A reference proofreading method, system, electronic device and storage medium

Abstract

The application relates to the technical field of document proofreading, and discloses a reference document proofreading method and system, an electronic device and a storage medium, which comprise a captured publishing agency format specification change event, a citation feature data set extracted from an academic database and a dynamically constructed rule feature library; in response to an operation instruction of a user interface, the operation instruction is converted into a regular expression rule set, and the parameters of the regular expression are optimized; based on the optimized rule set and a sample set, a comprehensive evaluation numerical value is calculated and generated, and a rule is activated when the comprehensive evaluation numerical value is higher than a preset quality threshold value; in response to a target reference document, a corresponding activated rule in the rule feature library is called to perform regular matching, and a plurality of levels of verification of authenticity are performed on a regular matching result to generate a proofreading report comprising a verification result. The application can improve the accuracy and reliability of document proofreading, avoid the uncertainty of manual proofreading, and improve the proofreading efficiency.

Description

Technical Field

[0001] This invention belongs to the field of literature review, and in particular relates to a method, system, electronic device and storage medium for reviewing references. Background Technology

[0002] In today's academic research field, the pace of knowledge innovation and dissemination is accelerating, with academic achievements constantly emerging in the form of papers, monographs, and other publications. References, as an important component of academic research, not only demonstrate respect for and cite the research findings of predecessors but also serve as a vital link for academic exchange and knowledge transmission. They provide readers with clues to further explore related research, contributing to the in-depth development of academic research.

[0003] With the rapid development of internet technology, access to academic information has become more convenient, and academic databases contain a vast amount of academic literature, providing researchers with abundant resources. However, this has also led to a sharp increase in the number of references, placing higher demands on the review and verification of references.

[0004] In academic publishing, different journals and publishers have varying requirements for the format and citation of references. These requirements typically include the order of references, the format of bibliographic entries (such as the way author names are written and the abbreviation of journal names), and the use of punctuation. Accurate citation of references not only reflects the author's professional competence but also relates to the standardization and credibility of academic achievements.

[0005] In current technology, the review of references mainly relies on manual work. Editors or scholars need to spend a lot of time and energy checking the format and content of each reference against the journal's requirements. This manual review method is inefficient and easily affected by human factors such as fatigue and negligence, leading to inaccurate results. Furthermore, as academic standards are updated and journal requirements change, manual review struggles to adapt to these changes in real time, further increasing the difficulty of the review process.

[0006] Therefore, this application provides a reference review method to solve the above-mentioned technical problems. Summary of the Invention

[0007] The purpose of this invention is to provide a reference review method, system, electronic device, and storage medium to solve the technical problem that manual review is difficult to adapt to the impact of various changing factors, resulting in low review efficiency in the prior art.

[0008] To address the aforementioned technical problems, this invention provides a method for reviewing references, comprising:

[0009] In response to captured publishing institution format specification change events and citation feature datasets extracted from academic databases, the format specification change events and citation feature datasets are fused and analyzed to dynamically construct a rule feature library, wherein the rule feature library includes the distribution pattern of bibliographic item formats;

[0010] In response to user interface operation commands, based on the format distribution pattern of the bibliographic entries in the rule feature library, the operation commands are converted into a regular expression rule set. At the same time, a test sample set is generated according to the sample features in the rule feature library. The matching test between the rule set and the sample set is performed. The error type distribution of unmatched samples is analyzed and the parameters of the regular expression are optimized.

[0011] Based on the optimized regular expression rule set and the sample set, the logical contradictions of the rules in the rule set and the coverage defects of the samples in the sample set are verified and calculated, and a comprehensive evaluation value is generated. When the comprehensive evaluation value is higher than a preset quality threshold, the rule is activated.

[0012] The activated rules are encoded into a structured storage format according to feature dimensions, a version control index is established to record the rule generation time and change history, and a change tracking mechanism is set up synchronously to respond to and update the rule feature library;

[0013] In response to the target reference, the corresponding structured rules stored in the rule feature library are invoked to perform regular expression matching, and the authenticity of the regular expression matching results is verified at multiple levels to generate a review report including the verification results.

[0014] In some specific embodiments, in response to captured publishing institution format specification change events and citation feature datasets extracted from academic databases, the format specification change events and the citation feature datasets are fused and analyzed to dynamically construct a rule feature library, wherein the rule feature library includes the distribution pattern of bibliographic item formats, and further includes:

[0015] Periodically scan the reference format pages on the official websites of publishing institutions and identify format specification change events through content hash comparison;

[0016] Analyze the frequency and distribution patterns of author names and key fields in the title structure of the citation feature dataset in the academic database.

[0017] The captured format specification change events are fused with the frequency and distribution patterns of the main fields analyzed to generate a rule feature library with timestamp version identifiers.

[0018] When a change in the format specification is detected, a comparison between the old and new versions is performed, and a rule feature library update instruction is generated.

[0019] In some specific embodiments, in response to operation instructions from the user interface, based on the format distribution patterns of the bibliographic entries in the rule feature library, the operation instructions are converted into a set of regular expression rules. Simultaneously, a test sample set is generated according to the sample features in the rule feature library. A matching test is performed between the rule set and the sample set. The error type distribution of unmatched samples is analyzed, and the parameters of the regular expression are optimized. Further steps include:

[0020] Analyze the format distribution pattern of the bibliographic entries in the rule feature library to generate a regular expression template including named capture groups;

[0021] Based on the regular expression template, the user's drag-and-drop and check-in operations on the interface are mapped to regular expression parameter configurations in real time.

[0022] Based on the sample features recorded in the rule feature library, a test sample set covering multiple scenarios is constructed;

[0023] Perform batch matching tests between the rule set and the sample set, statistically analyze the error type distribution data of unmatched samples, and adjust the character range definition and quantifier parameters of the named capture group based on the error type distribution data of the unmatched samples.

[0024] In some specific embodiments, based on the optimized regular expression rule set and the sample set, logical contradictions in the rules in the rule set and coverage defects in the samples in the sample set are verified and calculated to generate a comprehensive evaluation value. When the comprehensive evaluation value is higher than a preset quality threshold, the rule is activated, further including:

[0025] The regular expression rule set and the associated sample set are input into the pre-trained validation model;

[0026] Verify logical contradictions in the rules in the rule set, and identify blind spots in sample coverage and missing boundary conditions;

[0027] Calculate the comprehensive evaluation value of the logical defect rate and sample coverage of the rules in the rule set. When the comprehensive evaluation value is higher than the preset quality threshold, the rule is activated and marked.

