Intelligent typesetting and management method and system for journal papers

Abstract

The invention discloses a journal paper intelligent typesetting and management method and system, and belongs to the technical field of management, and the method comprises the steps: constructing a journal typesetting parameter knowledge base, and carrying out the structural representation and semantic description of the typesetting requirements of various periodicals through the ontology and semantic web technology, and obtaining a standardized journal typesetting parameter ontology library; introducing a graph-based reference link analysis method for the recognized structured paper content, and converting a reference relationship in the paper into a semantic link graph; the content is uniquely identified through a Hash algorithm, and rapid retrieval and updating are achieved.

Description

Technical Field

[0001] The present invention belongs to the field of management technology, and in particular relates to a journal paper intelligent typesetting and management method and system. Background Art

[0002] The intelligent typesetting and management system for journal articles faces many technical challenges in practical applications. First, the typesetting requirements of different journals vary greatly. Each journal has its own specific column settings, font size selection, line spacing control, etc. This requires the system to be able to flexibly adapt to the typesetting parameters of various journals and automatically generate corresponding typesetting styles based on the parameters. Secondly, the content of journal articles involves multiple elements such as text, charts, formulas, etc., and these elements have different display methods during typesetting. How to accurately identify these elements and give them appropriate treatment during the typesetting process is a major technical challenge.

[0003] Furthermore, the content of journal articles also involves a large number of citations and references. How to automatically identify these citation links during typesetting and generate a standardized reference index is also a big problem. Finally, due to the large number of journal articles and the diversity of content, the speed and stability of the typesetting system are very high. How to ensure the efficiency and quality of typesetting while making the system efficient and robust is another difficult problem facing developers. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a journal article intelligent typesetting and management method and system to solve the problems existing in the above-mentioned prior art.

[0005] To achieve the above-mentioned purpose, the present invention provides a journal paper intelligent typesetting and management method, comprising:

[0006] Use ontology and semantic web technology to perform structured representation and semantic description of the typesetting requirements of various journals, and build a journal typesetting parameter database based on the description;

[0007] Generate a typesetting parameter configuration file based on the journal typesetting parameter database, and build a recognition rule library based on the multimodal content recognition method and the typesetting parameter configuration file;

[0008] Obtaining a journal article to be identified, identifying the journal article to be identified based on the identification rule base, and generating structured article content;

[0009] The graph-based citation link analysis method converts the citation relationship in the structured paper content into a semantic link graph;

[0010] The semantic connection graph is uniquely identified based on a hash algorithm, and journal papers are retrieved and managed based on the unique identification.

[0011] Preferably, the process of constructing a journal typesetting parameter database based on the description includes:

[0012] Acquiring typesetting requirement data of a journal, wherein the typesetting requirement data of the journal includes typesetting parameters and rules;

[0013] Performing structured description and semantic description processing based on the typesetting requirement data of the journal to obtain structured data of the typesetting requirement of the journal;

[0014] A journal typesetting parameter database is constructed based on the structured data of the journal typesetting requirements.

[0015] Preferably, the process of generating a typesetting parameter configuration file based on the journal typesetting parameter database includes:

[0016] According to the journal typesetting parameter ontology library, obtaining typesetting parameters matching the target journal;

[0017] A corresponding typesetting parameter configuration file is generated for the typesetting parameters that match the target journal.

[0018] Preferably, the process of constructing a recognition rule library based on the multimodal content recognition method and the typesetting parameter configuration file includes:

[0019] Obtain historical typeset papers, and parse the papers based on the corresponding typeset parameter configuration files to obtain different types of content;

[0020] The different types of content are identified based on processing technology to build the identification rule library.

[0021] Preferably, the process of identifying the different types of content based on the processing technology and constructing the identification rule base includes:

[0022] Recognize the text content of the different types of content based on natural language processing technology and build a text recognition rule library;

[0023] Recognize the chart contents of the different types of content based on computer vision technology and build a chart recognition rule library;

[0024] Identify the formula contents of the different types of content based on mathematical formula recognition technology, and build a formula recognition rule library;

[0025] The text recognition rule base, the chart recognition rule base and the formula recognition rule base are integrated to generate the multi-modal recognition rule base.

