A document association information generation system and method
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUAZHONG AGRI UNIV
- Filing Date
- 2026-04-29
- Publication Date
- 2026-07-21
Smart Images
Figure CN122432294A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of computer information processing technology, and specifically relates to a system and method for generating document-related information. Background Technology
[0002] With the deepening application of artificial intelligence (AI) technology in research assistance, more and more systems are attempting to use intelligent models such as large language models to automate the analysis of literature, especially research literature, to support research decision-making tasks such as research topic selection, theory construction, method selection, and trend prediction. However, research literature itself is characterized by its vast quantity, uneven quality, and severe information redundancy. If a full-text input method is directly used to input a large-scale, unfiltered collection of literature into an AI model for analysis, the following problems will arise: large-scale literature input leads to excessively long model processing time, significantly increasing analysis costs and consuming excessive computational resources; due to the randomness and noise interference of input data, the results of multiple executions of the same task differ significantly; low-relevance or low-impact literature dilutes core information, reducing the reliability of analytical conclusions; and the lack of standardized input preprocessing procedures makes it difficult to reproduce and verify the analysis results.
[0003] Compared to full-text input methods, other methods exist, such as simple threshold screening based solely on single criteria like publication date or citation status; and human experience-based screening that relies on researchers manually selecting representative literature, which is inefficient and highly subjective. However, all of these methods lack systematic, automated, and reproducible mechanisms for literature denoising and subset construction, making it difficult to guarantee the quality and stability of AI-analyzed input.
[0004] Furthermore, traditional methods typically output flat lists of literature, lacking structured information such as citation relationships, research context, and key nodes. This limits subsequent intelligent analysis modules, such as large language models, to understanding based solely on isolated abstracts and titles, making it difficult to grasp the knowledge evolution path, community structure, and key influential nodes within the field. Consequently, the analytical conclusions lack depth, interpretability, and a global perspective. Summary of the Invention
[0005] To address the problems raised in the background art, this invention proposes a system and method for generating document-related information.
[0006] A document-related information generation system for achieving one of the objectives of this invention includes: The document retrieval module is used to obtain an initial set of documents with preset fields from external databases; The document value assessment and ranking module is used to assess the document value of the documents in the initial document set, and to rank the documents in the initial document set according to the assessment results, and output the ranked document set. The association information generation module is used to retrieve the sorted document set, filter the documents according to preset filtering rules to obtain a high-value document subset; perform association structure analysis on the high-value document subset to generate document association information; The subset output interaction module is used to output the high-value document subset and document association information in a standard structured data format.
[0007] Furthermore, it also includes a data storage module for storing the initial document set, the sorted document data, and the standardized preprocessed high-value document subset data and their document association information.
[0008] Furthermore, it also includes a subset quality verification module for performing multi-dimensional quality verification on the high-value literature subset. The multi-dimensional quality verification includes the following verification indicators: average citation frequency of the high-value literature subset, proportion of core authors, and proportion of high-impact journals. Core authors refer to authors whose H-index is not lower than a preset threshold of 20. High-impact journals refer to journals with a JCR (Journal Citation Reports) quartile ranking. Specific verification methods include: When the average citation frequency of the high-value literature subset is lower than the first preset threshold, or the proportion of core authors is lower than the second preset threshold, or the proportion of high-impact journals is lower than the third preset threshold, the high-value literature subset is determined to not meet the multi-dimensional quality verification requirements. The first, second, and third preset thresholds can be configured according to different disciplines or user needs, and their default values are 50%, 30%, and 60%, respectively. The subset quality verification module performs the following adjustments based on the type of verification metric that was not met: If the average citation frequency of the high-value document subset does not meet the standard, the association information generation module is triggered to expand the specified number in the preset screening rules to increase the size of the subset. If the proportion of core authors does not meet the standard, the document value assessment and ranking module will be triggered to increase the weight coefficient of citation frequency and then re-execute the ranking process. If the proportion of high-impact journals does not meet the standard, the literature value assessment and ranking module will be triggered to increase the weight coefficient of the journal score and then re-execute the ranking process. After completing the above adjustment operations, the subset quality verification module re-triggers the association information generation module to generate a new high-value document subset and performs verification again until the high-value document subset meets the multi-dimensional quality verification requirements.
[0009] Furthermore, the document value assessment and sorting module sorts the initial document set using a first sorting mode and a second sorting mode. When the missing rate of the citation frequency field is lower than a preset threshold, the first sorting mode is activated, and when the missing rate of the citation frequency field reaches or exceeds the preset threshold, the module switches to the second sorting mode.