[0028] When the comprehensive evaluation value is lower than the preset quality threshold, error type distribution data is extracted for the corresponding rules, the process is reconstructed, and the parameters of the regular expression are optimized accordingly.

[0029] In some specific embodiments, in response to a target reference, the corresponding structured rules stored in the rule feature library are invoked to perform regular expression matching, and the regular expression matching results are subjected to multi-level authenticity verification to generate a review report including the verification results, further including:

[0030] Based on the journal type of the target reference selected by the user, the corresponding activation rule in the rule feature library is invoked to perform regular expression matching;

[0031] Extract the metadata of successfully matched bibliographic entries and initiate authenticity verification of fields such as title, author, and journal name to the academic database;

[0032] Perform semantic completion on fields that fail to match, generate candidate metadata, and then re-initiate authenticity verification;

[0033] Based on the authenticity verification results, a review report is generated, which integrates data on format errors and authenticity deviations, and generates a structured review report containing the error location, error type, and correction plan.

[0034] In some specific embodiments, when the format specification change event is detected, a comparison of the differences between the old and new versions is performed and a rule feature library update instruction is generated, further including:

[0035] The specific change points of the new and old format specification change events are located using text difference algorithms;

[0036] Analyze the impact range of the change points on the distribution pattern of the recorded item formats in the rule feature library;

[0037] Based on the scope of influence, reconstruct the format distribution model of the rule feature library and generate a new version identifier;

[0038] The version change record is written into the historical trajectory of the rule feature library and the associated rule update notification is triggered.

[0039] In some specific embodiments, the multi-level authenticity verification further includes:

[0040] Basic field validation includes accurately comparing the metadata of bibliographic entries that have been successfully matched by regular expressions with the corresponding data in the academic database, and detecting differences in title spelling and author name order;

[0041] Semantic completion verification includes completing the complete bibliographic metadata for fields that fail to match regular expressions through semantic analysis, and performing cross-database joint query verification;

[0042] Variation compatibility checks include performing multilingual spelling variant compatibility checks on the author name field, handling abbreviations, aliases, and culturally different variants.

[0043] Based on the same concept, the present invention also provides a reference review system, comprising:

[0044] The rule feature library construction module is configured to respond to captured publishing institution format specification change events and citation feature datasets extracted from academic databases, and to perform fusion analysis on the format specification change events and the citation feature datasets to dynamically construct a rule feature library, wherein the rule feature library includes the distribution pattern of bibliographic item formats;

[0045] The matching test module is configured to respond to operation commands from the user interface. Based on the distribution pattern of the bibliographic item format in the rule feature library, it converts the operation commands into a set of regular expression rules. At the same time, it generates a test sample set according to the sample features in the rule feature library, performs matching tests between the rule set and the sample set, analyzes the error type distribution of unmatched samples, and optimizes the parameters of the regular expression.

[0046] The comprehensive evaluation module is configured to verify and calculate the logical contradictions of the rules in the rule set and the coverage defects of the samples in the sample set based on the optimized regular expression rule set and the sample set, generate a comprehensive evaluation value, and activate the rule when the comprehensive evaluation value is higher than a preset quality threshold.

[0047] The rule update module is configured to encode the activated rules into a structured storage format according to feature dimensions, establish a version control index to record the rule generation time and change history, and synchronously set a change tracking mechanism to respond to and update the rule feature library.

[0048] The document review module is configured to respond to the target reference by calling the corresponding structured rules stored in the rule feature library to perform regular expression matching, performing multi-level authenticity verification on the regular expression matching results, and generating a review report including the verification results.

[0049] Based on the same concept, the present invention also provides an electronic device, including: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of a reference review method.

[0050] Based on the same concept, the present invention also provides a computer-readable storage medium storing a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform the steps of a reference review method.

[0051] Compared with existing technologies, its advantages are as follows:

[0052] This invention discloses a reference review method, system, electronic device, and storage medium, which can improve the accuracy and reliability of reference review, avoid the uncertainty of manual review, and improve review efficiency. Attached Figure Description

[0053] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0054] Figure 1 This is a flowchart illustrating some specific embodiments of the reference review method of the present invention;

[0055] Figure 2 This is one of the flowcharts of a reference review method according to another specific embodiment of the present invention;

[0056] Figure 3 This is a second flowchart illustrating a reference review method of the present invention in another specific embodiment;

[0057] Figure 4 This is the third flowchart of another specific embodiment of the reference review method of the present invention;

[0058] Figure 5 This is the fourth flowchart of another specific embodiment of the reference review method of the present invention;

[0059] Figure 6 This is a schematic diagram of the structure of a reference review system according to some specific embodiments of the present invention;

[0060] Figure 7 This is a schematic diagram of the structure of an electronic device according to some specific embodiments of the present invention;

[0061] In the diagram, 710 is the processor; 720 is the memory; 730 is the input device; and 740 is the output device. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0063] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the application. The singular forms “a,” “said,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0064] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0065] It should be understood that although the terms first, second, third, etc., may be used in the embodiments of this application, these descriptions should not be limited to these terms. These terms are only used to distinguish the descriptions. For example, first may also be referred to as second without departing from the scope of the embodiments of this application, and similarly, second may also be referred to as first.

[0066] Depending on the context, the words “if” or “suppose” as used here can be interpreted as “when” or “in response to determination” or “in response to detection.” Similarly, depending on the context, the phrases “if determination” or “if detection (of the stated condition or event)” can be interpreted as “when determination” or “in response to determination” or “when detection (of the stated condition or event)” or “in response to detection (of the stated condition or event).”

[0067] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that an article or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or device that includes said element.

[0068] It should be noted that any symbols and / or numbers present in the specification that are not marked in the accompanying drawings are not reference numerals.

[0069] Reference Figure 1 A method for reviewing references, including:

[0070] S101, in response to the captured publishing institution format specification change event and the citation feature dataset extracted from the academic database, the format specification change event and the citation feature dataset are fused and analyzed to dynamically construct a rule feature library, wherein the rule feature library includes the distribution pattern of bibliographic item formats;

[0071] S102, in response to the operation command of the user interface, based on the distribution pattern of the format of the recorded items in the rule feature library, the operation command is converted into a regular expression rule set, and a test sample set is generated according to the sample features in the rule feature library. The matching test between the rule set and the sample set is performed, the error type distribution of the unmatched samples is analyzed, and the parameters of the regular expression are optimized.

