Automatic optimization method and system for periodical checking process

Abstract

The invention relates to the technical field of periodical checking, and discloses an automatic optimization method and system of a periodical checking process, when a new manuscript is received, a periodical corresponding to submission information is automatically matched according to the submission information of the new manuscript, and the new manuscript is sent to a periodical terminal corresponding to the periodical; the periodical terminal obtains an initial manuscript reviewer set; according to the manuscript characteristics of the new manuscript and the interest association strength and the relationship matching strength between the new manuscript and each candidate manuscript reviewer, obtaining a potential association degree value between each candidate manuscript reviewer and the new manuscript; obtaining implicit knowledge relationship strength between the candidate reviewer and the new manuscript according to the potential association degree value between each candidate reviewer and the new manuscript; and optimizing the initial manuscript reviewer set according to the matching degree between the state space and the action space, obtaining a target manuscript reviewer set, and carrying out periodical review recommendation on the target manuscript reviewer set. According to the method, blindness and uncertainty in the manuscript reviewing process can be reduced, and the reviewing efficiency and quality are improved.

Description

Technical Field

[0001] The present invention belongs to the field of journal review and proofreading, and in particular relates to an automatic optimization method and system for a journal review and proofreading process. Background Art

[0002] In the academic publishing sector, journal proofreading is a critical component in ensuring the quality of scholarly output, upholding academic standards, and promoting academic progress. High-quality proofreading can identify errors, logical loopholes, and data in manuscripts, enhancing the journal's overall academic level and reputation, and providing a reliable foundation for scholarly exchange and knowledge dissemination. However, with the deepening of academic research and the increasing subdivision of disciplinary fields, the volume of manuscript submissions to journals has rapidly increased, posing unprecedented challenges to journal proofreading. On the one hand, the surge in manuscript volume has led to a heavy review workload, placing pressure on reviewers to process a large volume of manuscripts within a limited timeframe. This can lead to fatigue and negligence, which in turn compromises review quality. On the other hand, the interdisciplinary integration of disciplinary fields has led to an increasing degree of specialization in manuscripts, placing higher demands on reviewers' expertise and capabilities. Traditional reviewer selection methods often rely on the editor's subjective judgment and experience, failing to accurately match the manuscript's specialized field with the reviewer's interests and expertise. This results in inefficient reviewing and can even lead to unprofessional reviewers.

[0003] Currently, many journals select reviewers based on a simple match of the professional fields to which the manuscripts belong. Editors will screen out reviewers whose professional fields match the fields of the manuscripts from the reviewer database to form a set of candidate reviewers. However, the limitation of this method is that the professional field is only a broad classification and cannot accurately reflect the specific research directions and interests of the reviewers. For example, in the field of "medical informatics", the reviewer's research direction may cover medical image processing, medical big data analysis, electronic medical record management and other aspects, while the manuscript may only involve one specific sub-direction. If matching is performed solely based on professional fields, the selected reviewers may not be familiar with the content of the manuscript and may not be able to provide high-quality review opinions.

[0004] In addition to professional fields, factors such as reviewers' interests and relationships with authors and journals also affect the quality and efficiency of reviews. For example, if a reviewer is interested in the research topic of a manuscript, they may be more engaged in the review and provide more valuable suggestions. Furthermore, if a reviewer has a good cooperative relationship with the author or journal, they may be more willing to accept the review assignment, and the review process will be smoother. However, existing reviewer selection methods often ignore these factors and fail to comprehensively consider the strength of the reviewer's interest association and relationship matching, resulting in less accurate reviewer selection.

[0005] Therefore, this application provides an automatic optimization method for journal review process to solve the above technical problems. Summary of the Invention

[0006] The purpose of the present invention is to provide a method and system for automatically optimizing the journal review process, so as to solve the technical problem that the existing technology does not comprehensively consider the reviewer's interest association strength and relationship matching strength, resulting in inaccurate reviewer selection.

[0007] In order to solve the above technical problems, the present invention provides a method for automatically optimizing the journal review process, comprising: When a new manuscript is received, the journal corresponding to the submission information is automatically matched based on the submission information of the new manuscript, and the new manuscript is sent to the journal terminal corresponding to the journal; and the journal terminal retrieves candidate reviewers based on the journal field to which the new manuscript belongs to obtain an initial reviewer set; Obtaining a potential correlation value between each candidate reviewer and the new manuscript based on the manuscript features of the new manuscript and the interest correlation strength and relationship matching strength between each candidate reviewer and the new manuscript; wherein, the first observation manuscript and the second observation manuscript are screened out using the difference between the L2 norm between the review feature vector of each candidate reviewer and the manuscript feature vector of the new manuscript and the L2 norm corresponding to the manuscript that each candidate reviewer has completed reviewing, and the relationship matching strength is set using the cosine similarity relationship between the manuscript feature vectors of the first observation manuscript and the second observation manuscript and the manuscript feature vector of the new manuscript; Obtaining the implicit knowledge relationship strength between each candidate reviewer and the new manuscript based on the potential correlation value between each current candidate reviewer and the new manuscript; The action space of each candidate reviewer is set according to their historical review data, and the state space of the reviewer is set using the implicit knowledge relationship strength corresponding to the candidate reviewer. The initial reviewer set is optimized according to the matching degree between the state space and the action space to obtain the target reviewer set, and journal review recommendations are made for the target reviewer set.

[0008] Furthermore, when a new manuscript is received, the journal corresponding to the submission information is automatically matched based on the submission information of the new manuscript, and the new manuscript is sent to the journal terminal corresponding to the journal; and the journal terminal retrieves candidate reviewers based on the journal field to which the new manuscript belongs, and obtains an initial reviewer set, including: When a new manuscript is received, the submission information of the new manuscript is retrieved and the journal to which the author wants to submit is determined based on the submission information; Sending the new manuscript to the journal terminal corresponding to the journal according to the name of the journal; The journal terminal extracts information from the new manuscript to obtain metadata corresponding to the new manuscript, wherein the metadata includes technical subdivision fields and keyword fields; Classify the new manuscript according to the metadata corresponding to the new manuscript, and obtain the journal field to which the new manuscript belongs; The reviewers corresponding to the journal field to which the new manuscript belongs are retrieved from the database to form an initial reviewer set.

[0009] Furthermore, the potential correlation value between each candidate reviewer and the new manuscript is obtained based on the manuscript features of the new manuscript and the interest correlation strength and relationship matching strength between the new manuscript and each candidate reviewer, including: Retrieving the manuscript features of the new manuscript and generating a manuscript feature vector using the manuscript features of the new manuscript; wherein the manuscript features of the new manuscript include the average weight of keywords, the distribution ratio of core research content, the ratio of theoretical content to experimental content, and the ratio of references; Retrieving the review features of the candidate reviewers and generating a review feature vector using the review features of the candidate reviewers; wherein the review features of the candidate reviewers include average review time, average review quality score, rejection rate, and review extension rate; The manuscript feature vector of the new manuscript and the review feature vector of each candidate reviewer are used to obtain the potential correlation value corresponding to each reviewer.