[0010] Furthermore, the first sorting mode is a sorting mode based on a multi-indicator weighted comprehensive score, including: normalizing the citation frequency, publication year, and journal score of the literature, calculating a comprehensive score by weighting using preset weight coefficients, and sorting in descending order based on the comprehensive score. When the first sorting mode is enabled, zero-padding is required for any missing values present in this mode to ensure the integrity of the weighted comprehensive score.
[0011] Furthermore, the second sorting mode is a hierarchical comparison sorting mode based on multi-dimensional sorting keys, specifically including comparing and sorting the documents in the order of priority of primary sorting key, secondary sorting key and third sorting key, and using the next level sorting key for sorting when the values of the previous level sorting keys are consistent.
[0012] In the process of denoising and constructing high-value subsets of literature, the reliability of the ranking rules directly determines the quality of the subsequent high-value literature subsets. However, literature data obtained in batches from external academic databases generally suffers from missing citation frequency fields. This is due to reasons such as: newly published literature not yet having its citation data included in the database, incomplete data returned by API interfaces, and inconsistent field mappings between different databases. Research has found that when the citation frequency missing rate is low (e.g., below 30%), using the first ranking mode can fully utilize information from the three dimensions of citation frequency, publication year, and journal score to quantify the value of literature in a refined manner, resulting in high ranking discrimination. However, when the citation frequency missing rate is high (e.g., reaching or exceeding 30%), the ranking results become distorted. In contrast, hierarchical comparison ranking relies only on the relative size relationship of each field, does not require precise numerical values, has a higher tolerance for missing citation frequencies, and produces more robust ranking results.
[0013] Furthermore, the multi-dimensional sorting keys of the document value assessment and sorting module include a primary sorting key, a secondary sorting key, and a tertiary sorting key. When the values of the previous level sorting keys are consistent, the next level sorting key is used for sorting.
[0014] Furthermore, the primary sorting key is the citation frequency of the document, the secondary sorting key is the publication year of the document, and the tertiary sorting key is the journal score of the document.
[0015] Furthermore, the association information generation module performs association structure analysis on the high-value document subset, and the method for generating document association information includes: Extract the reference list and cited list of each document in the high-value document subset, and construct a directed citation network among the documents; Based on the directed citation network, the link analysis algorithm is used to calculate the citation influence of each document, and the core document nodes are determined according to the citation influence. Based on the directed citation network, several document clusters are obtained by dividing the network using a community detection algorithm, and the shortest path algorithm is used to extract the key citation paths between the document clusters. The core document nodes, document clusters, and citation paths between clusters obtained from the analysis are aggregated and statistically analyzed to form a structured descriptive text, which serves as the document association information.
[0016] Furthermore, the subset output interaction module is also used to perform standardized preprocessing of field information on the high-value literature subset. This standardized preprocessing involves standardizing the naming format of author names and journal names in the high-value literature subset, and removing special characters from the abstract field. Specifically, this includes: firstly, standardizing the field information of the high-value literature subset; secondly, uniformly adopting the English format of "full surname spelling + first letter of given name" for author names, with Chinese authors undergoing simultaneous pinyin conversion, and the first letter of the given name capitalized while the full surname spelling is lowercase; thirdly, uniformly adopting the official English standard translation of the journal name, and using the standard full Chinese name if there is no official English translation, avoiding the use of journal abbreviations or acronyms; and fourthly, removing special characters from the abstract field, including removing garbled characters, escape characters, abnormal punctuation marks (not in Chinese or English), and invisible characters, while retaining standard expressions such as uppercase and lowercase letters, numbers, Chinese commas / periods / colons, and English periods / commas / colons.
[0017] Furthermore, the preset filtering rule of the association information generation module is to select a specified number of documents from the top of the sorted document set, and the specified number can be dynamically adjusted according to actual needs.
[0018] Furthermore, the subset output interaction module has a temporary caching function, which can temporarily cache high-value document subsets after standardized preprocessing.
[0019] Furthermore, the literature retrieval module obtains the initial literature set from external academic databases through a standardized application programming interface.
[0020] Furthermore, the multi-dimensional sorting keys of the document value assessment and sorting module include a primary sorting key, a secondary sorting key, and a tertiary sorting key. When the values of the previous level sorting keys are consistent, the subsequent level sorting key is used for sorting.
[0021] Furthermore, the standard structured data format of the subset output interaction module is the general JSON data format.
[0022] Furthermore, when the subset quality verification module detects that a subset of high-value documents does not meet the verification requirements, it can also trigger the document value assessment and ranking module to adjust the weight coefficients of the multi-dimensional ranking key and then re-execute the ranking process.