[0026] Preferably, the process of generating structured paper content includes:

[0027] Obtaining a journal article to be identified, identifying the journal article to be identified based on the identification rule base, and generating an identification result;

[0028] The journal article to be identified is identified based on the typesetting parameters corresponding to the identification result, and the structured article content is generated.

[0029] Preferably, the process of converting the citation relationship in the structured paper content into a semantic link graph by the graph-based citation link analysis method includes:

[0030] Preprocessing the structured paper content to obtain citation information of the paper;

[0031] Constructing a citation relationship graph of the paper content based on the citation information of the paper;

[0032] Processing the reference relationship graph of the paper content based on a community discovery algorithm to obtain a paper community with close semantic correlation;

[0033] For each paper community, extract keywords that can represent the semantic features of the community and construct a semantic vector representation of the paper community;

[0034] The semantic link graph is constructed based on the semantic vector representation of the paper community.

[0035] Preferably, the semantic connection graph is uniquely identified based on a hash algorithm, and the process of retrieving and managing journal papers based on the unique identification includes:

[0036] Acquire key information based on the semantic link graph and construct a content feature vector;

[0037] Based on the content feature vector, a unique identification code of the content is calculated using a hash algorithm as an index key of the cache;

[0038] Search the layout cache to see if there is a layout result that matches the current content identification code. If so, directly return the cached result. At the same time, establish a mapping relationship between the layout result and the content identification code and store it in the layout cache.

[0039] To achieve the above purpose, the present invention also provides a journal paper intelligent typesetting and management system, including:

[0040] A parameter database construction module is used to use ontology and semantic web technology to perform structured representation and semantic description of the typesetting requirements of various journals, and to construct a journal typesetting parameter database based on the description;

[0041] An identification rule base construction module is used to generate a typesetting parameter configuration file based on the journal typesetting parameter database, and to construct an identification rule base based on a multimodal content recognition method and the typesetting parameter configuration file;

[0042] A structured generation module, used for obtaining journal articles to be identified, identifying the journal articles to be identified based on the identification rule base, and generating structured article content;

[0043] A semantic link graph generation module is used to convert the citation relationship in the structured paper content into a semantic link graph based on a graph-based citation link analysis method;

[0044] The retrieval management module is used to uniquely identify the semantic connection graph based on a hash algorithm, and to retrieve and manage journal papers based on the unique identification.

[0045] Compared with the prior art, the present invention has the following advantages and technical effects:

[0046] The present invention discloses an intelligent journal typesetting system. The system constructs a journal typesetting parameter knowledge base, and uses ontology and semantic web technology to perform structured representation of typesetting requirements. Based on this, a typesetting parameter configuration file is generated, and a recognition rule base is constructed using a multimodal content recognition method. Through reference link analysis, the paper reference relationship is converted into a semantic link graph. The present invention optimizes the memory management and task scheduling of the typesetting engine, and uses multi-threaded parallel computing to improve efficiency. The present invention realizes the intelligence and automation of journal typesetting, greatly improves typesetting efficiency and quality, and provides strong support for academic publishing. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0048] Figure 1 This is a flow chart of the intelligent typesetting and management method for journal articles according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0050] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0051] Embodiment 1

[0052] like Figure 1As shown, this embodiment provides a journal paper intelligent typesetting and management method, including:

[0053] Use ontology and semantic web technology to perform structured representation and semantic description of the typesetting requirements of various journals, and build a journal typesetting parameter database based on the description;

[0054] Generate a typesetting parameter configuration file based on the journal typesetting parameter database, and build a recognition rule library based on the multimodal content recognition method and the typesetting parameter configuration file;

[0055] Obtaining a journal article to be identified, identifying the journal article to be identified based on the identification rule base, and generating structured article content;

[0056] The graph-based citation link analysis method converts the citation relationship in the structured paper content into a semantic link graph;

[0057] The semantic connection graph is uniquely identified based on a hash algorithm, and journal papers are retrieved and managed based on the unique identification.