[0010] Furthermore, the potential correlation value is obtained by the following steps, including: Retrieve the manuscript feature vector of the new manuscript; Retrieve the manuscript feature vector of each candidate reviewer's reviewed manuscript; Using the manuscript feature vector of the new manuscript and the manuscript feature vectors of the manuscripts reviewed by each candidate reviewer, obtaining the interest association strength between the new manuscript and each candidate reviewer; Retrieve the review feature vector of each candidate reviewer; Obtaining the relationship matching strength between the new manuscript and each candidate reviewer using the manuscript feature vector of the new manuscript and the review feature vector of each candidate reviewer; The interest association strength and relationship matching strength between the new manuscript and each candidate reviewer are used to obtain a potential association value corresponding to the new manuscript and each reviewer.

[0011] Furthermore, the relationship matching strength is obtained through the following steps, including: The L2 norm between the review feature vector of each candidate reviewer and the manuscript feature vector of the new manuscript is obtained by using the review feature vector of each candidate reviewer and the manuscript feature vector of the new manuscript as the target norm; Retrieve the L2 norm of each candidate reviewer's completed manuscript as the observation norm; Performing difference processing on the target norm and each observation norm to obtain the difference between the target norm and each observation norm; comparing the difference between the target norm and each observed norm with a preset difference reference value; The completed manuscript corresponding to the observation norm that is not less than the preset difference reference value is used as the first observation manuscript; The completed manuscript corresponding to the observation norm lower than the preset difference reference value is used as the second observation manuscript; The potential correlation value corresponding to the new manuscript and each reviewer is obtained by using the manuscript feature vectors corresponding to the first observed manuscript and the second observed manuscript and the manuscript feature vector of the new manuscript.

[0012] Furthermore, the implicit knowledge relationship strength between each candidate reviewer and the new manuscript is obtained based on the potential correlation value between each candidate reviewer and the new manuscript, including: Retrieve the potential correlation value between each current candidate reviewer and the new manuscript; Retrieve the local dynamic knowledge graph corresponding to the journal review process from the database; The weight parameters between the candidate reviewers and the technical sub-fields and key fields to which the new manuscript belongs are obtained from the local dynamic knowledge graph, and the implicit knowledge relationship strength between each candidate reviewer and the new manuscript is obtained using the weight parameters.

[0013] Furthermore, weight parameters between candidate reviewers and the technical sub-fields and key fields to which the new manuscript belongs are obtained from the local dynamic knowledge graph, and the implicit knowledge relationship strength between each candidate reviewer and the new manuscript is obtained using the weight parameters, including: Retrieving, from the local dynamic knowledge graph, a weight coefficient between the candidate reviewer and the technical sub-field to which the new manuscript belongs as a first weight parameter; Retrieve key fields from the new manuscript; Compare the key fields in the new manuscript with the local dynamic knowledge graph to obtain the key fields existing in the local knowledge graph; Retrieve the weight coefficient between the key fields in the local knowledge graph and the candidate reviewers as the second weight parameter; The implicit knowledge relationship strength between each candidate reviewer and the new manuscript is obtained according to the potential correlation value in combination with the first weight parameter and the second weight parameter.

[0014] Furthermore, the action space of each candidate reviewer is set based on their historical review data, and the state space of the reviewer is set using the implicit knowledge relationship strength corresponding to the candidate reviewer. The initial reviewer set is optimized based on the matching degree between the state space and the action space to obtain the target reviewer set, and journal review recommendations are made for the target reviewer set, including: Set the action space of each candidate reviewer based on their historical review data; Setting the reviewer's state space using the implicit knowledge relationship strength between each candidate reviewer and the new manuscript; Determining whether the action space of the candidate reviewer and the state space of the reviewer meet the matching requirements; The candidate reviewer whose action space and state space of the candidate reviewer meet the matching requirements is selected as the target reviewer; Target reviewers are screened to generate a target reviewer set, and the target reviewer set is used to optimize the initial reviewer set, and the target reviewer set is recommended for journal review.

[0015] Furthermore, determining whether the action space of the candidate reviewer and the state space of the reviewer meet the matching requirements includes: Retrieve each element contained in the action space of the candidate reviewer; Retrieving each element contained in the state space of the candidate reviewer; When each element contained in the candidate reviewer's action space and each element contained in the reviewer's state space meet the preset constraint conditions, it is determined that the candidate reviewer's action space and the reviewer's state space meet the matching requirements.

[0016] Based on the same concept, the present invention also provides an automatic optimization system for journal review process, comprising: The automatic journal matching and initial reviewer set acquisition module is used to automatically match the journal corresponding to the submission information of the new manuscript based on the submission information of the new manuscript when receiving the new manuscript, and send the new manuscript to the journal terminal corresponding to the journal; and the journal terminal retrieves candidate reviewers based on the journal field to which the new manuscript belongs to obtain the initial reviewer set; a potential correlation value acquisition module, configured to obtain a potential correlation value between each candidate reviewer and the new manuscript based on the manuscript features of the new manuscript and the interest correlation strength and relationship matching strength between each candidate reviewer and the new manuscript; wherein the first observation manuscript and the second observation manuscript are screened out using the difference between the L2 norm between the review feature vector of each candidate reviewer and the manuscript feature vector of the new manuscript and the L2 norm corresponding to the manuscripts that each candidate reviewer has completed reviewing, and the relationship matching strength is set using the cosine similarity relationship between the manuscript feature vectors of the first observation manuscript and the second observation manuscript and the manuscript feature vector of the new manuscript; An implicit knowledge relationship strength acquisition module is used to acquire the implicit knowledge relationship strength between each candidate reviewer and the new manuscript based on the potential correlation value between each candidate reviewer and the new manuscript; The optimization matching module is used to set the action space of candidate reviewers based on their historical review data, and to set the state space of reviewers using the implicit knowledge relationship strength corresponding to the candidate reviewers. The initial reviewer set is optimized according to the matching degree between the state space and the action space, and the target reviewer set is obtained. The target reviewer set is then recommended for journal review.