[0023] A method for generating document-related information to achieve the second objective of this invention includes: Retrieve an initial set of documents with preset fields from an external database; The literature in the initial literature set is evaluated for its value, and the literature in the initial literature set is sorted according to the evaluation results, and the sorted literature set is output. The sorted document set is retrieved, and the sorted document set is filtered according to a preset filtering rule to obtain a high-value document subset; the association structure analysis of the high-value document subset is performed to generate document association information; The high-value document subset and document association information are output in a standard structured data format.
[0024] A non-transitory computer-readable storage medium for achieving the third objective of the present invention stores a computer program thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the method for generating document association information.
[0025] A computer program product for achieving the fourth objective of the present invention includes a computer program / instruction that, when executed by a processor, implements the steps of the document association information generation method.
[0026] The beneficial effects of this invention include: 1. By screening and constructing a high-quality subset, the number of documents input into the artificial intelligence model is significantly reduced, thereby reducing the consumption of computing resources and analysis time; 2. The standardized screening process ensures the consistency of input data, making the results of artificial intelligence analysis more reproducible and stable; 3. Eliminate interference from low-quality literature, enabling AI analysis to focus on high-value literature and improve the academic credibility of the analysis conclusions; 4. The method allows for flexible parameter adjustment to adapt to the processing needs of literature collections of different disciplines and sizes; 5. No manual intervention is required, achieving full automation from literature retrieval to subset construction; 6. By constructing a directed citation network and extracting core document nodes, document clusters, and key citation paths, structured document association information is generated, transforming a flat document list into a knowledge graph with semantic relationships, providing an interpretable global perspective for subsequent intelligent analysis; community detection algorithms and shortest path algorithms are used to automatically divide research subfields and identify knowledge evolution paths, enabling researchers to intuitively grasp the context and turning points of discipline development, improving the depth and accuracy of research decisions; document association information and high-value document subsets are output together to form a standardized data package of document collections and relational structures, which can be directly connected to artificial intelligence analysis systems such as large language models, making up for the lack of structured relationships in pure text input; 7. By setting a dual-mode adaptive sorting strategy, the system automatically selects between weighted comprehensive score sorting and hierarchical comparison sorting based on the missing rate of the citation frequency field in the literature collection, thus solving the problem of unreliable sorting results caused by missing data. This strategy can stably output a high-quality subset of high-value literature under different data quality conditions. In particular, it avoids the systematic exclusion of newly published literature with potential high value but missing data due to lack of citations, which significantly improves the robustness and adaptability of the technical solution described in this invention. Attached Figure Description
[0027] Figure 1 This is an integrated architecture diagram of the method of the present invention in an artificial intelligence analysis system; Figure 2 This is an overall flowchart of the method of the present invention; Figure 3 A schematic diagram of the selection of a subset of high-value documents; Figure 4 This is a flowchart for subset quality verification. Detailed Implementation
[0028] The following detailed embodiments are provided to explain the technical solutions of the present invention, so that those skilled in the art can understand the present invention. The scope of protection of the present invention is not limited to the following specific embodiments. Any modifications or improvements made by those skilled in the art that incorporate the technical solutions of the present invention but differ from the following detailed embodiments are also within the scope of protection of the present invention.
[0029] Example 1 A system for generating document-related information. See also: Figure 1 The system includes: The document retrieval module is used to obtain an initial set of documents with preset fields from external databases; The document value assessment and ranking module is used to assess the document value of the documents in the initial document set, and to rank the documents in the initial document set according to the assessment results, and output the ranked document set. The association information generation module is used to retrieve the sorted document set, filter the documents according to preset filtering rules to obtain a high-value document subset; perform association structure analysis on the high-value document subset to generate document association information; The subset output interaction module is used to output the high-value document subset and document association information in a standard structured data format.
[0030] In one embodiment, the system further includes a subset quality verification module for performing multi-dimensional quality verification on the high-value literature subset. The multi-dimensional quality verification includes the following verification indicators: average citation frequency of the high-value literature subset, proportion of core authors, and proportion of high-impact journals. Specific verification methods include: When the average citation frequency of the high-value literature subset is lower than the first preset threshold, or the proportion of core authors is lower than the second preset threshold, or the proportion of high-impact journals is lower than the third preset threshold, the high-value literature subset is determined to not meet the multi-dimensional quality verification requirements. The subset quality verification module performs the following adjustments based on the type of verification metric that was not met: If the average citation frequency of the high-value document subset does not meet the standard, the association information generation module is triggered to expand the specified number in the preset screening rules to increase the size of the subset. If the proportion of core authors does not meet the standard, the document value assessment and ranking module will be triggered to increase the weight coefficient of citation frequency and then re-execute the ranking process. If the proportion of high-impact journals does not meet the standard, the literature value assessment and ranking module will be triggered to increase the weight coefficient of the journal score and then re-execute the ranking process. After completing the above adjustment operations, the subset quality verification module re-triggers the association information generation module to generate a new high-value document subset and performs verification again until the high-value document subset meets the multi-dimensional quality verification requirements.
