is analyzed and key nodes and intensive collaboration groups within the academic community are identified, where key nodes usually include influential researchers and institutions. The resource selection selects high-impact academic resources based on the results of academic collaboration network analysis. The academic collaboration network analysis evaluates the academic influence in the academic collaboration network by identifying key nodes and intensive collaboration groups in the academic collaboration network, and selects the resources most closely associated with core collaborators by combining insights into the academic collaboration network with academic resources. High-impact academic resources generally include publications written by core nodes in the academic collaboration network, and the core nodes refer to researchers and institutions with high centrality values ​​in the academic collaboration network.

[0011] The functions of the resource selection module include: resource ranking and final selection; The resource ranking is to sort the academic resources according to two major factors, specifically: Text similarity score, by calculating the text similarity between each resource and the target document, a similarity score is assigned to each resource. The higher the similarity score, the more relevant the content of the resource is to the target document; Network centrality, by calculating the centrality score of the nodes associated with the resource in the academic cooperation network, reflects the academic influence and importance of the resource. The higher the centrality, the more important the position of the node in the academic cooperation network; The resource ranking will be based on the combination of these two factors. First, the resources will be initially screened based on text similarity, and then they will be weighted and sorted based on the centrality of the academic cooperation network, ensuring that not only the content-related resources are retained, but also the resources with high academic influence are given priority. The final selection is doubly screened based on the resources' text similarity scores and academic collaboration network centrality scores.

[0012] The functions of the monitoring and feedback module include: continuous monitoring and feedback loop; The continuous monitoring function dynamically updates the academic cooperation network and similarity measurement by tracking the latest publications and cooperation relationships. Specifically, it regularly tracks the latest publications, research results and ongoing cooperation in the academic field, obtains literature updates in real time through the integrated academic resource library, and identifies new research trends and key papers; dynamically updates the nodes and cooperation relationships of the academic cooperation network based on the latest publications and cooperation data, and uses the newly collected data to update the text similarity measurement model, so that the relevance and similarity scores of resources in the academic cooperation network are more accurate. The text similarity measurement model is based on the cosine similarity calculation method used in the text similarity module. The model is used to evaluate the similarity between academic resources and target documents during the dynamic update process; The feedback loop continuously improves the resource recommendation and ranking algorithm of the system through the interaction between users and the system. Specifically, feedback is collected from the interaction between users and recommended resources. Users evaluate the resources recommended by the system. The evaluation includes marking the relevance, quality and actual application effect of the resources. Based on user feedback, the selection and ranking algorithms of the recommendation system are adjusted. Specifically, user feedback is used as a reward signal in reinforcement learning to optimize the recommendation strategy. User-labeled feedback is used to train the supervised learning model. The supervised learning model analyzes and processes user feedback. The system fine-tunes the recommendation strategy in a short period of time and applies the adjusted algorithm to future resource selection and sorting.

[0013] Compared with the prior art, the present invention has the following beneficial effects: The present invention adopts a semantic vector construction module combined with latent semantic analysis technology, so that text similarity analysis not only focuses on the surface overlap of vocabulary, but also deeply explores the potential semantic relationship between texts. Even if two documents do not use the same keywords, as long as they are similar at the semantic level, they can be correctly identified as related, thereby improving the accuracy and comprehensiveness of academic resource screening.

[0014] By dynamically constructing and analyzing academic cooperation networks, the present invention can reflect the cooperation relationships and changes between researchers and institutions in real time. The cooperation network not only relies on the number of citations, but also combines the cooperation frequency and influence of researchers, providing a more comprehensive and accurate academic influence assessment.

[0015] The present invention conducts a comprehensive evaluation of resources by combining text similarity and influence analysis of the collaboration network. This comprehensive ranking method ensures that not only content-related documents are selected, but also resources with higher academic influence are given priority, thereby enhancing the academic value and influence of the recommendation results.