[0017] Compared with the prior art, the beneficial effects are: The present invention discloses a method and system for automatically optimizing the journal review process. By comprehensively considering the journal field to which a new manuscript belongs, the manuscript characteristics, the reviewer's interest correlation strength and the relationship matching strength, it can more comprehensively and accurately evaluate the compatibility of the reviewer and the manuscript, thereby significantly improving the accuracy of reviewer matching. The introduction of implicit knowledge relationship strength can effectively discover potential high-quality reviewers who have deep knowledge reserves in professional fields but may not be fully noticed by traditional methods. These reviewers may not be included in the conventional reviewer database due to their unique research direction or lack of fame, but this method can discover the potential connection between them and the manuscript, significantly improving the accuracy of reviewer identification. Utilizing the idea of ​​reinforcement learning, the action space is set according to the reviewer's historical review data, and the state space is set in combination with the implicit knowledge relationship strength. The matching degree between the two is calculated to optimize the reviewer set. This dynamic optimization mechanism can adjust the reviewer set in real time according to the reviewer's actual performance and the characteristics of the manuscript, ensuring that the reviewer always maintains a high review quality and efficiency. By accurately optimizing and recommending reviewer sets, we can reduce blindness and uncertainty in the review process and improve review efficiency. At the same time, suitable reviewers can more accurately identify problems in the manuscript and provide valuable revision suggestions, thereby improving the quality of review. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Other features, objects and advantages of the present application will become more apparent upon reading the detailed description of non-limiting embodiments made with reference to the following drawings: Figure 1 It is a flowchart of some specific embodiments of a method for automatically optimizing a journal review process according to the present invention; Figure 2 This is a system block diagram of some specific embodiments of an automatic optimization system for a journal review process of the present invention; Figure 3 It is a structural diagram of a local dynamic knowledge graph of an automatic optimization method for a journal review process of the present invention. DETAILED DESCRIPTION

[0019] To make the objectives, technical solutions, and advantages of this application more clear, this application will be further described in detail below with reference to the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of this application.

[0020] The terms used in the examples of this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms "a," "the," and "the" used in the examples of this application and the appended claims are also intended to include plural forms, and unless the context clearly indicates otherwise, "a plurality" generally includes at least two.

[0021] It should be understood that the term "and / or" as used herein is merely a description of the relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.

[0022] It should be understood that although the terms first, second, third, etc. may be used to describe in the embodiments of the present application, these descriptions should not be limited to these terms. These terms are only used to distinguish the descriptions. For example, without departing from the scope of the embodiments of the present application, the first may also be referred to as the second, and similarly, the second may also be referred to as the first.

[0023] As used herein, the words "if" and "if" may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to the determination" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)," depending on the context.

[0024] It should also be noted that the terms "include," "comprises," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a product or device comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such product or device. In the absence of further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the product or device comprising the element.

[0025] It should be noted in particular that any symbols and / or numbers in the specification that are not marked in the accompanying drawings are not drawing marks.

[0026] Reference Figure 1 , a method for automatically optimizing the journal review process, including: S101, when a new manuscript is received, the journal corresponding to the submission information of the new manuscript is automatically matched based on the submission information, and the new manuscript is sent to the journal terminal corresponding to the journal; and the journal terminal retrieves candidate reviewers based on the journal field to which the new manuscript belongs to obtain an initial reviewer set; S102, obtaining a potential correlation value between each candidate reviewer and the new manuscript based on the manuscript features of the new manuscript and the interest correlation strength and relationship matching strength between each candidate reviewer and the new manuscript; wherein, the first observation manuscript and the second observation manuscript are screened out using the difference between the L2 norm between the review feature vector of each candidate reviewer and the manuscript feature vector of the new manuscript and the L2 norm corresponding to the manuscripts that each candidate reviewer has completed reviewing, and the relationship matching strength is set using the cosine similarity relationship between the manuscript feature vectors of the first observation manuscript and the second observation manuscript and the manuscript feature vector of the new manuscript; S103, obtaining the implicit knowledge relationship strength between each candidate reviewer and the new manuscript based on the potential correlation value between each candidate reviewer and the new manuscript; S104: Set the action space of each candidate reviewer based on their historical review data, and set the state space of the reviewer using the implicit knowledge relationship strength corresponding to the candidate reviewer. Optimize the initial reviewer set based on the matching degree between the state space and the action space, obtain the target reviewer set, and recommend journals for the target reviewer set.

[0027] Specifically, in an embodiment of the present invention, when a new manuscript is received, the submission information of the new manuscript is retrieved, and the journal to which the author wants to submit is determined based on the submission information; the new manuscript is sent to the journal terminal corresponding to the journal according to the name of the journal; the journal terminal extracts information from the new manuscript and obtains metadata corresponding to the new manuscript, wherein the metadata includes technical sub-fields and keyword fields; the new manuscript is classified according to the metadata corresponding to the new manuscript, and the journal field to which the new manuscript belongs is obtained; and reviewers corresponding to the journal field to which the new manuscript belongs are retrieved from the database to form an initial reviewer set.

[0028] It is understandable that in the above embodiment, when it is determined through the submission information of the new manuscript that the author has determined the specific journal for submission, then the journal terminal is determined according to the journal determined by the author; when the author does not clearly specify the specific journal for submission in the submission information of the new manuscript, then the keywords and abstract information contained in the submission information are retrieved, and the journal names that match them are screened out for the author to choose based on the keywords and abstract information, and the journal selected by the author is recorded, and the keywords and abstract information are determined according to the journal selected by the author; and the new manuscript is sent to the journal terminal corresponding to the journal to which it belongs according to the determined journal; and the journal terminal performs information extraction on the new manuscript, and obtains the metadata corresponding to the manuscript, which mainly covers technical sub-fields and keyword fields. The new manuscript is classified according to the extracted metadata to clarify the journal field to which it belongs. From the pre-constructed database, the reviewers corresponding to the journal field to which the new manuscript belongs are retrieved, thereby forming an initial reviewer set.

[0029] The new manuscript features are thoroughly analyzed, while also considering the strength of interest association and relationship match between each candidate reviewer and the new manuscript. Interest association strength is based on factors such as the reviewer's previous research interests and the alignment of their published work with the manuscript's topic. Relationship match strength considers any prior collaborations and academic exchanges between the reviewer, author, and journal. Taking these factors into account, a potential association value is calculated between each candidate reviewer and the new manuscript, reflecting the potential compatibility between the reviewer and the manuscript at both the professional and relationship levels. Based on this calculated potential association value, the strength of the implicit knowledge relationship between the reviewer and the new manuscript is further explored. Implicit knowledge relationships encompass the reviewer's depth of knowledge in a specific field, their innovative capabilities, and their potential connections to the knowledge covered in the manuscript. An action space is established for each candidate reviewer based on their historical review data. This action space encompasses the reviewer's review behaviors, such as review speed and the level of detail provided in their comments. The state space of the reviewer is set using the implicit knowledge relationship strength corresponding to the candidate reviewer. The state space reflects the reviewer's knowledge state and ability level in the current manuscript context. The matching degree between the state space and the action space is calculated, and the initial reviewer set is optimized based on the matching degree to finally obtain the target reviewer set. The optimized target reviewer set is recommended for journal review, and the manuscript is assigned to a suitable reviewer for review. The present invention, by comprehensively considering factors in multiple dimensions, can accurately match reviewers across interdisciplinary fields, ensuring that manuscripts from different disciplinary backgrounds can receive professional and comprehensive review.