[0072] S103, based on the optimized regular expression rule set and the sample set, verify and calculate the logical contradictions of the rules in the rule set and the coverage defects of the samples in the sample set, generate a comprehensive evaluation value, and activate the rule when the comprehensive evaluation value is higher than the preset quality threshold.

[0073] S104, after activation, the rule is encoded into a structured storage format according to the feature dimension, a version control index is established to record the rule generation time and change history, and a change tracking mechanism is set up simultaneously to respond to and update the rule feature library;

[0074] S105, in response to the target reference, the corresponding structured rules stored in the rule feature library are called to perform regular expression matching, and the authenticity of the regular expression matching results is verified at multiple levels to generate a review report including the verification results.

[0075] Specifically, in this embodiment of the invention, in response to a change in reference format specifications announced on the official website of a publishing institution (e.g., the journal *Medical Research* changed the author name format from "surname + comma + space + first name abbreviation" to "surname + space + first name abbreviation"), the change event is captured by an automated program; simultaneously, a citation feature dataset is extracted from an academic database (e.g., analyzing the references of 100,000 CNKI papers, and statistically showing that the "Li,Y." format appears in the author field in 68% of cases, the "Li Y" format in 30% of cases, and the "Y.Li" format in 2% of cases), and the change event and the feature dataset are fused and analyzed: first, the key differences in the change event are extracted (e.g., the deletion of commas), and combined with the format distribution patterns in the citation feature dataset (e.g., the frequency of comma usage in the author field), a rule feature library with version identifier V2.1 is dynamically constructed. This feature library records the format distribution patterns of bibliographic items, including the new distribution pattern of the author field being {surname}{space}{first name abbreviation} in 100%, and the first letter of the title field being capitalized in 92% of cases (e.g., the title "Deep learning" matches the capitalization rule, while "deep learning" is considered an anomaly).

[0076] Users select the "No comma in author name" option through a visual interface and submit their operation instructions. Based on the distribution patterns of bibliographical entries in the rule feature library, the operation instructions are converted into a set of regular expression rules:

[0077] Original rules (?)<last_name> [^,]+),\s(?<first_initial> [AZ]\.);

[0078] Optimized to (?)<last_name> [^\s]+)\s(?<first_initial> [AZ]\.);

[0079] Simultaneously, a test sample set of 500 samples (including compliant samples such as "ZhangS." and "Wang H.", and erroneous samples such as "Zhang,S." and "Wang H") is generated based on the sample features in the rule feature library. After performing a matching test between the rule set and the sample set, the error type distribution of the unmatched samples is analyzed: 30% of the samples containing commas are unmatched, and 25% of the samples with name abbreviations without punctuation are unmatched. Based on this, the regular expression parameters are optimized: the name field capture group is expanded to (?).<first_initial> [AZ]\.?) to accommodate cases without punctuation.

[0080] Based on the optimized regular expression rule set and sample set, the logical contradictions of the rules in the rule set and the coverage defects of the samples in the sample set were verified and calculated: 2 logical contradictions in the rules themselves were detected (e.g., the rule requires "a space between the last name and the first name" but the sample contains "ZhangS." without a space), and 1 type of sample coverage defect was identified (the test sample set did not contain a variant of "full name spelling" such as "Zhang San"); a comprehensive evaluation value was generated: 2 logical contradictions, 87% sample coverage, and a comprehensive score of 89 points; when the comprehensive score is higher than the preset quality threshold of 85 points, the rule is activated. The activated rule is encoded into a structured storage format according to the feature dimension (e.g., stored as a JSON structure {"rule_id":"AUTH_V2.1","regex":"..."}), a version control index is established (recording version number: V2.1, parent version: V2.0, change summary: commas are removed), and a change tracking mechanism is set up synchronously: when a new specification requiring "the author's last name to be capitalized" is detected, the rule feature library is automatically updated to V2.2.

[0081] In response to the target reference "Zhang S. et al. Medical AI. 2023", the AUTH_V2.1 rules stored in the rule feature library are invoked to perform regular expression matching: capturing the author field "Zhang S."; multi-level verification of the matching results is performed: first, basic field verification is performed, and a "missing comma" deviation is detected compared with the academic database metadata (CNKI records the author as "Zhang,S."); semantic completion verification is performed for the uncommon abbreviation "et al.", and it is verified after being completed as "et alii"; finally, a transformation compatibility verification is performed to be compatible with the edit distance difference between "ZhangS" and "Zhang, S." (the calculated character difference value is 2, which is lower than the threshold of 3); a review report containing the verification results is generated: {Error location: author field, error type: format deviation, suggested correction: "Zhang, S."}.

[0082] In some applications, in response to captured publishing institution format change events and citation feature datasets extracted from academic databases, the format change events and citation feature datasets are fused and analyzed to dynamically construct a rule feature library. This rule feature library includes the distribution patterns of bibliographic item formats, including periodically scanning the reference format pages of publishing institution websites and identifying format change events through content hash comparison; analyzing the frequency and distribution patterns of author names and key fields in the citation feature datasets from the academic database; fusing the captured format change events with the analyzed frequency and distribution patterns of the key fields to generate a rule feature library with timestamp version identifiers; and when a format change event is detected, performing a comparison between the old and new versions and generating a rule feature library update instruction.

[0083] Understandably, in this application, in response to captured publishing institution format change events and academic database citation feature datasets, change detection is achieved by periodically scanning the reference format pages of publishing institution websites: an automated program is configured to scan the target journal's website (such as *Natural Science Reviews*) daily, calculate the hash value of the page text for comparison, and trigger a change event when a hash value mismatch is detected. For example, the scan found that the journal *Cell Research* had updated its reference format. The old format was "Author.Title[J].Journal Name, Year, Volume (Issue):Page Number", and the new format is "Author.Title[J].Journal Name, Publication Year, Volume Number (Issue Number):Start Page-End Page". The key change points are identified as "Year → Publication Year", "Volume (Issue) → Volume Number (Issue Number)", and "Page Number → Start Page-End Page". Meanwhile, citation feature datasets were extracted from academic databases for analysis: Statistical analysis of references from 500,000 documents revealed that in the author field, the format "last name + comma + first name abbreviation" (e.g., "Zhang,S.") accounted for 58%, and the format "last name + space + first name abbreviation" (e.g., "ZhangS.") accounted for 37%; in the volume / issue field, the format "numerical brackets" (e.g., "15(2)") accounted for 62%, and the format "text prefix" (e.g., "Vol.15,No.2") accounted for 35%. Feature fusion was performed between the captured change events and the analyzed distribution patterns: For changes in the author field, the data was mapped to the distribution pattern where "last name + comma + first name abbreviation" accounted for 58%; for changes in the volume / issue field, the data was mapped to the format "text prefix," which accounted for only 35% but was mandated by the new standard. Generate a rule feature library with timestamp version identifiers (e.g., version number ref_rules_v20250801). The author field records data that is compatible with both formats (comma format 58%, space format 37%), while the volume / issue field mandates the "volume number (issue number)" format (covering 35% of the data). When a change event is detected again, a difference comparison is performed: for example, the journal *Medical Frontiers* now requires "all uppercase author surnames." This is compared to the old feature library's feature where author surnames had no case sensitivity. An update instruction is generated, including the field change type, a description of the new rule, and the existing data compatibility coefficient (in this case, only 8% of the citation data conforms to the uppercase format). Ultimately, this triggers the rule feature library to update to the new version ref_rules_v20250810 and marks the new constraint on the author field.