[0031] The first sorting mode is a sorting mode based on a multi-indicator weighted comprehensive score, specifically including: normalizing the citation frequency, publication year, and journal score of the literature, calculating the comprehensive score by weighting using preset weight coefficients, and sorting in descending order based on the comprehensive score. In this embodiment, the normalization method adopts minimum-maximum normalization, and the default weight coefficients are configured as 0.6 for citation frequency, 0.3 for publication year, and 0.1 for journal score.
[0032] The second sorting mode is a hierarchical comparison sorting mode based on multi-dimensional sorting keys. Specifically, it includes: comparing and sorting documents sequentially according to the priority order of the primary sorting key, secondary sorting key, and tertiary sorting key; when the values of the previous level sorting keys are the same, the next level sorting key is used for sorting. The primary sorting key is the citation frequency of the document, the secondary sorting key is the publication year of the document, and the tertiary sorting key is the journal score of the document. For example, first sort by citation frequency in descending order; if the citation frequencies are the same, sort by publication year in descending order; if both are the same, sort by journal score in descending order.
[0033] In one embodiment, the primary sorting key is the citation frequency of the document, the secondary sorting key is the publication year of the document, and the third sorting key is the journal score of the document.
[0034] In one embodiment, the subset output interaction module is further used to perform standardized preprocessing of field information on the high-value literature subset. The standardized preprocessing includes standardizing the naming format of the author names and journal names of the high-value literature subset, and removing special characters from the abstract field.
[0035] In one embodiment, the association information generation module performs association structure analysis on a subset of high-value documents, specifically including the following steps: Step 1: Construct a directed reference network The system retrieves the reference list and cited list (i.e., which documents cited the document) for each document in the high-value document subset via the OpenAlex or CrossRef API. All documents appearing in the references or cited documents are treated as network nodes, denoted as V, with a total of N = |V| nodes. For each document... i References j Establish a reference relationship from node j Pointing to node i The directed edge e j→i This forms a set of directed edges E. The final directed reference network graph is G=(V,E).
[0036] Step 2: Calculate citation impact and identify core literature nodes This embodiment uses a Personalized PageRank algorithm with restart to calculate the citation influence of each document, in order to more accurately reflect its relative importance within a limited set of documents. The specific steps are as follows: Set the damping factor α = 0.85 (meaning there is an 85% probability of continuing along the outgoing chain and a 15% probability of jumping back to the initial seed node set).
[0037] Define a personalized vector v: designate each document in the high-value document subset as a seed node and assign it an equal initial weight, i.e., for the seed node set S (size m=120). v i =1 / m, if i∈S, otherwise v i =0.
[0038] Iteratively calculate the PageRank value r until convergence: r (t+1) =α×r (t) ×M+(1-α)×v Where M is the transition probability matrix (if node i Excessive d i Then M i→j =1 / d i For dangling nodes that have not left the chain, jump evenly to all nodes.
[0039] After convergence, the top k documents with the highest values in r are selected as core document nodes. In this embodiment, k = max(10, ⌈0.1×N⌉), which means at least 10 documents or 10% of the total number of nodes. The titles, authors, and citation impact scores of these core document nodes are recorded as structured data.
[0040] Step 3: Divide the literature into clusters and extract key citation paths This embodiment uses the Louvain community detection algorithm to partition document clusters: Initially, each node is treated as an independent community. The first stage is modularity optimization, which includes: traversing each node and calculating the modularity increment ΔQ after moving it to any neighboring community; the modularity calculation formula is: Where L is the total number of sides, A ij For elements of the adjacency matrix (directed edges are considered bidirectional), d i For nodes i The degree (in-degree + out-degree), δ(c) i ,c j Indicator node i and j Determine if the node is in the same community. Move the node to the community that maximizes ΔQ; otherwise, leave it in its current position. Repeat this iteration until the modularity no longer increases.
[0041] The second phase is network cohesion, which includes: compressing the communities obtained in the first phase into supernodes, constructing a new weighted network, and returning to the first phase for further optimization.