[0030] In some of the applications, the potential correlation value between each candidate reviewer and the new manuscript is obtained based on the manuscript features of the new manuscript and the interest association strength and relationship matching strength between the new manuscript and each candidate reviewer, including retrieving the manuscript features of the new manuscript and generating a manuscript feature vector using the manuscript features of the new manuscript; wherein the manuscript features of the new manuscript include the average keyword weight, the core research content distribution ratio, the ratio of theoretical content to experimental content and the reference ratio; retrieving the review features of the candidate reviewers and generating a review feature vector using the review features of the candidate reviewers; wherein the review features of the candidate reviewers include the average review time, the average review quality score, the rejection rate and the review extension rate; and obtaining the potential correlation value corresponding to each reviewer using the manuscript feature vector of the new manuscript and the review feature vector of each candidate reviewer.

[0031] Specifically, this application first retrieves the manuscript features of the new manuscript. These features include the average keyword weight, the distribution ratio of core research content (i.e., the ratio of core research content words to the total number of words), the ratio of theoretical content to experimental content, and the proportion of references. Based on these retrieved manuscript features, they are converted into a manuscript feature vector. A feature vector is a mathematical representation that can represent the multi-dimensional characteristics of a manuscript in a vector form, facilitating subsequent calculations and analysis. Simultaneously, review features are retrieved for each candidate reviewer, including the average review time, average review quality score, rejection rate, and review extension rate. Using these review features, a review feature vector is generated for each candidate reviewer, also quantifying the reviewer's review characteristics in vector form. Using the generated new manuscript feature vector and the review feature vectors of each candidate reviewer, an algorithm is used to determine the potential correlation value for each reviewer. This value comprehensively reflects the potential connection between the reviewer and the new manuscript at the feature level.

[0032] As can be seen, by converting manuscript and reviewer features into vector form, manuscript and reviewer characteristics are precisely quantified, avoiding the uncertainty of subjective judgment and providing a more objective and accurate description of manuscript and reviewer characteristics. The selected manuscript and reviewer features cover multiple dimensions, including manuscript content structure (core research content distribution, theoretical to experimental content ratio), academic value (average keyword weighting, reference ratio), and reviewer efficiency (average review time, review delay rate) and quality (average review quality score, rejection rate). This allows for accurate assessment of reviewer compatibility with new manuscripts and improves the accuracy of potential correlation values. Using feature vectors to calculate potential correlation values ​​leverages the efficiency of vector operations to quickly determine the potential correlation between each reviewer and the new manuscript. This allows for the rapid screening and ranking of a large number of candidate reviewers, improving reviewer selection efficiency and shortening the review cycle. By comparing the potential correlation values ​​between different reviewers and new manuscripts, we can intuitively see which reviewers have a higher match with the manuscripts, and thus give priority to these reviewers for review.

[0033] In some applications, the potential correlation value is obtained through the following steps, including retrieving the manuscript feature vector of the new manuscript; retrieving the manuscript feature vector of the manuscript reviewed by each candidate reviewer; using the manuscript feature vector of the new manuscript and the manuscript feature vector of the manuscript reviewed by each candidate reviewer to obtain the interest correlation strength between the new manuscript and each candidate reviewer; retrieving the review feature vector of each candidate reviewer; using the manuscript feature vector of the new manuscript and the review feature vector of each candidate reviewer to obtain the relationship matching strength between the new manuscript and each candidate reviewer; and using the interest correlation strength and relationship matching strength between the new manuscript and each candidate reviewer to obtain the potential correlation value corresponding to the new manuscript and each reviewer.

[0034] Specifically, in this application, the interest association strength is obtained by the following formula:

[0035] Among them, Int(p) represents the interest association strength between the new manuscript and each candidate reviewer; n represents the number of manuscripts reviewed by each candidate reviewer; p represents the manuscript feature vector of the new manuscript; p i represents the manuscript feature vector corresponding to the i-th reviewed manuscript; t now Indicates the time when the new manuscript is received; t i represents the receiving time corresponding to the i-th reviewed manuscript; λ represents a preset time decay coefficient, and the value range of the time decay coefficient is 0.2-1.4; Represents the cosine similarity between the manuscript feature vector of the new manuscript and the manuscript feature vector corresponding to the i-th reviewed manuscript; The potential correlation value between the new manuscript and each reviewer is obtained by the following formula:

[0036] Among them, Rel(p, r) represents the potential correlation value corresponding to the new manuscript and each reviewer; p represents the manuscript feature vector of the new manuscript; r represents the review feature vector of the candidate reviewer; M represents the relationship matching strength between the new manuscript and each candidate reviewer; x represents the interest sensitivity coefficient, and the value range of the interest sensitivity coefficient is 0.4-0.7.

[0037] First, retrieve the manuscript feature vector of the new manuscript and the manuscript feature vector of each candidate reviewer’s reviewed manuscript. Using the manuscript feature vector of the new manuscript and the manuscript feature vector of each candidate reviewer’s reviewed manuscript, calculate the interest association strength between the new manuscript and each candidate reviewer through the formula. The formula takes into account the number of candidate reviewers’ reviewed manuscripts (n), the feature vectors of the new manuscript and the reviewed manuscripts (p and p i), the time of receipt of new manuscripts and reviewed manuscripts (t now and t i ), a preset time decay coefficient (λ), and the cosine similarity between the feature vectors of the two. The time decay coefficient reflects the impact of time on the strength of the interest association, while the cosine similarity measures the degree of feature similarity between the new manuscript and the reviewed manuscript. The review feature vector of each candidate reviewer is retrieved, along with the manuscript feature vector of the new manuscript. The manuscript feature vector of the new manuscript and the review feature vector of each candidate reviewer are used to calculate the relationship matching strength between the new manuscript and each candidate reviewer. After obtaining the interest association strength and relationship matching strength between the new manuscript and each candidate reviewer, another formula is used to calculate the potential association value corresponding to the new manuscript and each reviewer. This formula takes into account the interest association strength (Int(p)), relationship matching strength (M), the new manuscript feature vector (p), the candidate reviewer's review feature vector (r), and the interest sensitivity coefficient (x). The interest sensitivity coefficient is used to adjust the weight of the interest association strength in the calculation of the potential association value.