[0016] Through the continuous monitoring module and feedback loop mechanism, the present invention can track the latest academic resources and cooperation dynamics in real time and automatically optimize the resource recommendation strategy. User feedback, as a key part of the optimization algorithm, can help the system continuously improve the accuracy of resource screening and ranking, further enhancing the system's adaptability and long-term effectiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It is understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 It is a system architecture diagram of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the invention claimed for protection, but is only for selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention.

[0020] Please refer to Figure 1 Schematic diagram of a network academic resource association selection system provided by an embodiment of the present invention, including: A semantic vector construction module, which uses text preprocessing, TF-IDF vectorization and latent semantic analysis technology to convert the original text data in the academic resources into a semantic vector form that can be used for similarity comparison, and provides the constructed semantic vector to the text similarity module for subsequent calculation; A text similarity module, which compares the semantic vectors of different academic resources and the target document by using cosine similarity calculation and threshold selection to obtain a similarity score, passes the similarity score to the resource selection module for sorting, and feeds back the similarity score to the monitoring feedback module; A network construction module, which is responsible for analyzing the cooperation relationship between researchers and institutions in the academic resources and passing the cooperation relationship information to the cooperation network association module; A collaboration network association module, which analyzes the academic influence of academic resources and researchers, obtains academic influence information through a core node identification method, and provides the academic influence information to the resource selection module; A resource selection module, which performs weighted ranking of academic resources according to the similarity scores of the academic resources and the academic influence of the academic cooperation network, and feeds the selected academic resources back to the monitoring feedback module; Monitoring and feedback module, which is responsible for monitoring the latest status of the system and dynamically updating the academic collaboration network and similarity measurement by continuously tracking academic publications, research collaboration dynamics and user feedback. This module interacts with the text similarity module, collaboration network association module and resource selection module to improve the system's recommendation strategy and weighted ranking through feedback loops.

[0021] The functions of the semantic vector construction module include: text preprocessing, TF-IDF vectorization, and latent semantic analysis; The text preprocessing includes text tokenization, removal of stop words, and standardization of content. By uniformly preprocessing all documents, it is ensured that common words or formats do not distort the analysis, wherein the standardization content includes converting uppercase and lowercase letters and removing special symbols, and the special symbols include punctuation marks, number symbols, and currency symbols; The TF-IDF vectorization converts each document into a high-dimensional vector using the TF-IDF weighting scheme, thereby creating a document-term matrix in which each entry reflects the importance of the word in the document relative to the entire corpus. Words that appear frequently in the document but are rare in the entire corpus will receive higher weights, thereby strengthening the expression of unique terms. The latent semantic analysis uses singular value decomposition technology to decompose the original high-dimensional TF-IDF matrix into three matrices, namely U, Σ, and V, where U is a left singular vector, that is, each column is a feature vector of a text; Σ is a diagonal matrix containing all potential semantic concepts; V is a right singular vector, that is, each row is a feature vector of a word; During the dimensionality reduction process, the top k largest singular values ​​and their corresponding singular vectors are selected to reduce the dimension and retain the most important semantic information. Specifically, semantically similar texts will be close to each other in the latent semantic space after dimensionality reduction. Even if they do not use the same keywords, they can be correctly identified as related as long as they are similar at the semantic level. It is easier to obtain conceptual similarity when comparing documents through vector space without being misled by superficial term mismatches.

[0022] The functions of the text similarity module include: cosine similarity calculation and threshold selection; The cosine similarity calculation calculates the cosine similarity between the vector of a given target resource and the vectors of all other documents in the corpus. The cosine similarity score ranges from 0 to 1, where 1 indicates that the documents are exactly the same in the vector space and 0 indicates that they have almost no similarity in the space. The specific calculation of cosine similarity is shown in the formula: Among them, A and B represent the vector representation of two documents respectively. represents the dot product of vectors, ||A|| and ||B|| represent the respective vector moduli; The threshold selection is performed by setting a similarity threshold. Any document with a similarity score higher than the threshold is considered to be closely related to the target topic and is retained, while documents with a similarity score lower than the threshold are filtered out. The threshold is selected by observing the natural gap in the distribution of similarity scores and identifying the natural dividing point by hierarchical clustering of the similarity scores. A similarity score threshold is set at this dividing point as the boundary for judging the relevance of the document to the target topic.