[0058] To further optimize the scheme, the process of constructing a journal typesetting parameter database based on the description includes:

[0059] Acquiring typesetting requirement data of a journal, wherein the typesetting requirement data of the journal includes typesetting parameters and rules;

[0060] Performing structured description and semantic description processing based on the typesetting requirement data of the journal to obtain structured data of the typesetting requirement of the journal;

[0061] A journal typesetting parameter database is constructed based on the structured data of the journal typesetting requirements.

[0062] Acquire journal typesetting requirement data, the journal typesetting requirement data including typesetting parameters and rules; perform structured description and semantic description processing on the journal typesetting requirement data to obtain journal typesetting requirement structured data that is easy for a machine to understand; construct a journal typesetting body library based on the journal typesetting requirement structured data, the journal typesetting body library including standardized typesetting parameters and rules;

[0063] Specifically, when obtaining journal typesetting requirement data, crawler technology can be used to automatically extract relevant typesetting parameters and rules from the journal's official website, such as paper format, font size, line spacing, page margins, etc. For the extracted unstructured typesetting requirement data, natural language processing technologies such as named entity recognition and relationship extraction are used to convert it into a structured JSON or XML format, and semantic annotation is performed, such as marking "small four" as fontSize, to facilitate machine understanding and processing. On this basis, a journal typesetting entity library is constructed, and ontology description languages ​​such as OWL are used to define the core concepts, attributes and their relationships in the field of journal typesetting, so as to achieve standardized representation of typesetting parameters and rules.

[0064] Further optimizing the scheme, the process of generating a typesetting parameter configuration file based on the journal typesetting parameter database includes:

[0065] According to the journal typesetting parameter ontology library, obtaining typesetting parameters matching the target journal;

[0066] A corresponding typesetting parameter configuration file is generated for the typesetting parameters that match the target journal.

[0067] Further optimizing the solution, the process of constructing the recognition rule library based on the multimodal content recognition method and the typesetting parameter configuration file includes:

[0068] Obtain historical typeset papers, and parse the papers based on the corresponding typeset parameter configuration files to obtain different types of content;

[0069] The different types of content are identified based on processing technology to build the identification rule library.

[0070] Further optimizing the solution, the process of identifying the different types of content based on the processing technology and building the identification rule base includes:

[0071] Recognize the text content of the different types of content based on natural language processing technology and build a text recognition rule library;

[0072] Recognize the chart contents of the different types of content based on computer vision technology and build a chart recognition rule library;

[0073] Identify the formula contents of the different types of content based on mathematical formula recognition technology, and build a formula recognition rule library;

[0074] The text recognition rule base, the chart recognition rule base and the formula recognition rule base are integrated to generate the multi-modal recognition rule base.

[0075] According to the journal typesetting parameter ontology library, typesetting parameters matching the target journal are obtained; for the typesetting parameters, corresponding typesetting parameter configuration files are generated; the journal papers to be identified are obtained, and the papers are parsed according to the typesetting parameter configuration files to obtain different types of content such as text, charts and formulas; for the text content, a text recognition rule library is constructed by using natural language processing technology; for the chart content, a chart recognition rule library is constructed by using computer vision technology; for the formula content, a formula recognition rule library is constructed by using mathematical formula recognition technology; the text recognition rule library, the chart recognition rule library and the formula recognition rule library are integrated to construct a multimodal content recognition rule library;

[0076] Specifically, firstly, according to the semantic description and ontology relationship in the journal typesetting parameter ontology library, a rule-based matching algorithm is used to obtain the typesetting parameters that match the target journal, such as the font size of the paper title is small two, the font size of the body is small four, etc. Then, for these typesetting parameters, the corresponding typesetting parameter configuration file is generated in JSON format for subsequent paper parsing and typesetting. Then, the PDF file of the journal paper to be identified is obtained, and the paper is parsed using a PDF parsing library such as PDFBox, and different types of content such as text, charts and formulas are extracted through the layout analysis algorithm. For the extracted text content, natural language processing technologies such as Chinese word segmentation, part-of-speech tagging, and named entity recognition are used to construct a text recognition rule library; for the extracted chart content, computer vision technologies such as edge detection, region segmentation, and feature extraction are used to construct a chart recognition rule library; for the extracted formula content, mathematical formula recognition technologies such as symbol recognition, structural analysis, and semantic interpretation are used to construct a formula recognition rule library. Finally, the recognition rule libraries of text, charts, and formulas are integrated to construct a multimodal content recognition rule library.