[0038] As can be understood, by considering the feature vectors of the candidate reviewer's previously reviewed manuscripts, time factors, and cosine similarity, we can more accurately measure the strength of the interest association between a new manuscript and the candidate reviewer. The introduction of a time decay coefficient makes the influence of recently reviewed manuscripts on the interest association strength greater, which is more realistic, as reviewers' interests and research directions may change over time. Calculating the relationship match strength by combining the manuscript feature vector of the new manuscript with the review feature vector of the candidate reviewer comprehensively considers the degree of match between the manuscript and the reviewer in multiple aspects, providing a more comprehensive basis for calculating the potential association value. By utilizing the interest association strength and relationship match strength, and adjusting them with the interest sensitivity coefficient, we can reasonably calculate the potential association value for the new manuscript and each reviewer. The restricted range of the interest sensitivity coefficient ensures that the interest association strength plays an appropriate role in the calculation of the potential association value, avoiding over-reliance on any one factor. This ensures that the potential association value more accurately reflects the overall compatibility between the new manuscript and the reviewer. Based on the calculated potential association value, appropriate reviewers can be selected more accurately. By comprehensively considering interest associations and relationship matching, we ensure that selected reviewers are both interested in the manuscript topic and possess the appropriate reviewing skills and experience, thereby improving review quality and efficiency. By factoring in time, this method is adaptable to dynamic changes in reviewers' interests and research directions. Reviewers' interests may shift over time, and this method can promptly reflect these changes, ensuring accurate and effective reviewer selection.

[0039] In some applications, the relationship matching strength is obtained through the following steps, including using the review feature vector of each candidate reviewer and the manuscript feature vector of the new manuscript to obtain the L2 norm between the review feature vector of each candidate reviewer and the manuscript feature vector of the new manuscript as the target norm; retrieving the L2 norm corresponding to the manuscript completed by each candidate reviewer as the observation norm; performing difference processing on the target norm and each observation norm to obtain the difference between the target norm and each observation norm; comparing the difference between the target norm and each observation norm with a preset difference reference value; taking the completed manuscript corresponding to the observation norm not lower than the preset difference reference value as the first observation manuscript; taking the completed manuscript corresponding to the observation norm lower than the preset difference reference value as the second observation manuscript; using the manuscript feature vectors corresponding to the first observation manuscript and the second observation manuscript and the manuscript feature vector of the new manuscript to obtain the potential correlation value corresponding to the new manuscript and each reviewer.

[0040] Specifically, in this application, the relationship matching strength between the new manuscript and each candidate reviewer is obtained by the following formula:

[0041] Where M represents the matching strength between the new manuscript and each candidate reviewer; a represents the number of first-observed manuscripts; b represents the number of second-observed manuscripts; P ai represents the manuscript feature vector of the first observed manuscript of i; P bi represents the manuscript feature vector of the i-th second observation manuscript. Sim(p, p ai ) represents the cosine similarity between the manuscript feature vector of the new manuscript and the manuscript feature vector of the i-th first observation manuscript; Sim(p, p bi ) represents the cosine similarity between the manuscript feature vector of the new manuscript and the manuscript feature vector of the i-th second observation manuscript; Sim(p, r) represents the cosine similarity between the manuscript feature vector of the new manuscript and the review feature vector of the candidate reviewer; Using the review feature vector of each candidate reviewer and the manuscript feature vector of the new manuscript, the L2 norm between the two is calculated and used as the target norm. The L2 norm is a mathematical method for measuring the length or distance of a vector and is used here to quantify the degree of difference between the review feature vector and the manuscript feature vector. The L2 norm corresponding to the manuscript that each candidate reviewer has completed review is retrieved as the observation norm. The target norm is subtracted from each observation norm to obtain the difference between the target norm and each observation norm. The difference between the target norm and each observation norm is compared with a preset difference reference value. Completed manuscripts corresponding to observation norms that are not lower than the preset difference reference value are classified as first observation manuscripts; completed manuscripts corresponding to observation norms that are lower than the preset difference reference value are classified as second observation manuscripts. Using the manuscript feature vectors corresponding to the first and second observation manuscripts and the manuscript feature vector of the new manuscript, the relationship matching strength between the new manuscript and each reviewer is calculated using a formula. The formula takes into account the number of the first observation manuscript (a), the number of the second observation manuscript (b), and the manuscript feature vectors of the first observation manuscript and the second observation manuscript (P ai and P bi ).

[0042] As can be understood, calculating the target norm and the observed norm quantifies the difference between the reviewer's feature vector and the manuscript's feature vector, providing a basis for subsequent manuscript classification and relationship matching strength calculation. By comparing the difference between the target norm and the observed norm with a preset reference difference, the reviewed manuscripts are appropriately classified into the first observed manuscript and the second observed manuscript. This allows for distinguishing manuscripts with significantly different features from the new manuscript, and those with lesser differences. The relationship matching strength calculated using the above steps and formula more accurately measures the relationship matching between the new manuscript and each reviewer. This allows for better matching of new manuscripts with appropriate reviewers during the reviewer selection process, effectively improving review quality and efficiency. This method is adaptable to new manuscripts with diverse characteristics. By considering the differences in the reviewer's previous review features, this method can calculate a reasonable relationship matching strength, regardless of whether the new manuscript is unique or shares some similarity with previous manuscripts, providing an effective reference for reviewer selection.

[0043] like Figure 3As shown, in some of the applications, the implicit knowledge relationship strength between each candidate reviewer and the new manuscript is obtained based on the potential correlation value between each current candidate reviewer and the new manuscript, including retrieving the potential correlation value between each current candidate reviewer and the new manuscript; retrieving the local dynamic knowledge graph corresponding to the journal review process from the database; obtaining the weight parameters between the technical sub-fields and key fields to which the candidate reviewers and the new manuscript belong from the local dynamic knowledge graph, and using the weight parameters to obtain the implicit knowledge relationship strength between each candidate reviewer and the new manuscript.

[0044] Furthermore, in this application, weight parameters between the candidate reviewers and the technical sub-fields and key fields to which the new manuscript belongs are obtained from the local dynamic knowledge graph, and the weight parameters are used to obtain the implicit knowledge relationship strength between each candidate reviewer and the new manuscript, including retrieving the weight coefficient between the candidate reviewers and the technical sub-field to which the new manuscript belongs from the local dynamic knowledge graph as a first weight parameter; retrieving the key fields in the new manuscript; comparing the key fields in the new manuscript with the local dynamic knowledge graph to obtain the key fields existing in the local knowledge graph; retrieving the weight coefficient between the key fields existing in the local knowledge graph and the candidate reviewers as a second weight parameter; and obtaining the implicit knowledge relationship strength between each candidate reviewer and the new manuscript based on the potential correlation value combined with the first weight parameter and the second weight parameter.