[0042] After convergence, several document clusters are obtained, each containing a group of documents closely related in terms of citation. The system automatically generates topic tags for each cluster: extracting high-frequency words from the titles and abstracts of all documents within the cluster (after filtering for stop words), and taking the three keywords with the highest word frequency as cluster tags.
[0043] Each cluster is treated as a supernode, and the citation relationships between clusters are calculated: if there is a citation edge from any document in cluster A to any document in cluster B, a directed edge from A to B is added to the hypergraph, with a weight equal to the total number of citations between clusters. Then, Dijkstra's shortest path algorithm is used to calculate the shortest path between each pair of supernodes (using hop count or the reciprocal of the weight as the distance), and the top 5 paths with the highest betweenness centrality connecting different clusters are extracted as key citation paths. Each path is formatted as: cluster label 1 → cluster label 2 → ... → cluster label n, representing the evolution or migration of research ideas.
[0044] Step 4: Generate structured description text The core document nodes, document clusters, and citation paths between clusters obtained from the above analysis are aggregated and statistically analyzed, and structured descriptive text is generated according to a preset template as document association information.
[0045] [Document Association Structure Information] Total number of nodes: {N}, Total number of references: {E}.
[0046] Core literature nodes (Top {k} of citation influence): 1. Title: [Title 1], Author: [Author 1], Citation Impact Score: [Score 1] 2. Title: [Title 2], Author: [Author 2], Citation Impact Score: [Score 2] ... Document clustering (a total of {X} clusters): Cluster 1: Tags [Tag A, Tag B, Tag C], containing {size1} documents, representing documents: [List 2-3 core document titles].
[0047] Cluster 2: Tags [Tag D, Tag E, Tag F], containing {size2} documents, representing documents: ... ... Key inter-cluster reference paths (research context): Path 1: {Cluster A} → {Cluster B} → {Cluster C} (Citation count: {weight}) Path 2: ... Example 2 A method for generating document association information. See also Figure 2The method includes the following steps: Step S101: Obtain an initial set of documents with preset fields from an external database. Document data is obtained from an external database (such as OpenAlex Academic Database) via an API interface, and this document data constitutes the initial set of documents. In one embodiment, the initial set of documents is obtained according to user-defined search conditions, which are: keyword "Machine Learning", time range 2020-2024, and no specific restrictions on journal source. The system obtains a total of 1850 relevant documents, each containing preset fields such as title, author list, publication year, citation frequency, DOI, abstract, and source journal. These preset fields are basic metadata fields required for document preprocessing. In addition to the above fields, extended metadata fields such as author affiliation, funding project, and document type can be added according to research analysis needs. All preset fields are obtained from the standardized interface of the external database. The external database can be OpenAlex Academic Database, or other conventional academic literature external databases such as Web of Science, CNKI, and Wanfang, and document data acquisition is completed through the standardized API interface corresponding to each database. Step S102: Evaluate the literature value of the literature in the initial literature set, sort the literature in the initial literature set according to the evaluation results, and output the sorted literature set.
[0048] In this embodiment, the document value assessment and ranking module adopts a hierarchical ranking strategy: First, the missing rate of the citation frequency field in the initial document set is calculated. If the missing rate is less than 30%, the first ranking mode (weighted comprehensive score ranking) is activated; if the missing rate reaches or exceeds 30%, the second ranking mode (hierarchical comparison ranking) is switched. In the first ranking mode, the citation frequency, publication year, and journal score of each document are min-max normalized, and then a comprehensive score is calculated by weighting at 0.6, 0.3, and 0.1, and the documents are ranked in descending order of comprehensive score. In the second ranking mode, the documents are first ranked in descending order of citation frequency, then in descending order of publication year, and finally in descending order of journal score.
[0049] In one embodiment, the second sorting mode includes sorting according to the following rules: Primary sorting key: Citation frequency, sorted in descending order; Secondary sorting key: Publication year, sorted in descending order; Third sorting key: Journal score, sorted in descending order; In one embodiment, the process further includes sorting journals based on their scores calculated using JCR quartiles and journal impact factors, where the journal scores serve as a third sorting key and are arranged in descending order. Journals without JCR quartiles or impact factors are scored as 0.
[0050] After sorting, the documents are arranged from highest to lowest citation frequency; those with the same citation frequency are arranged from newest to oldest publication year; and those with both the same citation frequency and publication year are arranged from highest to lowest journal score. Table 1 below illustrates the logical diagram of the document quality assessment and sorting process.