[0023] The functions of the network construction module include: data collection, collaborative graph construction, and edge weighting; The data collection is done by collecting data about the co-authors and affiliations of authors from a corpus of academic resources. Specifically, for a set of publications in a given field, a list of researchers involved in each publication is extracted and their institutions are recorded. Each academic resource provides metadata that can be used as raw data for constructing an academic collaboration network. The metadata includes author names, co-author lists, and author affiliations. The collaboration graph construction uses the collected data to construct a graphical representation of an academic collaboration network, wherein the academic collaboration network refers to a network constructed by analyzing researchers, institutions and the collaboration relationships between them in academic resources, wherein nodes represent researchers and institutions, and edges represent the collaboration relationships between them; The edge weighting is achieved by weighting the edge according to the number of collaborations, the number of co-authored papers, and the number of citation impacts; specifically, the edge weight can be calculated in the following way: for each collaboration, the edge weight can be measured according to the number of co-authored papers; for the weighting of citation impact, it can be weighted according to the number of citations of the cited document and the journal impact factor.

[0024] It should be noted that if two researchers collaborate on one or more papers, an advantage will be formed between them. The advantage can connect the institutions that jointly participate in the research project. The academic collaboration network can be enriched by the weights on the edges. Specifically, the weight can be the number of papers jointly published by the two researchers, or the impact of their collaboration can be considered (such as adding the number of citations of the papers they co-authored). Higher weights mean that these nodes have stronger or more important collaborative links. The result of this step is a network graph of co-authors / collaborations, which contains the structure of the academic collaboration in the dataset.

[0025] In graph terms, it is usually an undirected graph of co-authors, which may be bipartite if we distinguish researcher-to-paper relationships before projection, but ultimately we consider an author-author network for our analysis.

[0026] The functions of the cooperation network association module include: core node identification, community analysis, and resource selection; The core node identification analyzes the academic collaboration network, identifies the most important and influential nodes in the collaboration graph, and calculates the centrality of the nodes to evaluate the importance of each node in the academic collaboration network. The centrality includes the following three types: Degree centrality: indicates how many direct collaborators a researcher has collaborative relationships with, measuring the breadth of their collaboration in the academic collaboration network; Betweenness centrality: indicates that the researcher is in the intermediary position of the shortest collaboration path between other researchers, reflecting his / her bridging role in the academic collaboration network; Closeness centrality: represents the average distance between a researcher and other researchers, measuring their closeness in the academic collaboration network; According to the calculated centrality values, nodes are divided into high centrality and low centrality. High centrality nodes refer to researchers and institutions with high connectivity and central positions in the academic cooperation network. The centrality values ​​of these nodes are in the top 30% of all nodes or reach a certain preset threshold. Low centrality nodes refer to nodes with fewer connections and peripheral positions in the network. Their centrality values ​​are lower than the preset threshold or ranked in the bottom 30%. The community analysis identifies communities and clusters of researchers who frequently collaborate through the Louvain method, which is an algorithm for network community detection. The Louvain method optimizes modularity to identify groups of nodes that collaborate closely in the academic collaboration network, revealing closely connected research teams or expert groups in sub-fields. In this way, the structure of the academic collaboration network is analyzed and key nodes and intensive collaboration groups within the academic community are identified, where key nodes usually include influential researchers and institutions. The resource selection selects high-impact academic resources based on the results of academic collaboration network analysis. The academic collaboration network analysis evaluates the academic influence in the academic collaboration network by identifying key nodes and intensive collaboration groups in the academic collaboration network, and selects the resources most closely associated with core collaborators by combining insights into the academic collaboration network with academic resources. High-impact academic resources generally include publications written by core nodes in the academic collaboration network. The core nodes refer to researchers and institutions with high centrality values ​​in the academic collaboration network. These nodes generally play a key role in the collaboration network, connecting multiple other researchers and institutions, and have extensive academic influence.