[0045] Specifically, in this application, the implicit knowledge relationship strength is obtained by the following formula:

[0046] Where S(p,r) represents the strength of the implicit knowledge relationship between the candidate reviewer and the new manuscript; Rel(p,r) represents the potential correlation value between the new manuscript and each reviewer; w d and w r denote the first weight parameter and the second weight parameter respectively; and, .

[0047] First, retrieve the potential correlation value between each current candidate reviewer and the new manuscript. Retrieve the local dynamic knowledge graph corresponding to the journal's review process from the database. Retrieve the weight coefficient between the candidate reviewer and the technical sub-field to which the new manuscript belongs from the local dynamic knowledge graph and use it as the first weight parameter. This weight coefficient reflects the reviewer's expertise or influence in a specific technical sub-field. Retrieve the key fields in the new manuscript; these key fields can summarize the core content and research direction of the new manuscript. Compare the key fields in the new manuscript with the local dynamic knowledge graph to identify the key fields in the local knowledge graph. Retrieve the weight coefficient between the key fields in the local knowledge graph and the candidate reviewer as the second weight parameter. This weight coefficient reflects the reviewer's familiarity with or depth of research related to these key fields. Based on the potential correlation value, combined with the first and second weight parameters, calculate the strength of the implicit knowledge relationship between each candidate reviewer and the new manuscript.

[0048] As can be understood, in this application, the introduction of a local dynamic knowledge graph allows for a deeper exploration of the implicit knowledge relationships between candidate reviewers and new manuscripts. These implicit knowledge relationships are not based solely on superficial feature matching but also take into account the reviewer's expertise and experience in specific technical fields and key fields, resulting in a more comprehensive and in-depth assessment of the relationship between the reviewer and the new manuscript. By calculating the strength of the implicit knowledge relationship using the potential correlation value combined with the first and second weight parameters, the compatibility between the reviewer and the new manuscript can be more accurately measured. During the reviewer selection process, reviewers who not only match the new manuscript in terms of features but also have a high degree of expertise and experience are selected, thereby improving review quality and efficiency. The local dynamic knowledge graph provides data support for deriving the weight parameters. By retrieving relevant information from the knowledge graph, existing knowledge resources can be fully utilized, duplication of construction and data collection efforts can be avoided, and the efficiency and reliability of the system can be improved. In complex technical fields, new manuscripts may involve multiple sub-fields and key fields. By comprehensively considering the weighting parameters of technical sub-fields and key fields, this method can better adapt to the characteristics of complex technical fields, accurately assess the relationship between reviewers and new manuscripts, and ensure that reviewers have sufficient expertise and capabilities to review new manuscripts. The local dynamic knowledge graph dynamically adjusts as knowledge is updated and developed. Therefore, this method can adapt to changes in knowledge, promptly reflecting the reviewers' latest expertise and experience in different technical fields and key fields, and ensuring the accuracy and timeliness of reviewer selection.

[0049] In some of these applications, the action space of the candidate reviewers is set according to the historical review data of each candidate reviewer, and the state space of the reviewers is set using the implicit knowledge relationship strength corresponding to the candidate reviewers, the initial reviewer set is optimized according to the matching degree between the state space and the action space, the target reviewer set is obtained, and the target reviewer set is recommended for journal review, including setting the action space of the candidate reviewers according to the historical review data of each candidate reviewer; setting the state space of the reviewers using the implicit knowledge relationship strength between each candidate reviewer and the new manuscript; determining whether the action space of the candidate reviewers and the state space of the reviewers meet the matching requirements; taking the candidate reviewers whose action space of the candidate reviewers and the state space of the reviewers meet the matching requirements as target reviewers; screening the target reviewers to generate a target reviewer set, and using the target reviewer set to optimize the initial reviewer set, and recommending the target reviewer set for journal review.

[0050] Specifically, in this application, the action space structure of the candidate reviewer is as follows:

[0051] Among them, A represents the action space structure of the candidate reviewers; k represents the number of technical sub-fields included in the number of manuscripts reviewed by each candidate reviewer; Gi represents the proportion of reviewed manuscripts corresponding to the i-th technical sub-field; Yi represents the review extension rate corresponding to the i-th technical sub-field; u01 represents the first normalization coefficient, which is used to convert Projected to the interval [0, 1]; Δt represents the average review time; t represents the preset reference time, ranging from 20 days to 60 days; L represents the rejection rate; The state space structure of the reviewer is as follows:

[0052] Where B represents the state space of the reviewer; u02 represents the second normalization coefficient, which is used to convert Projected to the interval [0, 1]; Q represents the implicit knowledge relationship strength between each candidate reviewer and the new manuscript; Qp represents the average implicit knowledge relationship strength between the candidate reviewer and the manuscripts reviewed by the candidate reviewer; J represents the proportion of the number of times the new manuscript author appears in the manuscripts reviewed by the candidate reviewer; Sp represents the cosine similarity between the manuscript feature vector of the new manuscript and the review feature vector of the candidate reviewer and the cosine similarity between the review feature vector of the candidate reviewer and the manuscript feature vector of the corresponding reviewed manuscript; u03 represents the third normalization coefficient, which is used to convert Project to the interval [0, 1]; Each candidate reviewer's action space is set based on their historical review data. This includes the number of technical sub-fields covered in their reviewed manuscripts (k), the proportion of reviewed manuscripts in each technical sub-field (Gi), and the review delay rate (Yi). These metrics are projected to the [0, 1] interval using the first normalization coefficient (u01). The action space also includes the relationship between the average review time (Δt) and the preset reference time (t), as well as the rejection rate (L). These factors comprehensively reflect the candidate reviewer's work performance, efficiency, and expertise distribution during their historical review process. The reviewer's state space is set using the strength of the implicit knowledge relationship between each candidate reviewer and the new manuscript. The state space structure consists of multiple elements. The second normalization coefficient (u02) projects the implicit knowledge relationship strength (Q) to the [0, 1] range. The average implicit knowledge relationship strength corresponding to the candidate reviewer's reviewed manuscripts (Qp), the proportion of the new manuscript's author appearances in the candidate reviewer's reviewed manuscripts (J), and the difference (Sp) between the cosine similarity between the new manuscript's feature vector and the candidate reviewer's review feature vector and the cosine similarity between the candidate reviewer's review feature vector and the feature vector of the reviewed manuscripts (Sp) are also considered. This difference is projected to the [0, 1] range using the third normalization coefficient (u03). These elements comprehensively reflect the state information between the candidate reviewer and the new manuscript in terms of knowledge association, author association, and feature similarity. The candidate reviewer's action space and the reviewer's state space are then determined to determine whether they meet the matching requirements. Candidate reviewers who meet the matching requirements are selected as target reviewers. The target reviewers are screened to generate a target reviewer set, which is then used to optimize the initial reviewer set. Finally, the target reviewer set is recommended for journal review.