[0084] Furthermore, in this application, when the format specification change event is detected, a comparison of the differences between the old and new versions is performed, and a rule feature library update instruction is generated. This includes locating the specific change points of the old and new format specification change events using a text difference algorithm; analyzing the impact range of the change points on the format distribution pattern of the recorded items in the rule feature library; reconstructing the format distribution model of the rule feature library based on the impact range, generating a new version identifier; writing the version change record into the historical trajectory of the rule feature library, and triggering an associated rule update notification.

[0085] Understandably, in this application, when a change in a publishing institution's formatting specifications is detected, a comparison of the old and new versions is performed, and an update instruction for the rule feature library is generated. First, the text difference algorithm is used to parse the old and new formatting specifications line by line to locate specific change points and mark the change type and location. For example, the journal *Environmental Science* changed the reference title format from "《Title》" to "Title" (removing the book title marks). The difference algorithm identifies that the deletion operation occurred at characters 5-15 of the 3rd line, and the change type is "format simplification". Next, the impact of this change point on the distribution pattern of bibliographic item formats in the rule feature library is analyzed: based on the citation feature dataset, statistics show that 82% of the literature uses the original book title mark format. After the new specification forcibly removes the book title marks, the distribution pattern of the title field needs to be recalculated. The impact covers all rule entries in the rule feature library that involve title format (such as title validation rules, title capture rules, and 12 other rules). The format distribution model of the rule feature library was reconstructed based on the scope of impact: the original "with book titles" distribution item (originally accounting for 82%) was deleted, and a new "without book titles" mandatory distribution item was added (initial coverage rate of 0%, requiring dynamic adaptation). A new version identifier "ref_rules_v20250901_EM1" was generated (where "EM" indicates an emergency update and "1" represents the first iteration). Finally, the version change record was written into the rule feature library history: recording the change time, the changed journal, the number of affected rules, and the data compatibility coefficient (in this case, the compatibility between the old and new data is only 18%), and triggering associated rule update notifications—sending warning emails to subscribers (containing a change summary and a list of affected rules), and simultaneously pushing rule update instructions to the review system via API (instruction code "RU_UPDATE_TITLE_FORMAT").

[0086] In some applications, in response to user interface operation commands, based on the distribution patterns of the bibliographic item formats in the rule feature library, the operation commands are converted into a set of regular expression rules. Simultaneously, a test sample set is generated based on the sample features in the rule feature library. Matching tests are performed between the rule set and the sample set. The error type distribution of unmatched samples is analyzed, and the parameters of the regular expression are optimized. This includes parsing the distribution patterns of the bibliographic item formats in the rule feature library and generating a regular expression template including named capture groups. Based on the regular expression template, user drag-and-drop and check-out operation commands in the interface are mapped to regular expression parameter configurations in real time. Based on the sample features recorded in the rule feature library, a test sample set covering multiple scenarios is constructed. Batch matching tests are performed between the rule set and the sample set. The error type distribution data of unmatched samples is statistically analyzed, and the character range definition and quantifier parameters of the named capture groups are adjusted based on the error type distribution data of the unmatched samples.

[0087] Understandably, in this application, responding to user interface commands, the rule set is constructed and optimized based on the distribution patterns of bibliographic entries in the rule feature library. First, the distribution patterns of bibliographic entries in the rule feature library are parsed to generate a regular expression template containing named capture groups. For example, if the distribution pattern of the author field in the rule feature library is "last name all uppercase + space + first name abbreviation" (e.g., "ZHANG S." accounts for 80%), then a basic template (?) is generated.<last_name> [AZ]+)\s(?<first_initial> [AZ]\.), where the last_name capture group corresponds to the name field, and the first_initial capture group corresponds to the name abbreviation field.

[0088] Based on this template, user operation commands in the interface are mapped to regular expression parameter configurations in real time: when the user drags the "Last Name Length Limit" slider to "3-8 characters" and checks the "Select Punctuation for First Name Abbreviation" option in the visual interface, the operation command is converted into parameter constraints—the last_name capture group is modified to [AZ]{3,8} to limit the character length, and the first_initial capture group is expanded to [AZ]\.? to be compatible with formats without punctuation (such as "S" and "S." can both be matched).

[0089] Simultaneously, a test sample set is constructed based on the sample features recorded in the rule feature library:

[0090] Positive examples (compliant with the new regulations):

[0091] "ZHANG S. Medical AI, 2025" (surname length 5 characters + initials with punctuation);

[0092] "WANG X Physics Review, 2024" (surname length 4 characters + initials without punctuation);

[0093] Counterexamples (covering common errors):

[0094] "Zhang S...." (surname not capitalized);

[0095] "LIU YJ..." (Extremely long initials);

[0096] The final test set contains 200 samples, covering 92% of the format variants in the feature library.

[0097] Perform batch matching tests between the rule set and the sample set:

[0098] Run the regular expression rule set to match all samples;

[0099] Distribution of error types for unmatched samples:

[0100] Error Type A: Last name not capitalized (accounting for 40%, such as "chen H." in the sample);

[0101] Error type B: Abbreviations without punctuation but required by the rules (accounting for 35%, such as "LEE T");

[0102] Error type C: Surname length exceeds the limit (accounting for 25%, such as "CHRISTOPHER K.")