[0027] It should be noted that by selecting resources connected to these core network nodes, the system not only has relevant academic resources (because academic collaboration networks are usually formed around related research fields), but also has high influence and authority. This method can find resources that are likely to be recommended because "important people work on them", complementing the content similarity method, which can find resources because "they discuss similar things".

[0028] If the text similarity process described earlier can be seen as content-based filtering, then the collaborative network process is a form of social filtering based on academic relationships, which helps to discover important papers that might be missed by pure content search. For example, a foundational paper in the field, but using different terminology, might be missed by text similarity, but if it was written by a leading researcher in a field, it would be picked up here.

[0029] It should be noted that if certain researchers have a high centrality in the network or dominate collaborative clusters, the work they produce or participate in will often be of higher quality and influence. Specifically, the citation impact of these publications is positively correlated with the network centrality of the authors. Papers co-authored by researchers with extensive connections and influence usually accumulate more citations and recognitions, indicating that they have higher academic influence.

[0030] The functions of the resource selection module include: resource ranking and final selection; The resource ranking is to sort the academic resources according to two major factors, specifically: Text similarity score, by calculating the text similarity between each resource and the target document, a similarity score is assigned to each resource. The higher the similarity score, the more relevant the content of the resource is to the target document; Network centrality, by calculating the centrality score of the nodes associated with the resource in the academic cooperation network, reflects the academic influence and importance of the resource. The higher the centrality, the more important the position of the node in the academic cooperation network; The resource ranking will be based on the combination of these two factors. First, the resources will be initially screened based on text similarity, and then they will be weighted and sorted based on the centrality of the academic cooperation network, ensuring that not only the content-related resources are retained, but also the resources with high academic influence are given priority. The final selection is double-screened based on the text similarity score and academic collaboration network centrality score of the resources; Resources with high text similarity and network centrality. If a resource scores high in both text similarity and network centrality, it is considered to be the most relevant and academically influential resource and is selected first. In addition, resources related to influential collaborators in the academic network are given priority. These resources usually reflect the latest progress in the field or have high academic value. After comprehensive screening of these two aspects, academic resources that are highly relevant to the target topic and have high academic influence are finally selected. The field refers to the relevant research field in the academic cooperation network.

[0031] The functions of the monitoring and feedback module include: continuous monitoring and feedback loop; The continuous monitoring function dynamically updates the academic cooperation network and similarity measurement by tracking the latest publications and cooperation relationships. Specifically, it regularly tracks the latest publications, research results and ongoing cooperation in the academic field, obtains literature updates in real time through the integrated academic resource library, and identifies new research trends and key papers; dynamically updates the nodes and cooperation relationships of the academic cooperation network based on the latest publications and cooperation data, and uses the newly collected data to update the text similarity measurement model, so that the relevance and similarity scores of resources in the academic cooperation network are more accurate; the text similarity measurement model is based on the cosine similarity calculation method used in the text similarity module, and the model is used to evaluate the similarity between academic resources and target documents during the dynamic update process; The feedback loop continuously improves the resource recommendation and ranking algorithm of the system through the interaction between users and the system. Specifically, feedback is collected from the interaction between users and recommended resources. Users evaluate the resources recommended by the system. The evaluation includes marking the relevance, quality and actual application effect of the resources. According to user feedback, the selection and ranking algorithms of the recommendation system are adjusted. Specifically, user feedback is used as a reward signal in reinforcement learning to optimize the recommendation strategy. User-labeled feedback is used to train the supervised learning model. The supervised learning model analyzes and processes user feedback. The system fine-tunes the recommendation strategy in a short period of time and applies the adjusted algorithm to future resource selection and sorting. The supervised learning model is based on the random forest algorithm. The supervised learning model optimizes the recommendation strategy and weighted ranking through supervised learning based on the interaction feedback between the user and the system and the user's evaluation of the recommended resources.