[0051] Table 1 Step S103: Retrieve the sorted document set and filter the documents according to preset filtering rules to obtain a high-value document subset. The system selects the high-value document subset according to the preset filtering rules and performs association structure analysis on the high-value document subset to generate document association information. The preset filtering rules are: select a specified number of documents (e.g., 100) from the sorted documents. This specified number can be dynamically adjusted according to the actual scientific research analysis needs of different disciplines and document set sizes. After adjustment, the subset selection operation is re-executed. After this step, the 100 most cited documents from 1850 documents are selected to form a high-value subset. See also Figure 3 This shows a schematic diagram of subset selection.
[0052] This step specifically includes: S103.1: Obtain a subset of high-value documents according to the preset screening rules.
[0053] S103.2: Extract the reference list and cited list of each document in the high-value document subset, and construct a directed citation network among the documents.
[0054] S103.3: Based on the directed citation network, a link analysis algorithm (such as the PageRank algorithm) is used to calculate the citation influence of each document, and the core document nodes (such as the top 10% in terms of influence) are determined according to the citation influence.
[0055] S103.4: Based on the directed citation network, several document clusters are obtained by dividing the network using a community discovery algorithm (such as the Louvain algorithm), and the key citation paths between each document cluster are extracted using a shortest path algorithm (such as the Dijkstra algorithm).
[0056] S1035: Aggregate and statistically analyze the core document nodes (such as title, author, score), document clusters (such as tags, size, representative documents), and citation paths between clusters to form structured descriptive text, which serves as document association information. For example: [Literature Association Structure Information] Total number of nodes: 156, Total number of citation relationships: 423. Core literature nodes (Top 10 in citation influence): 1. Title: "Application of Deep Learning in Medical Image Analysis", Author: "ZW", Citation influence score: 0.087. Literature cluster division (3 clusters in total): Cluster 1: Tags [Deep Learning, Neural Networks, Representation Learning], containing 45 articles, representative articles: [Article A, Article D]. Cluster 2: Tags [Model Compression, Edge Computing], containing 38 articles. Key citation paths between clusters (research context): Path 1: Cluster 1 → Cluster 2 (citation count: 23); Path 2: Cluster 1 → Cluster 3 (citation count: 15).
[0057] Step S104: Output the high-value literature subset in a standard structured data format and input it into the artificial intelligence analysis. The system first performs standardized preprocessing on the field information of the high-value literature subset, including unifying author names to the format of "full last name + first letter of first name", removing special characters from the abstract field, and unifying journal names to the official standard translation. Then, the high-value literature subset is output in a standard JSON format, with the following data structure: { "high_value_subset":[ { "id": "W123456789", "title":"Deep Learning for Medical Image Analysis", "authors":["Smith,J.","Chen,L.","Wang,Y."], "year": 2023 "citations":245, "journal":"Nature Medicine" "abstract":"This paper presents.. "doi":"10.1038 / s41591-023-02245-3" }, / / ...99 other documents ], "metadata": { "total_original":1850, "total_subset":100, "selection_criteria":"top_100_by_citations", "avg_citations": 156.7 } } The standard structured data format described in this invention is a general key-value pair structured data format that can be recognized by computer programs. In this embodiment, JSON format is used, but it can also be adapted to other conventional standard structured data formats such as XML and CSV. The output fields are consistent with the preset fields of the initial document set.
[0058] The aforementioned subset of high-value documents in JSON format and its associated information are stored in a local cache database and then pushed to the artificial intelligence analysis module via a RESTful API interface in a streaming manner. The transmission process uses the JSON-RPC data interaction protocol to ensure the integrity and consistency of data transmission. This subset of high-value documents and its associated information can then be input into the artificial intelligence analysis module to perform tasks such as research gap identification and trend prediction.
[0059] Example 3 This embodiment adds a multi-index weighted ranking function to the existing embodiment 2. The remaining steps (S101, S103, S104) are completely consistent with embodiment 2, except that the literature quality assessment and ranking rules in step S102 are optimized, as follows: In step S102, the system uses a weighted scoring method to comprehensively evaluate the literature and obtain a comprehensive score. The scoring formula is: The overall score is calculated as follows: w1 × Normalized Citation Frequency + w2 × Normalized Publication Year + w3 × Normalized Journal Score. In this embodiment, the weighting coefficients are configured as follows: Citation Frequency 0.6, Publication Year 0.3, and Journal Score 0.1, i.e., weighting coefficients w1 = 0.6, w2 = 0.3, and w3 = 0.1. These weighting coefficients can be flexibly configured and adjusted according to different research analysis needs. The normalization method is minimum-maximum normalization. The normalization calculation formula for each indicator is: Normalized Value = (Original Value of Indicator - Minimum Value of Indicator) / (Maximum Value of Indicator - Minimum Value of Indicator). After calculation, each indicator value is mapped to the range of 0-100.