[0053] As can be understood, the action space takes into account candidate reviewers' historical review data, including their distribution across technical fields, review efficiency (delay rate, average review time), and rejection rate. This allows for a comprehensive assessment of candidate reviewers' professional competence, work efficiency, and review quality. The state space, based on the strength of the implicit knowledge relationship between candidate reviewers and the new manuscript, accurately measures the compatibility between candidate reviewers and the new manuscript in terms of knowledge, authors, and characteristics, ensuring that selected reviewers are well-suited to the new manuscript. By matching the action and state spaces, reviewers with both a strong historical review performance and a high degree of compatibility with the new manuscript can be selected, thereby improving the accuracy of reviewer selection and reducing review quality issues caused by inappropriate reviewer selection. Optimizing the initial reviewer set using the target reviewer set can improve the efficiency and quality of the review process. Reviewers in the target reviewer set are more likely to complete their reviews quickly and accurately, shortening the review cycle and improving the overall quality of the journal. This method comprehensively considers multiple factors and can adapt to the review needs of new manuscripts of different types and fields. Whether it is a new manuscript with complex technology or a case with special reviewer requirements, it can ensure the selection of suitable reviewers by properly setting the action space and state space and optimizing the recommendation based on the matching degree.

[0054] In some of the applications, determining whether the action space of the candidate reviewer and the state space of the reviewer meet the matching requirements includes retrieving each element contained in the action space of the candidate reviewer; retrieving each element contained in the state space of the candidate reviewer; when each element contained in the action space of the candidate reviewer and each element contained in the state space of the reviewer meet the preset constraints, then determining that the action space of the candidate reviewer and the state space of the reviewer meet the matching requirements.

[0055] Specifically, in this application, the constraints are as follows:

[0056] Where A1, A2, and A3 represent the first, second, and third elements in the candidate reviewer's action space, respectively; B1, B2, and B3 represent the first, second, and third elements in the candidate reviewer's state space, respectively; and θ1, θ2, and θ3 represent the preset thresholds corresponding to the relationships between the first, second, and third elements in the candidate reviewer's action and state spaces, respectively. These thresholds are obtained by fitting historical review data.

[0057] Each element in the candidate reviewer's action space and state space is retrieved separately. Action space elements reflect the candidate reviewer's characteristics based on historical review data, such as their review status in specific technical fields and review efficiency indicators. State space elements reflect the candidate reviewer's status with the new manuscript, such as knowledge connections and author relationships. Each element in the action space (A1, A2, A3) is compared with each corresponding element in the state space (B1, B2, B3) to determine whether they meet the preset constraints. The preset constraints define the relationship between the corresponding elements in the action and state spaces, and compliance is determined using preset thresholds (θ1, θ2, θ3). The preset thresholds are fitted based on historical review data and are determined based on the relationships between elements in the action and state spaces from a large number of past review cases. When each element in the action space and each element in the state space meet the preset constraints, the candidate reviewer's action space and the reviewer's state space are considered to be matched.

[0058] As can be understood, the matching requirements between the action space and the state space are quantified through the preset constraints and thresholds. This effectively improves the objectivity and accuracy of the decision-making process, avoids the uncertainty caused by subjective judgment, and enhances the scientific nature of reviewer selection. The preset thresholds are fitted based on historical review data, and these matching criteria are derived from successful review cases. Using these historically data-based thresholds for matching decisions can better adapt to actual review requirements and increase the probability of selecting suitable reviewers. Matching requirements are only considered met when all corresponding elements in the action space and state space meet the constraints. This decision-making method accurately selects reviewers with both a good historical review performance and a high degree of compatibility with the new manuscript, thereby improving review quality and efficiency. Clear matching criteria and thresholds based on historical data make the reviewer selection process more stable and predictable. The same criteria and thresholds are applied to each reviewer selection, reducing variability in reviewer selection due to human or random factors, effectively improving the overall stability of the journal's review process. This technical solution is adaptable to complex review scenarios because the action and state spaces contain information across multiple dimensions, and constraints are tailored to these dimensions. Whether facing new manuscripts in different technical fields or with special reviewer requirements, appropriate reviewers can be selected through reasonable matching decisions.

[0059] For the method steps disclosed in the above embodiments, for the purpose of simple description, the method steps are expressed as a series of action combinations. However, those skilled in the art should be aware that the embodiments of the present invention are not limited by the order of the actions described, because according to the embodiments of the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present invention.

[0060] like Figure 2 As shown, the present invention also provides an automatic optimization system for a journal review process, comprising: The journal automatic matching and initial reviewer set acquisition module 201 is used to automatically match a journal corresponding to the submission information of the new manuscript based on the submission information of the new manuscript when receiving the new manuscript, and send the new manuscript to the journal terminal corresponding to the journal; and the journal terminal retrieves candidate reviewers based on the journal field to which the new manuscript belongs to obtain an initial reviewer set; The potential correlation value acquisition module 202 is configured to acquire a potential correlation value between each candidate reviewer and the new manuscript based on the manuscript features of the new manuscript and the interest correlation strength and relationship matching strength between each candidate reviewer and the new manuscript; wherein the first observation manuscript and the second observation manuscript are selected by using the difference between the L2 norm between the review feature vector of each candidate reviewer and the manuscript feature vector of the new manuscript and the L2 norm corresponding to the manuscripts that each candidate reviewer has completed reviewing, and the relationship matching strength is set by using the cosine similarity relationship between the manuscript feature vectors of the first observation manuscript and the second observation manuscript and the manuscript feature vector of the new manuscript; The implicit knowledge relationship strength acquisition module 203 is used to acquire the implicit knowledge relationship strength between each candidate reviewer and the new manuscript based on the potential correlation value between each candidate reviewer and the new manuscript; The optimization matching module 204 is used to set the action space of the candidate reviewer based on the historical review data of each candidate reviewer, and set the state space of the reviewer using the implicit knowledge relationship strength corresponding to the candidate reviewer, optimize the initial reviewer set according to the matching degree between the state space and the action space, obtain the target reviewer set, and recommend journal review to the target reviewer set.