[0032] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention has various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A network academic resource association selection system, characterized in that: include: A semantic vector construction module, which uses text preprocessing, TF-IDF vectorization and latent semantic analysis technology to convert the original text data in the academic resources into a semantic vector form that can be used for similarity comparison, and provides the constructed semantic vector to the text similarity module for subsequent calculation; A text similarity module, which compares the semantic vectors of different academic resources and the target document by using cosine similarity calculation and threshold selection to obtain a similarity score, passes the similarity score to the resource selection module for sorting, and feeds back the similarity score to the monitoring feedback module; A network construction module, which is responsible for analyzing the cooperation relationship between researchers and institutions in academic resources and passing the cooperation relationship information to the cooperation network association module; A collaboration network association module, which analyzes the academic influence of academic resources and researchers, obtains academic influence information through a core node identification method, and provides the academic influence information to the resource selection module; A resource selection module, which performs weighted ranking of academic resources according to the similarity scores of the academic resources and the academic influence of the academic cooperation network, and feeds the selected academic resources back to the monitoring feedback module; Monitoring and feedback module, which is responsible for monitoring the latest status of the system and dynamically updating the academic collaboration network and similarity measurement by continuously tracking academic publications, research collaboration dynamics and user feedback. This module interacts with the text similarity module, collaboration network association module and resource selection module to improve the system's recommendation strategy and weighted ranking through feedback loops.

2. According to claim 1, a network academic resource association selection system is characterized by: The functions of the semantic vector construction module include: text preprocessing, TF-IDF vectorization, and latent semantic analysis; The text preprocessing includes text tokenization, removal of stop words, and standardization of content. All documents are uniformly preprocessed to ensure that common words and formats do not distort the analysis, wherein the standardization content includes conversion of uppercase and lowercase letters and removal of special symbols. The TF-IDF vectorization converts each document into a high-dimensional vector using the TF-IDF weighting scheme, thereby creating a document-term matrix in which each entry reflects the importance of the word in the document relative to the entire corpus. Words that appear frequently in the document but are rare in the entire corpus will receive higher weights, thereby strengthening the expression of unique terms. The latent semantic analysis uses singular value decomposition technology to decompose the original high-dimensional TF-IDF matrix into three matrices, namely U, Σ, and V, where U is a left singular vector, that is, each column is a feature vector of a text; Σ is a diagonal matrix containing all potential semantic concepts; V is a right singular vector, that is, each row is a feature vector of a word; In the dimensionality reduction process, the first k largest singular values ​​and their corresponding singular vectors are selected to reduce the dimension and retain the most important semantic information.

3. The network academic resource association selection system according to claim 1, characterized in that: The functions of the text similarity module include: cosine similarity calculation and threshold selection; The cosine similarity calculation calculates the cosine similarity between the vector of a given target resource and the vectors of all other documents in the corpus. The cosine similarity score ranges from 0 to 1, where 1 indicates that the documents are exactly the same in the vector space and 0 indicates that they have almost no similarity in the space. The specific calculation of cosine similarity is shown in the formula: Among them, A and B represent the vector representation of two documents respectively. represents the dot product of vectors, ||A|| and ||B|| represent the respective vector moduli; The threshold selection is performed by setting a similarity threshold. Any document with a similarity score higher than the threshold is considered to be closely related to the target topic and is retained, while documents with a similarity score lower than the threshold are filtered out. The threshold is selected by observing the natural gap in the distribution of similarity scores and identifying the natural dividing point by hierarchical clustering of the similarity scores. A similarity score threshold is set at this dividing point as the boundary for judging the relevance of the document to the target topic.

4. The network academic resource association selection system according to claim 1, characterized in that: The functions of the network construction module include: data collection, collaborative graph construction, and edge weighting; Said data collection,is done by collecting data on authors’ co-authors and affiliations from a corpus of,academic sources; The collaboration graph construction uses the collected data to construct a graphical representation of an academic collaboration network, wherein the academic collaboration network refers to a network constructed by analyzing researchers, institutions and the collaboration relationships between them in academic resources, wherein nodes represent researchers and institutions, and edges represent the collaboration relationships between them; The edge weighting is performed by weighting the edge according to the number of collaborations, the number of co-authored papers, and the number of citation impacts.