[0103] Optimize parameters based on error distribution data:

[0104] For error type A: Expand the character range of the last_name capture group from [AZ] to [A-Za-z] for lowercase compatibility, but add a case conversion flag (to generate intermediate rules (?)).<last_name> [A-Za-z]+)\s→Automatically converts to uppercase when outputting);

[0105] For error type B: Modify the quantifier parameter, optimizing the first_initial group from [AZ]\. to [AZ]\.? (the question mark indicates matching 0 or 1 punctuation marks);

[0106] For error type C: Add a length validation sub-rule to trigger a warning when the number of characters in the captured name is greater than 8;

[0107] Comparison of rules before and after optimization:

[0108] Original rules:

[0109] (?<last_name> [AZ]{3,8})\s(?<first_initial> [AZ]\.);

[0110] Optimized rules:

[0111] (?<last_name> [A-Za-z]{3,8})\s(?<first_initial> [AZ]\.?);

[0112] Additional validation: if len(last_name)>8 then flag("Last name is too long");

[0113] Test and verification:

[0114] The original unmatched sample "chen H." was successfully matched after optimization (the last name was automatically capitalized to "CHEN").

[0115] Originally unmatched sample "LEE T" → successfully matched after optimization (compatible with samples without punctuation "T");

[0116] Sample "CHRISTOPHER K." → Captures the last name "CHRISTOPHER" (length 11) and triggers a warning.

[0117] In some applications, based on the optimized regular expression rule set and the sample set, logical contradictions in the rules in the rule set and coverage defects in the samples in the sample set are verified and calculated to generate a comprehensive evaluation value. When the comprehensive evaluation value is higher than a preset quality threshold, the rule is activated. This includes inputting the regular expression rule set and the associated sample set into a pre-trained verification model; verifying logical contradictions in the rules in the rule set, identifying blind spots in sample coverage and missing boundary conditions; calculating the comprehensive evaluation value of the logical defect rate and sample coverage rate of the rules in the rule set; activating and marking the rule when the comprehensive evaluation value is higher than the preset quality threshold; and extracting error type distribution data for the corresponding rule when the comprehensive evaluation value is lower than the preset quality threshold, reconstructing the process, and optimizing the parameters of the regular expression accordingly.

[0118] Understandably, in this application, based on the optimized regular expression rule set and sample set, rule validation and activation are performed. The regular expression rule set and its associated test sample set are input into a pre-trained validation model (a natural language processing model based on a deep learning architecture). This model verifies the self-consistency of the rules through syntax tree parsing and logical reasoning. For example, inputting the author field rule set (?<last_name> [A-Za-z]{3,8})\s(?<first_initial> [AZ]\.?) and 200 test samples (including positive example "ZHANGS." and negative example "chen H."), verify the model to perform the following operations:

[0119] Logical contradiction verification:

[0120] Detecting rule conflicts: If two constraints exist simultaneously in the rule set: Constraint A: Name abbreviations must contain punctuation (the first_initial group must match [AZ]\.) Constraint B: Punctuation is optional (the first_initial group is [AZ]\.?), the model marks the conflict type as "quantifier conflict" and outputs a conflict report: Conflict location: first_initial capture group; Conflict type: punctuation necessity conflict; Conflict rule ID: Rule_Author_2025-001;

[0121] Sample coverage defect identification:

[0122] Boundary condition missing detection: The test sample set does not contain the following boundary scenarios: authors with double last names (e.g., "O'BrienT.") or first names containing hyphens (e.g., "Smith-John A."). The model identifies blind spots and outputs a defect report: Missing sample type: compound last names; Missing proportion: 12% of the feature library distribution; Boundary condition: unprocessed apostrophes / hyphens;

[0123] Comprehensive evaluation numerical calculation: A weighted algorithm is used to generate a comprehensive score.

[0124] Logical defect rate (weight 60%): Defect rate = number of contradictory rules / total number of rules. For example: 2 contradictory rules were detected (out of a total of 50) → defect rate 4% → score 96 (weight value 57.6).

[0125] Sample coverage (weight 40%): Coverage = number of matched samples / total number of samples. For example: 184 matches out of 200 samples → coverage 92% → score 92 (weight value 36.8).

[0126] Overall score: 57.6 + 36.8 = 94.4 points (preset quality threshold 85 points);

[0127] Activation determination:

[0128] 94.4>85 → Activate the rule and mark it as effective;

[0129] Handling of non-compliance rules: When the overall score is 72 points (below 85 points):

[0130] Error type distribution: Major errors: Surname length exceeded the limit (25%) → sample "CHRISTOPHER K."; Incorrect first name abbreviation format (30%) → sample "LEE T".

[0131] Reconstruction and optimization process: For length exceeding the limit: Add a dynamic length validation sub-rule:

[0132] (?<last_name> [A-Za-z]{3,8})(?(?<=.{8})[^.]*|)#Exceeds 8 characters and triggers an exception; For name abbreviation errors: Extended character set compatible with numeric codes: -[AZ]\.?;+[A-Z0-9]\.?#Compatible with "ZHANG R1.".

[0133] In some applications, in response to a target reference, the corresponding structured rules in the rule feature library are invoked to perform regular expression matching. Multi-level authenticity verification is performed on the regular expression matching results, and a review report including the verification results is generated. This includes invoking the corresponding active rule in the rule feature library to perform regular expression matching based on the journal type of the target reference selected by the user; extracting metadata of successfully matched bibliographic entries and initiating authenticity verification for fields such as title, author, and journal name in an academic database; performing semantic completion on fields that fail to match, generating candidate metadata, and re-initiating authenticity verification; and generating a review report based on the authenticity verification results, including integrating format errors and authenticity deviation data to generate a structured review report containing error locations, error types, and correction schemes.

[0134] Understandably, in this application, in response to the target reference "Application of Multimodal Deep Learning in Medical Imaging" (authors: Zhang S. et al, journal: Medical AI Review, 2025, Vol.12, No.3:45-58), the activation rules stored in the rule feature library are invoked to perform regular expression matching: based on the index rule library of the journal type "Medical AI Review" selected by the user, the volume and issue field rules are loaded (?). <volume>Vol\.\d+),\s(? <issue>The `No.d+` function successfully captured the volume number "Vol.12" and the issue number "No.3", but the author field rule (?) failed.<last_name> [AZ][az]+)\s(?<first_initial> [AZ]\. The abbreviation "etal" was not matched. Metadata (title, volume, issue, page number) of the successfully matched bibliographic entry was extracted and used to initiate authenticity verification with the academic database: Data from CNKI was retrieved via a distributed query interface, revealing a title record as "MultimodalDeep Learning in Medical Imaging" (abbreviation not expanded), indicating a title abbreviation deviation; the volume and issue information matched the database metadata. Semantic completion was performed on the unmatched author field "et al": A pre-trained model was used to restore the complete bibliographic entry "Zhang S., Li M., Wang H.", and author authenticity verification was re-initiated, confirming that all three authors were present in the database's author list. A review report was generated based on the verification results: The report integrated the volume and issue field format correctness markers, the title field semantic deviation (detected as "uncommon abbreviation," suggested correction to the full name), and the author field completion verification results. The output structured report included the error location (title field), error type (authenticity deviation), correction scheme (expanding the title abbreviation), and author field completion details (the original input "et al" was completed into a list of three authors), while also indicating that the volume and issue field verification passed.