[0077] To further optimize the solution, the process of generating structured paper content includes:

[0078] Obtaining a journal article to be identified, identifying the journal article to be identified based on the identification rule base, and generating an identification result;

[0079] The journal article to be identified is identified based on the typesetting parameters corresponding to the identification result, and the structured article content is generated.

[0080] Specifically, the identified structured paper content is first preprocessed, and regular expressions are used to match the reference format in the paper, such as "[1] Author. Paper title [J]. Journal name, year, volume (issue): page range", to extract the paper's citation information. Then, based on the extracted citation information, a citation relationship graph of the paper content is constructed and stored in the Neo4j graph database, where the paper is used as a node and the citation relationship is used as a directed edge. The Louvain community discovery algorithm is then used to process the citation relationship graph, and the modularity threshold is set to 4. It is then iterated to optimize and obtain the semantically closely related paper community division. For each paper community, the TF-IDF algorithm is used to extract the top 20 keywords with the largest weights, generate a 50-dimensional Word2Vec word vector, and calculate the semantic vector representation of the community through the mean of the word vector. The cosine similarity of the semantic vectors of different communities is calculated, and the threshold is set to 7. If the threshold is exceeded, an undirected semantic link edge is added between the corresponding community nodes. Finally, the semantic link information is fused on the basis of the citation relationship graph to construct a paper semantic link graph model. The DeepWalk algorithm is used to randomly walk and sample on the semantic link graph to generate node sequences, and the Skip-gram model is used to train a 128-dimensional node embedding vector. By calculating the similarity of the node embedding vectors, the implicit semantic relationship between the paper contents can be revealed, providing data support for intelligent analysis tasks such as paper recommendation and classification.

[0081] Further optimizing the scheme, the process of converting the citation relationship in the structured paper content into a semantic link graph by the graph-based citation link analysis method includes:

[0082] Preprocessing the structured paper content to obtain citation information of the paper;

[0083] Constructing a citation relationship graph of the paper content based on the citation information of the paper;

[0084] Processing the reference relationship graph of the paper content based on a community discovery algorithm to obtain a paper community with close semantic correlation;

[0085] For each paper community, extract keywords that can represent the semantic features of the community and construct a semantic vector representation of the paper community;

[0086] The semantic link graph is constructed based on the semantic vector representation of the paper community.

[0087] According to the citation relationship between papers, a citation relationship graph is constructed, in which nodes represent papers and edges represent citation relationships. The constructed citation relationship graph is preprocessed, and the threshold filtering method is used to remove noise edges with too few citations, and isolated nodes with a degree of 0 are removed to obtain the preprocessed citation relationship graph. On the preprocessed citation relationship graph, a community discovery algorithm based on modularity optimization is used to divide the community. By maximizing the modularity function, a heuristic optimization algorithm such as the simulated annealing algorithm is used to search for the optimal community division scheme. In the process of community division, the community granularity is controlled by setting a modularity gain threshold. If the modularity gain is less than the threshold, the community division is stopped to avoid over-division. For the divided communities, post-processing is performed according to the size of the community. For small communities with a number of nodes below the preset threshold, they are merged into the most similar large community to improve the semantic closeness of the community. The average semantic similarity between papers within each community is calculated as a measure of the semantic closeness of the community. If the community semantic closeness is lower than the preset threshold, the community is split into multiple sub-communities. Finally, a community of papers with closely related semantics is obtained;

[0088] Specifically, when constructing the citation relationship graph, the paper DOI is used as the unique identifier of the node, and the citation relationship is used as a directed edge. For edges with too few citations, the citation threshold is set to 5, that is, edges with less than 5 citations are removed, and isolated nodes with a degree of 0 are removed. On the preprocessed citation relationship graph, the Louvain algorithm is used for community division, and the modularity function is optimized by continuous iteration until the iteration is stopped when the modularity gain is less than 0001. In the post-processing stage, for communities with less than 10 nodes, the average semantic similarity with other communities is calculated, and they are merged into the community with the highest similarity. The semantic similarity uses the Word2Vec model to calculate the cosine similarity of the word vectors of the paper title and abstract. For communities with a semantic closeness lower than 5, the spectral clustering algorithm is used to further divide them into sub-communities.