5. The network academic resource association selection system according to claim 1, characterized in that: The functions of the cooperation network association module include: core node identification, community analysis, and resource selection; The core node identification analyzes the academic collaboration network, identifies the most important and influential nodes in the collaboration graph, and calculates the centrality of the nodes to evaluate the importance of each node in the academic collaboration network. The centrality includes the following three types: Degree centrality: indicates how many direct collaborators a researcher has collaborative relationships with, measuring the breadth of their collaboration in the academic collaboration network; Betweenness centrality: indicates that the researcher is in the intermediary position of the shortest collaboration path between other researchers, reflecting his / her bridging role in the academic collaboration network; Closeness centrality: represents the average distance between a researcher and other researchers, measuring their closeness in the academic collaboration network; Based on the calculated centrality values, nodes are divided into high centrality and low centrality. High centrality nodes highlight the "central" individuals and institutions that are at the core of the academic cooperation network; The community analysis uses the Louvain method to identify communities and clusters of researchers who frequently collaborate, and optimizes modularity to identify groups of nodes that collaborate closely in the academic collaboration network, revealing closely connected research teams and expert groups in sub-fields. In this way, the structure of the academic collaboration network is analyzed and key nodes and intensive collaboration groups within the academic community are identified, where key nodes usually include influential researchers and institutions. The resource selection selects high-impact academic resources based on the results of academic collaboration network analysis. The academic collaboration network analysis evaluates the academic influence in the academic collaboration network by identifying key nodes and intensive collaboration groups in the academic collaboration network, and selects the resources most closely associated with core collaborators by combining insights into the academic collaboration network with academic resources. High-impact academic resources generally include publications written by core nodes in the academic collaboration network, and the core nodes refer to researchers and institutions with high centrality values ​​in the academic collaboration network.

6. The network academic resource association selection system according to claim 1, characterized in that: The functions of the resource selection module include: resource ranking and final selection; The resource ranking is to sort the academic resources according to two major factors, specifically: Text similarity score, by calculating the text similarity between each resource and the target document, a similarity score is assigned to each resource. The higher the similarity score, the more relevant the content of the resource is to the target document. Network centrality, by calculating the centrality score of the nodes associated with the resource in the academic cooperation network, reflects the academic influence and importance of the resource. The higher the centrality, the more important the position of the node in the academic cooperation network; The resource ranking will be based on the combination of these two factors. First, the resources will be initially screened based on text similarity, and then they will be weighted and sorted based on the centrality of the academic cooperation network, ensuring that not only the relevant content resources are retained, but also the resources with high academic influence are given priority. The final selection is doubly screened based on the resources' text similarity scores and academic collaboration network centrality scores.

7. The network academic resource association selection system according to claim 1, characterized in that: The functions of the monitoring and feedback module include: continuous monitoring and feedback loop; The continuous monitoring function dynamically updates the academic cooperation network and similarity measurement by tracking the latest publications and cooperation relationships. Specifically, it regularly tracks the latest publications, research results and ongoing cooperation in the academic field, obtains literature updates in real time through the integrated academic resource library, and identifies new research trends and key papers; dynamically updates the nodes and cooperation relationships of the academic cooperation network based on the latest publications and cooperation data, and uses the newly collected data to update the text similarity measurement model, so that the relevance and similarity scores of resources in the academic cooperation network are more accurate. The text similarity measurement model is based on the cosine similarity calculation method used in the text similarity module. The model is used to evaluate the similarity between academic resources and target documents during the dynamic update process; The feedback loop continuously improves the resource recommendation and ranking algorithm of the system through the interaction between users and the system. Specifically, feedback is collected from the interaction between users and recommended resources. Users evaluate the resources recommended by the system. The evaluation includes marking the relevance, quality and actual application effect of the resources. Based on user feedback, the selection and ranking algorithms of the recommendation system are adjusted. Specifically, user feedback is used as a reward signal in reinforcement learning to optimize the recommendation strategy. User-labeled feedback is used to train the supervised learning model. The supervised learning model analyzes and processes user feedback. The system fine-tunes the recommendation strategy in a short period of time and applies the adjusted algorithm to future resource selection and sorting.

Images