[0135] Furthermore, in this application, the multi-level authenticity verification includes basic field verification, which involves accurately comparing the metadata of bibliographic entries that have successfully matched the regular expressions with the corresponding data in the academic database to detect differences in title spelling and author name order; semantic completion verification, which involves completing the complete bibliographic metadata of fields that failed to match the regular expressions through semantic analysis and performing cross-database joint query verification; and variant compatibility verification, which involves performing multilingual spelling variant compatibility verification on the author name field to handle abbreviations, aliases, and cultural variation variants.

[0136] Understandably, in this application, multi-level authenticity verification is implemented through a three-level verification mechanism:

[0137] Basic field validation: The metadata of bibliographic entries that are successfully matched by regular expressions is compared with authoritative records in the distributed academic database cluster at the field level. For example, in the title field validation, a spelling discrepancy was detected between the target document title "Deep Learning Applications" and the database record "Deep Learning Applications" ("Learning" vs "Learning"). In the author field validation, a difference in the order of names was found (the order of "Zhang, S." in the database "SanZhang" is reversed).

[0138] Semantic Completion Verification: For fields where regular matching fails, use a pre-trained semantic model to complete the full metadata. For example, for the original input "Wang et al" in the author field, after completing it to the full bibliographic item "Wang H., Zhang L., Chen T.", perform a cross-database joint query (retrieve databases such as CNKI and Web of Science simultaneously) to verify the authenticity of the three authors' signatures in the target literature "Synthesis of Nanomaterials".

[0139] Morphological Compatibility Verification: Perform a multi-language spelling variant compatibility verification for the author name field. For example, process the variants of the English author "Smith, J.":

[0140] Abbreviation Compatibility: Verify that "J. Smith" and "Smith J." are the same author;

[0141] Alias Handling: Identify the alias relationship between "Robert Smith" and "Bob Smith";

[0142] Cultural Differences: The Chinese author "Zhang Wei" is compatible with the spelling variants "Zhang Wei" and "Wei Zhang"

[0143] Confirm the equivalence of the variants by calculating the string edit distance (set the threshold ≤ 2) and analyzing cultural naming conventions.

[0144] The following combines Figures 2 to 5 to illustrate another specific embodiment of a method for reviewing and proofreading references according to the present invention:

[0145] As Figure 2 and Figure 3 shown,

[0146] Use the requests library and BeautifulSoup library in Python to write a web crawler program to regularly visit the official websites of each journal and search for pages related to the requirements for reviewing and proofreading references. Parse the content of the crawled pages, extract the key information required for review and proofreading, such as reference format templates, bibliographic item requirements, etc., and store them in a database.

[0147] Collect and extract a large amount of citation data included in CNKI. The citation data includes information such as the title, author, journal name, year, volume number, and page number of the literature.

[0148] Use data analysis tools (such as the pandas library in Python) to analyze the citation data, count the occurrence frequencies and format characteristics of different bibliographic items, and extract common reference bibliographic items and format patterns.

[0149] Journal or publishing house reviewers can use a visual interface to set and build review rules.

[0150] Based on the citation data mined and analyzed from publishing big data, rules are constructed (regular expression writing). According to the obtained review requirements and analysis results, regular expressions are written for different reference types. For example, the expression for citation of journal articles is as follows, where author is the author entry, title is the journal article title, etc. The matched entry can be used not only for format verification, but also for subsequent error identification.

[0151] (? <author> [^\.]+)(? <symb1> \s*\.)(? <title>[^\.]+)(?< / title> <symb2> \[)(? <type> J)(? <symb3> \])(? <symb4> \.)(? <pubunit> [^,]+)(? <symb5> ,)(? <pubyear> [^,]+)((((? <symb6> ,)(? <volume> [^\(^\s]+))?(? <symb7> \()(? <issue> [^\)]+)(? <symb8> \)))|((? <symb11> ,)(? <volume1> [^\(^\s]+)((? <symb12> \()(? <issue1> [^\)]+)(? <symb13> \)))?))(? <symb9> :)(? <page> [^\.]+)(? <symb10>\.)

[0152] Customize build rules via visual interface:

[0153] It supports publishing professionals, such as editors or reviewers, in setting personalized rules, constructing them through a visual interface, and generating regular expressions through a backend program. The visual interface is built using front-end development technologies (HTML, CSS, JavaScript, etc.) and front-end frameworks (Vue.js or React). Interface elements such as a rule display area, rule editing area, reference input area, and review result display area are designed, and corresponding interactive functions are implemented.

[0154] Rule matching testing and optimization:

[0155] Collect a batch of reference samples and test their matching using a custom regular expression. Record the matching results and analyze the reasons for any unsuccessful matches.

[0156] Based on the test results, the regular expressions were adjusted and optimized, such as modifying the character range and adding quantifiers, to improve the matching accuracy and coverage of the rules.

[0157] Validation and storage of review rules:

[0158] Large model validation:

[0159] The constructed review rules are input into the pre-trained large model, and questions are posed to the model, such as "According to this rule, do the following references meet the requirements? [Example of references]".

[0160] The model's response is analyzed to determine the rationality and correctness of the rule. If the model's response does not match expectations, the rule is further analyzed and modified. The rule's accuracy is statistically evaluated; only when the accuracy reaches 90% or higher is the rule preliminarily confirmed to meet the requirements.

[0161] Rule storage:

[0162] Create a database table using a database management system (MySQL or the domestically developed CNKI database KBASE) to store the review rules. The rule table can include fields such as rule number, rule description, scope of application, and regular expression.

[0163] like Figure 4 As shown, the validated review rules are inserted into the database table for easy subsequent access and management.

[0164] like Figure 5 As shown, references are reviewed according to the rules, and database verification is performed based on the bibliographic entries matched by regular expressions.