[0060] After calculating the comprehensive score for each document using the above formula, the system sorts the 1850 documents from highest to lowest comprehensive score. The comprehensive scores for the documents in Table 1 are then calculated, resulting in the document quality assessment and sorting diagram shown in Table 2 below: Table 2 Table 2 is arranged in descending order of comprehensive score. The comprehensive score is calculated as follows: Comprehensive score = 0.6 × normalized citation frequency + 0.3 × normalized publication year + 0.1 × journal score.
[0061] After sorting, the top 100 articles with the highest comprehensive scores are selected according to the preset rules of Example 2 to form a subset of high-value articles.
[0062] Example 4 This embodiment, based on embodiment 2, adds a subset quality verification step (step S105). The execution rules of steps S101-S104 are completely consistent with those of embodiment 2. After step S104 is executed, step S105 is added, forming a complete process of obtaining the initial literature set, quality assessment and ranking, generation of correlation information, selection of high-value subsets, subset standardization output, subset quality verification, and AI analysis input. Figure 4 As shown, it includes the following steps: Step S105: High-value subset quality verification. A high-value document subset stored in the local cache database is retrieved for quality assessment. Verification metrics include: 1. Average citation frequency of subsets; 2. Percentage of core authors (authors with an H-index ≥ 20); 3. Percentage of high-impact journals (referring to JCR Q1 journals, where Q1 refers to the top 25% of journals); The multi-dimensional quality verification described in this invention involves comprehensively verifying the core quality indicators of a subset of high-value literature. These indicators include the subset's average citation frequency, the proportion of core authors, and the proportion of high-impact journals. The multi-dimensional quality verification requires that each verification indicator simultaneously meet a preset threshold; failure to meet any indicator results in failure to pass the verification. The average citation frequency of the subset is calculated by summing the citation frequencies of all literature in the high-value subset and then dividing by the total number of articles in the subset.
[0063] The verification threshold is set as follows: Average citation frequency ≥ first preset threshold 50 The proportion of core authors is ≥ 30% of the second preset threshold. The proportion of high-impact journals is ≥ 60% of the third preset threshold. If all indicators meet the threshold requirements, the verification is considered successful; if any indicator falls below the threshold, the system automatically adjusts the selection strategy. Adjustment methods include: If the average citation frequency of the subset does not meet the target, the association information generation module is triggered to expand the specified number in the preset filtering rules (e.g., from 100 to 150 articles) to increase the size of the subset. If the proportion of core authors does not meet the standard, the document value assessment and ranking module will be triggered to increase the weight coefficient of citation frequency (e.g., from 0.6 to 0.7) and then re-execute the ranking process. If the proportion of high-impact journals does not meet the standard, the literature value assessment and ranking module will be triggered to increase the weight coefficient of the journal score (e.g., from 0.1 to 0.2) and then re-execute the ranking process.
[0064] After performing the differentiated adjustment operation, the system re-triggers the association information generation module to generate a new subset of high-value documents and executes step S105 again until all verification indicators reach the corresponding preset thresholds. In this embodiment, the system performs adaptive iterative adjustments. After several adjustments, the average citation frequency, core author ratio, and high-impact journal ratio of the subset all meet the threshold requirements, the verification is passed, and finally, a subset of high-value documents that meets the quality requirements and its association information are output.
[0065] Table 3 below is a comparison table of the effects of different document preprocessing methods, demonstrating the technical advantages of the method of the present invention compared with traditional methods.
[0066] Table 3 As can be seen, compared with the traditional direct analysis method, the method described in this invention reduces the analysis time by 86.2%, improves the stability of the results by 54.1%, improves the expert score by 38.7%, and reduces the computational cost by 85.2%, demonstrating significant advantages in terms of analysis efficiency, result reliability, and cost control.
[0067] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.
[0068] This invention also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the various steps of the method described in this invention.
[0069] This invention also provides a non-transitory computer-readable storage medium storing a computer program. The computer program includes program instructions that, when executed by a processor, implement the various steps of the method described in this invention, which will not be elaborated further here.
[0070] The computer-readable storage medium can be the data transmission apparatus or the internal storage unit of a computer device provided in any of the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium can also be the external storage device of the computer device, such as the plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on the computer device.
[0071] Furthermore, the computer-readable storage medium may include both internal storage units and external storage devices of the computer device. The computer-readable storage medium is used to store the computer program and other programs and data required by the computer device. The computer-readable storage medium may also be used to temporarily store data that is to be output or has already been output.