[0089] Further optimizing the scheme, the semantic connection graph is uniquely identified based on a hash algorithm, and the process of retrieving and managing journal papers based on the unique identification includes:

[0090] Acquire key information based on the semantic link graph and construct a content feature vector;

[0091] Based on the content feature vector, a unique identification code of the content is calculated using a hash algorithm as an index key of the cache;

[0092] Search the layout cache to see if there is a layout result that matches the current content identification code. If so, directly return the cached result. At the same time, establish a mapping relationship between the layout result and the content identification code and store it in the layout cache.

[0093] Specifically, in the typesetting quality assessment process, firstly, the typesetting result data to be evaluated is obtained according to the pre-established typesetting quality assessment rule base containing 20 evaluation dimensions. In terms of layout and content presentation, layout features (such as element alignment, spacing, hierarchy, etc.) and content features (such as font, color, readability, etc.) are extracted to form a 50-dimensional feature vector. The extracted feature vector is input into at least one pre-trained SVM typesetting quality assessment model, and the RBF kernel function is used to classify the features to obtain a comprehensive typesetting quality score of 0-100 points. If the comprehensive score is lower than the preset threshold of 80 points, it is judged that there is a typesetting quality defect, and local features are extracted for the quality defect area. The quality defect pattern recognition is performed through the trained CNN model, such as identifying defect patterns such as "font inconsistency" and "element overlap". According to the quality defect pattern recognition results, the diagnostic information matching the defect pattern is obtained from the diagnostic knowledge base, and a visual diagnostic feedback result containing the defect location, type, severity, etc. is generated. Associate the diagnostic feedback results with the typesetting result data to form a typesetting quality assessment report that includes scores, defect lists, optimization suggestions, etc., and display the assessment results and diagnostic feedback through the web interface. According to the assessment results and diagnostic feedback, determine the typesetting parameters (such as font size, line spacing, etc.) and rules (such as alignment, color matching, etc.) that need to be optimized, update the SVM typesetting quality assessment model and CNN defect pattern recognition model through incremental learning, and expand the diagnostic knowledge base at the same time. Continuously iterate and optimize the typesetting quality assessment and feedback mechanism, regularly evaluate model performance, continuously improve the accuracy of typesetting quality assessment (such as an increase of 5%) and the effectiveness of diagnostic feedback, and achieve continuous optimization of typesetting quality.

[0094] In a second aspect, this embodiment also discloses a journal paper intelligent typesetting and management system, including:

[0095] A parameter database construction module is used to use ontology and semantic web technology to perform structured representation and semantic description of the typesetting requirements of various journals, and to construct a journal typesetting parameter database based on the description;

[0096] An identification rule base construction module is used to generate a typesetting parameter configuration file based on the journal typesetting parameter database, and to construct an identification rule base based on a multimodal content recognition method and the typesetting parameter configuration file;

[0097] A structured generation module, used for obtaining journal articles to be identified, identifying the journal articles to be identified based on the identification rule base, and generating structured article content;

[0098] A semantic link graph generation module is used to convert the citation relationship in the structured paper content into a semantic link graph based on a graph-based citation link analysis method;

[0099] The retrieval management module is used to uniquely identify the semantic connection graph based on a hash algorithm, and to retrieve and manage journal papers based on the unique identification.

[0100] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for intelligent typesetting and management of journal articles, characterized in that: The following steps are involved: Use ontology and semantic web technology to structure and semantically describe the typesetting requirements of various journals, and build a journal typesetting parameter database based on the description; Generate a typesetting parameter configuration file based on the journal typesetting parameter database, and build a recognition rule library based on the multimodal content recognition method and the typesetting parameter configuration file; Obtaining a journal article to be identified, identifying the journal article to be identified based on the identification rule base, and generating structured article content; The graph-based citation link analysis method converts the citation relationship in the structured paper content into a semantic link graph; The semantic connection graph is uniquely identified based on a hash algorithm, and journal papers are retrieved and managed based on the unique identification.