[0165] Format proofreading;

[0166] The system provides a rule-calling function within the visual interface, allowing users to select appropriate rules for reference review. After the user inputs the references to be reviewed, the system retrieves the corresponding rules from the database and uses regular expressions to match and check the references. The review results are displayed in the review results area, providing prompts and suggestions for correction for any entries that do not conform to the rules. The system checks the format and citations of the references one by one according to the user-selected rules. For example, it checks whether the author's name format conforms to the rules and whether the journal name is correct. Any citations that do not conform to the rules are highlighted in red or otherwise prominently displayed in the review results area, along with suggested corrections.

[0167] Bibliographic item verification and proofreading:

[0168] Based on the bibliographic entries matched by regular expressions, relevant information is retrieved from academic databases (including bibliographic records indexed by CNKI). For example, the system searches for the existence of a document based on the author's name and title, and verifies the accuracy of the journal name. The system then uses the document title, year, and journal name to determine the author's correctness. If the regular expressions do not match any bibliographic entries, a publication model based on the Huazhi Big Data model is used for bibliographic entry identification, particularly effective for identifying authors, especially those with English authors. After identifying the bibliographic entries, their accuracy is verified. If the database search results are inconsistent with the information in the references, a corresponding prompt is displayed in the review results area, reminding the user to further verify and modify the information.

[0169] For the purpose of simplicity, the method steps disclosed in the above embodiments are described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0170] like Figure 6 As shown, the present invention also provides a reference review system, comprising:

[0171] The rule feature library construction module 201 is configured to respond to captured publishing institution format specification change events and citation feature datasets extracted from academic databases, and to perform fusion analysis on the format specification change events and the citation feature datasets to dynamically construct a rule feature library, wherein the rule feature library includes the distribution pattern of bibliographic item formats;

[0172] The matching test module 202 is configured to respond to operation instructions from the user interface, convert the operation instructions into a set of regular expression rules based on the distribution pattern of the bibliographic item format in the rule feature library, generate a test sample set according to the sample features in the rule feature library, perform matching tests between the rule set and the sample set, analyze the error type distribution of unmatched samples, and optimize the parameters of the regular expression.

[0173] The comprehensive evaluation module 203 is configured to verify and calculate the logical contradictions of the rules in the rule set and the coverage defects of the samples in the sample set based on the optimized regular expression rule set and the sample set, generate a comprehensive evaluation value, and activate the rule when the comprehensive evaluation value is higher than a preset quality threshold.

[0174] The rule update module 204 is configured to encode the activated rule into a structured storage format according to the feature dimension, establish a version control index to record the rule generation time and change history, and synchronously set a change tracking mechanism to respond to and update the rule feature library.

[0175] The document review module 205 is configured to respond to the target reference by calling the corresponding structured rules stored in the rule feature library to perform regular expression matching, performing multi-level authenticity verification on the regular expression matching results, and generating a review report including the verification results.

[0176] It is worth noting that although only some basic functional modules are disclosed in the embodiments of this invention, it does not mean that the composition of this system is limited to the above-mentioned basic functional modules. On the contrary, what this embodiment intends to express is that, based on the above-mentioned basic functional modules, those skilled in the art can arbitrarily add one or more functional modules in combination with existing technology to form an infinite number of embodiments or technical solutions. That is to say, this system is open rather than closed. The fact that this embodiment only discloses a few basic functional modules should not be considered as the scope of protection of the claims of this invention being limited to the disclosed basic functional modules. At the same time, for the convenience of description, the above device is described separately according to its functions as various units and modules. Of course, in implementing this invention, the functions of each unit and module can be implemented in one or more software and / or hardware.

[0177] like Figure 7 As shown, the present invention also provides an electronic device, including: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of a reference review method.

[0178] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. For example... Figure 7 The structure shown in this embodiment of the invention includes an electronic device comprising one or more processors 710 and a memory 720; the processors 710 in this electronic device may be one or more. Figure 7 Taking a processor 710 as an example; a memory 720 is used to store one or more programs; the one or more programs are executed by the one or more processors 710, so that the one or more processors 710 implement a reference review method as described in any one of the embodiments of the present invention.

[0179] The electronic device may also include an input device 730 and an output device 740.

[0180] The processor 710, memory 720, input device 730, and output device 740 in this electronic device can be connected via a bus or other means. Figure 7 Taking the example of a connection between China and Israel via a bus.

[0181] The memory 720 in this electronic device serves as a computer-readable storage medium, capable of storing one or more programs. These programs can be software programs, computer-executable programs, or modules, such as the program instructions / modules corresponding to the reference review method provided in this embodiment of the invention. The processor 710 executes various functional applications and data processing of the electronic device by running the software programs, instructions, and modules stored in the memory 720, thereby implementing the reference review method described in the above embodiment.

[0182] The memory 720 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device. Furthermore, the memory 720 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 720 may further include memory remotely located relative to the processor 710, which can be connected to the device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0183] Input device 730 can be used to receive input digital or character information, and to generate key signal inputs related to user settings and function control of the electronic device. Output device 740 may include display devices such as a display screen.

[0184] The present invention also provides a computer-readable storage medium storing a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform the steps of a reference review method.

[0185] Specifically, the computer storage medium in this embodiment of the invention can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. For example, a computer-readable storage medium can be—but is not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0186] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention. < / page> < / symb9> < / symb13> < / issue1> < / symb12> < / volume1> < / symb11> < / symb8> < / issue> < / symb7> < / volume> < / symb6> < / pubyear> < / symb5> < / pubunit> < / symb4> < / symb3> < / type> < / symb2> < / symb1> < / author> < / issue> < / volume>

Claims

1. A method for reviewing references, characterized in that, include: In response to captured publishing institution format specification change events and citation feature datasets extracted from academic databases, the format specification change events and citation feature datasets are fused and analyzed to dynamically construct a rule feature library, wherein the rule feature library includes the distribution pattern of bibliographic item formats; In response to user interface operation commands, based on the format distribution pattern of the bibliographic entries in the rule feature library, the operation commands are converted into a set of regular expression rules. At the same time, a test sample set is generated according to the sample features in the rule feature library. The matching test between the set of regular expression rules and the sample set is performed. The error type distribution of unmatched samples is analyzed and the parameters in the set of regular expression rules are optimized. Based on the optimized regular expression rule set and the sample set, the logical contradictions of the rules in the regular expression rule set and the coverage defects of the samples in the sample set are verified and calculated, and a comprehensive evaluation value is generated. When the comprehensive evaluation value is higher than a preset quality threshold, the rule is activated. The activated rules are encoded into a structured storage format according to feature dimensions, a version control index is established to record the rule generation time and change history, and a change tracking mechanism is set up synchronously to respond to and update the rule feature library; In response to the target reference, the corresponding structured rules stored in the rule feature library are invoked to perform regular expression matching, and the authenticity of the regular expression matching results is verified at multiple levels to generate a review report including the verification results.