[0072] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0073] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0074] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0075] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0076] The contents not described in detail in this specification are existing technologies known to those skilled in the art.
Claims
1. A system for generating document-related information, characterized in that, include: The document retrieval module is used to obtain an initial set of documents with preset fields from external databases; The document value assessment and ranking module is used to assess the document value of the documents in the initial document set, and to rank the documents in the initial document set according to the assessment results, and output the ranked document set. The association information generation module is used to retrieve the sorted document set, filter the sorted document set according to preset filtering rules, and obtain a high-value document subset. The high-value document subset is subjected to association structure analysis to generate document association information; The subset output interaction module is used to output the high-value document subset and document association information in a standard structured data format.
2. The document association information generation system as described in claim 1, characterized in that, It also includes a subset quality verification module for multi-dimensional quality verification of the high-value literature subset. The multi-dimensional quality verification includes the following verification indicators: average citation frequency of the subset, percentage of core authors, and percentage of high-impact journals. High-impact journals refer to journals in the JCR Q1 quartile. Specific verification methods include: When the average citation frequency of the subset is lower than the first preset threshold, or the proportion of core authors is lower than the second preset threshold, or the proportion of high-impact journals is lower than the third preset threshold, the high-value literature subset is determined to not meet the multi-dimensional quality verification requirements. The subset quality verification module performs the following adjustments based on the verification metrics that do not meet the multi-dimensional quality verification requirements: If the average citation frequency of the subset does not meet the standard, the association information generation module is triggered to expand the specified number in the preset filtering rules to increase the size of the subset; If the proportion of core authors does not meet the standard, the document value assessment and ranking module will be triggered to increase the weight coefficient of citation frequency and then re-execute the ranking process. If the proportion of high-impact journals does not meet the standard, the literature value assessment and ranking module will be triggered to increase the weight coefficient of the journal score and then re-execute the ranking process. After completing the above adjustment operations, the subset quality verification module re-triggers the association information generation module to generate a new high-value document subset and performs verification again until the high-value document subset meets the multi-dimensional quality verification requirements.
3. The document association information generation system as described in claim 1, characterized in that, The document value assessment and sorting module uses a first sorting mode and a second sorting mode to sort the initial document set. When the missing rate of the citation frequency field is lower than a preset threshold, the first sorting mode is activated. When the missing rate of the citation frequency field reaches or exceeds the preset threshold, the module switches to the second sorting mode.
4. The document association information generation system as described in claim 3, characterized in that, The first sorting mode is a sorting mode based on a multi-indicator weighted comprehensive score, which includes: normalizing the citation frequency, publication year and journal score of the literature respectively, calculating the comprehensive score by weighting through preset weight coefficients, and sorting in descending order based on the comprehensive score.
5. The document association information generation system as described in claim 3 or 4, characterized in that, The second sorting mode is a hierarchical comparison sorting mode based on multi-dimensional sorting keys. Specifically, it includes comparing and sorting the documents in the order of priority of primary sorting key, secondary sorting key and tertiary sorting key. When the values of the sorting keys at the previous level are consistent, the sorting key at the next level is used for sorting.
6. The document association information generation system as described in claim 5, characterized in that, The primary sorting key is the citation frequency of the document, the secondary sorting key is the publication year of the document, and the tertiary sorting key is the journal score of the document.
7. The document association information generation system as described in claim 1, characterized in that, The association information generation module performs association structure analysis on the high-value document subset, and the method for generating document association information includes: Extract the reference list and cited list of each document in the high-value document subset, and construct a directed citation network among the documents; Based on the directed citation network, the link analysis algorithm is used to calculate the citation influence of each document, and the core document nodes are determined according to the citation influence. Based on the directed citation network, several document clusters are obtained by dividing the network using a community detection algorithm, and the shortest path algorithm is used to extract the key citation paths between the document clusters. The core document nodes, document clusters, and citation paths between clusters obtained from the analysis are aggregated and statistically analyzed to form a structured descriptive text, which serves as the document association information.
8. A method for generating document association information in the system as described in claim 1, characterized in that, include: Retrieve an initial set of documents with preset fields from an external database; The literature in the initial literature set is evaluated for its value, and the literature in the initial literature set is sorted according to the evaluation results, and the sorted literature set is output. The sorted document set is retrieved, and the sorted document set is filtered according to a preset filtering rule to obtain a subset of high-value documents; The high-value document subset is subjected to association structure analysis to generate document association information; The high-value document subset and document association information are output in a standard structured data format.
9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the document association information generation method as described in claim 8.
10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the steps of the document association information generation method of claim 8.