2. The intelligent typesetting and management method for journal articles according to claim 1, characterized in that: The process of constructing a journal typesetting parameter database based on the description includes: Acquiring typesetting requirement data of a journal, wherein the typesetting requirement data of the journal includes typesetting parameters and rules; Performing structured description and semantic description processing based on the typesetting requirement data of the journal to obtain structured data of the typesetting requirement of the journal; A journal typesetting parameter database is constructed based on the structured data of the journal typesetting requirements.

3. The intelligent typesetting and management method for journal articles according to claim 1, characterized in that: The process of generating a typesetting parameter configuration file based on the journal typesetting parameter database includes: According to the journal typesetting parameter ontology library, obtaining typesetting parameters matching the target journal; A corresponding typesetting parameter configuration file is generated for the typesetting parameters that match the target journal.

4. The intelligent typesetting and management method for journal articles according to claim 3 is characterized in that: The process of constructing a recognition rule library based on the multimodal content recognition method and the typesetting parameter configuration file includes: Obtain historical typeset papers, and parse the papers based on the corresponding typeset parameter configuration files to obtain different types of content; The different types of content are identified based on processing technology to build the identification rule library.

5. The intelligent typesetting and management method for journal articles according to claim 4 is characterized in that: The process of identifying the different types of content based on the processing technology and constructing the identification rule base includes: Recognize the text content of the different types of content based on natural language processing technology and build a text recognition rule library; Recognize the chart contents of the different types of content based on computer vision technology and build a chart recognition rule library; Identify the formula contents of the different types of content based on mathematical formula recognition technology, and build a formula recognition rule library; The text recognition rule base, the chart recognition rule base and the formula recognition rule base are integrated to generate the multi-modal recognition rule base.

6. The intelligent typesetting and management method for journal articles according to claim 1, characterized in that: The process of generating structured paper content includes: Obtaining a journal article to be identified, identifying the journal article to be identified based on the identification rule base, and generating an identification result; The journal article to be identified is identified based on the typesetting parameters corresponding to the identification result, and the structured article content is generated.

7. The intelligent typesetting and management method for journal articles according to claim 1, characterized in that: The process of converting the citation relationship in the structured paper content into a semantic link graph by the graph-based citation link analysis method includes: Preprocessing the structured paper content to obtain citation information of the paper; Constructing a citation relationship graph of the paper content based on the citation information of the paper; Processing the reference relationship graph of the paper content based on a community discovery algorithm to obtain a paper community with close semantic correlation; For each paper community, extract keywords that can represent the semantic features of the community and construct a semantic vector representation of the paper community; The semantic link graph is constructed based on the semantic vector representation of the paper community.

8. The intelligent typesetting and management method for journal articles according to claim 1, characterized in that: The process of uniquely identifying the semantic connection graph based on a hash algorithm and retrieving and managing journal papers based on the unique identification includes: Acquire key information based on the semantic link graph and construct a content feature vector; Based on the content feature vector, a unique identification code of the content is calculated using a hash algorithm as an index key of the cache; Search the layout cache to see if there is a layout result that matches the current content identification code. If so, directly return the cached result. At the same time, establish a mapping relationship between the layout result and the content identification code and store it in the layout cache.

9. A system for intelligent typesetting and management of journal articles according to any one of claims 1 to 8, characterized in that: include: A parameter database construction module is used to use ontology and semantic web technology to perform structured representation and semantic description of the typesetting requirements of various journals, and to construct a journal typesetting parameter database based on the description; An identification rule base construction module is used to generate a typesetting parameter configuration file based on the journal typesetting parameter database, and to construct an identification rule base based on a multimodal content recognition method and the typesetting parameter configuration file; A structured generation module, used for obtaining journal articles to be identified, identifying the journal articles to be identified based on the identification rule base, and generating structured article content; A semantic link graph generation module is used to convert the citation relationship in the structured paper content into a semantic link graph based on a graph-based citation link analysis method; The retrieval management module is used to uniquely identify the semantic connection graph based on a hash algorithm, and to retrieve and manage journal papers based on the unique identification.

Images