2. The reference review method according to claim 1, characterized in that, In response to captured publishing institution format specification change events and citation feature datasets extracted from academic databases, the format specification change events and citation feature datasets are fused and analyzed to dynamically construct a rule feature library. The rule feature library includes the distribution patterns of bibliographic item formats and further includes: Periodically scan the reference format pages on the official websites of publishing institutions and identify format specification change events through content hash comparison; Analyze the frequency and distribution patterns of author names and key fields in the title structure of the citation feature dataset in the academic database. The captured format specification change events are fused with the frequency and distribution patterns of the main fields analyzed to generate a rule feature library with timestamp version identifiers. When a change in the format specification is detected, a comparison between the old and new versions is performed, and a rule feature library update instruction is generated.

3. The reference review method according to claim 1, characterized in that, In response to user interface operation commands, based on the format distribution patterns of the bibliographic entries in the rule feature library, the operation commands are converted into a set of regular expression rules. Simultaneously, a test sample set is generated according to the sample features in the rule feature library. A matching test is performed between the regular expression rule set and the sample set. The error type distribution of unmatched samples is analyzed, and the parameters of the regular expression are optimized. Further, this includes: Analyze the format distribution pattern of the bibliographic entries in the rule feature library to generate a regular expression template including named capture groups; Based on the regular expression template, the user's drag-and-drop and check-in operations on the interface are mapped to regular expression parameter configurations in real time. Based on the sample features recorded in the rule feature library, a test sample set covering multiple scenarios is constructed; Perform batch matching tests between the regular expression rule set and the sample set, statistically analyze the error type distribution data of unmatched samples, and adjust the character range definition and quantifier parameters of the named capture group based on the error type distribution data of the unmatched samples.

4. The reference review method according to claim 1, characterized in that, Based on the optimized regular expression rule set and the sample set, logical contradictions in the rules of the regular expression rule set and coverage defects in the samples of the sample set are verified and calculated to generate a comprehensive evaluation value. When the comprehensive evaluation value is higher than a preset quality threshold, the rule is activated, further including: The regular expression rule set and the associated sample set are input into the pre-trained validation model; Verify logical contradictions in the rules of the regular expression rule set, and identify blind spots in sample coverage and missing boundary conditions; Calculate the comprehensive evaluation value of the logical defect rate and sample coverage of the rules in the regular expression rule set. When the comprehensive evaluation value is higher than the preset quality threshold, the rule is activated and marked. When the comprehensive evaluation value is lower than the preset quality threshold, error type distribution data is extracted for the corresponding rules, the process is reconstructed, and the parameters of the regular expression are optimized accordingly.

5. The reference review method according to claim 1, characterized in that, In response to the target reference, the corresponding structured rules stored in the rule feature library are invoked to perform regular expression matching. Multi-level authenticity verification is performed on the regular expression matching results, and a review report including the verification results is generated, further including: Based on the journal type of the target reference selected by the user, the corresponding activation rule in the rule feature library is invoked to perform regular expression matching; Extract the metadata of successfully matched bibliographic entries and initiate authenticity verification of fields such as title, author, and journal name to the academic database; Perform semantic completion on fields that fail to match, generate candidate metadata, and then re-initiate authenticity verification; Based on the authenticity verification results, a review report is generated, which integrates data on format errors and authenticity deviations, and generates a structured review report containing the error location, error type, and correction plan.

6. A reference review method according to claim 2, characterized in that, When a format specification change event is detected, a comparison of the old and new versions is performed, and a rule feature library update instruction is generated, further including: The specific change points of the new and old format specification change events are located using text difference algorithms; Analyze the impact range of the change points on the distribution pattern of the recorded item formats in the rule feature library; Based on the scope of influence, reconstruct the format distribution model of the rule feature library and generate a new version identifier; The version change record is written into the historical trajectory of the rule feature library and the associated rule update notification is triggered.

7. A reference review method according to claim 5, characterized in that, The multi-level authenticity verification further includes: Basic field validation includes accurately comparing the metadata of bibliographic entries that have been successfully matched by regular expressions with the corresponding data in the academic database, and detecting differences in title spelling and author name order; Semantic completion verification includes completing the complete bibliographic metadata for fields that fail to match regular expressions through semantic analysis, and performing cross-database joint query verification; Variation compatibility checks include performing multilingual spelling variant compatibility checks on the author name field, handling abbreviations, aliases, and culturally different variants.

8. A reference review system, characterized in that, include: The rule feature library construction module is configured to respond to captured publishing institution format specification change events and citation feature datasets extracted from academic databases, and to perform fusion analysis on the format specification change events and the citation feature datasets to dynamically construct a rule feature library, wherein the rule feature library includes the distribution pattern of bibliographic item formats; The matching test module is configured to respond to operation commands from the user interface. Based on the distribution pattern of the bibliographic item format in the rule feature library, it converts the operation commands into a set of regular expression rules. At the same time, it generates a test sample set according to the sample features in the rule feature library, performs matching tests between the set of regular expression rules and the sample set, analyzes the error type distribution of unmatched samples, and optimizes the parameters of the regular expression. The comprehensive evaluation module is configured to verify and calculate the logical contradictions of the rules in the regular expression rule set and the coverage defects of the samples in the sample set based on the optimized regular expression rule set and the sample set, generate a comprehensive evaluation value, and activate the rule when the comprehensive evaluation value is higher than a preset quality threshold. The rule update module is configured to encode the activated rules into a structured storage format according to feature dimensions, establish a version control index to record the rule generation time and change history, and synchronously set a change tracking mechanism to respond to and update the rule feature library. The document review module is configured to respond to the target reference by calling the corresponding structured rules stored in the rule feature library to perform regular expression matching, performing multi-level authenticity verification on the regular expression matching results, and generating a review report including the verification results.

9. An electronic device, characterized in that, include: The system includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus; the memory stores a computer program, which, when executed by the processor, causes the processor to perform the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform the steps of the method according to any one of claims 1 to 7.

Images