[0061] It is worth noting that although only some basic functional modules are disclosed in the embodiment of the present invention, it does not mean that the composition of the present system is limited to the above basic functional modules. On the contrary, what this embodiment wants to express is that on the basis of the above basic functional modules, those skilled in the art can arbitrarily add one or more functional modules in combination with the existing technology to form an infinite number of embodiments or technical solutions. In other words, this system is open rather than closed. Just because this embodiment only discloses individual basic functional modules, it cannot be considered that the scope of protection of the claims of the present invention is limited to the disclosed basic functional modules. At the same time, for the convenience of description, the above devices are described in terms of functions, which are divided into various units and modules. Of course, when implementing the present invention, the functions of each unit and module can be implemented in the same or one or more software and / or hardware.

[0062] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements 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.

Claims

1. A method for automatically optimizing a journal review process, characterized in that: include: When a new manuscript is received, the journal corresponding to the submission information is automatically matched based on the submission information of the new manuscript, and the new manuscript is sent to the journal terminal corresponding to the journal; and the journal terminal retrieves candidate reviewers based on the journal field to which the new manuscript belongs to obtain an initial reviewer set; Obtaining a potential correlation value between each candidate reviewer and the new manuscript based on the manuscript features of the new manuscript and the interest correlation strength and relationship matching strength between each candidate reviewer and the new manuscript; wherein, the first observation manuscript and the second observation manuscript are screened out using the difference between the L2 norm between the review feature vector of each candidate reviewer and the manuscript feature vector of the new manuscript and the L2 norm corresponding to the manuscript that each candidate reviewer has completed reviewing, and the relationship matching strength is set using the cosine similarity relationship between the manuscript feature vectors of the first observation manuscript and the second observation manuscript and the manuscript feature vector of the new manuscript; Obtaining the implicit knowledge relationship strength between each candidate reviewer and the new manuscript based on the potential correlation value between each current candidate reviewer and the new manuscript; The action space of each candidate reviewer is set according to their historical review data, and the state space of the reviewer is set using the implicit knowledge relationship strength corresponding to the candidate reviewer. The initial reviewer set is optimized according to the matching degree between the state space and the action space to obtain the target reviewer set, and journal review recommendations are made for the target reviewer set.

2. The automatic optimization method for a journal review process according to claim 1 is characterized in that: When a new manuscript is received, the journal corresponding to the submission information is automatically matched based on the submission information of the new manuscript, and the new manuscript is sent to the journal terminal corresponding to the journal; and the journal terminal retrieves candidate reviewers based on the journal field to which the new manuscript belongs, and obtains an initial reviewer set, including: When a new manuscript is received, the submission information of the new manuscript is retrieved and the journal to which the author wants to submit is determined based on the submission information; Sending the new manuscript to the journal terminal corresponding to the journal according to the name of the journal; The journal terminal extracts information from the new manuscript to obtain metadata corresponding to the new manuscript, wherein the metadata includes technical subdivision fields and keyword fields; Classify the new manuscript according to the metadata corresponding to the new manuscript, and obtain the journal field to which the new manuscript belongs; The reviewers corresponding to the journal field to which the new manuscript belongs are retrieved from the database to form an initial reviewer set.

3. The automatic optimization method for a journal review process according to claim 1 is characterized in that: According to the manuscript features of the new manuscript and the interest association strength and relationship matching strength between the new manuscript and each candidate reviewer, a potential association value between each candidate reviewer and the new manuscript is obtained, including: Retrieving the manuscript features of the new manuscript and generating a manuscript feature vector using the manuscript features of the new manuscript; wherein the manuscript features of the new manuscript include the average weight of keywords, the distribution ratio of core research content, the ratio of theoretical content to experimental content, and the ratio of references; Retrieving the review features of the candidate reviewers and generating a review feature vector using the review features of the candidate reviewers; wherein the review features of the candidate reviewers include average review time, average review quality score, rejection rate, and review extension rate; The manuscript feature vector of the new manuscript and the review feature vector of each candidate reviewer are used to obtain the potential correlation value corresponding to each reviewer.

4. The automatic optimization method for a journal review process according to claim 1, characterized in that: The potential correlation value is obtained by the following steps, including: Retrieve the manuscript feature vector of the new manuscript; Retrieve the manuscript feature vector of each candidate reviewer's reviewed manuscript; Using the manuscript feature vector of the new manuscript and the manuscript feature vectors of the manuscripts reviewed by each candidate reviewer, obtaining the interest association strength between the new manuscript and each candidate reviewer; Retrieve the review feature vector of each candidate reviewer; Obtaining the relationship matching strength between the new manuscript and each candidate reviewer using the manuscript feature vector of the new manuscript and the review feature vector of each candidate reviewer; The interest association strength and relationship matching strength between the new manuscript and each candidate reviewer are used to obtain a potential association value corresponding to the new manuscript and each reviewer.

5. The method for automatically optimizing a journal review process according to claim 1 or 4, characterized in that: The relationship matching strength is obtained through the following steps, including: The L2 norm between the review feature vector of each candidate reviewer and the manuscript feature vector of the new manuscript is obtained by using the review feature vector of each candidate reviewer and the manuscript feature vector of the new manuscript as the target norm; Retrieve the L2 norm of each candidate reviewer's completed manuscript as the observation norm; Performing difference processing on the target norm and each observation norm to obtain the difference between the target norm and each observation norm; comparing the difference between the target norm and each observed norm with a preset difference reference value; The completed manuscript corresponding to the observation norm that is not less than the preset difference reference value is used as the first observation manuscript; The completed manuscript corresponding to the observation norm lower than the preset difference reference value is used as the second observation manuscript; The potential correlation value corresponding to the new manuscript and each reviewer is obtained by using the manuscript feature vectors corresponding to the first observed manuscript and the second observed manuscript and the manuscript feature vector of the new manuscript.

6. The method for automatically optimizing a journal review process according to claim 1, characterized in that: Obtain the implicit knowledge relationship strength between each candidate reviewer and the new manuscript based on the potential correlation value between each candidate reviewer and the new manuscript, including: Retrieve the potential correlation value between each current candidate reviewer and the new manuscript; Retrieve the local dynamic knowledge graph corresponding to the journal review process from the database; The weight parameters between the candidate reviewers and the technical sub-fields and key fields to which the new manuscript belongs are obtained from the local dynamic knowledge graph, and the implicit knowledge relationship strength between each candidate reviewer and the new manuscript is obtained using the weight parameters.

7. The method for automatically optimizing a journal review process according to claim 6, characterized in that: Obtaining weight parameters between candidate reviewers and the technical sub-fields and key fields to which the new manuscript belongs from the local dynamic knowledge graph, and using the weight parameters to obtain the implicit knowledge relationship strength between each candidate reviewer and the new manuscript, including: Retrieving, from the local dynamic knowledge graph, a weight coefficient between the candidate reviewer and the technical sub-field to which the new manuscript belongs as a first weight parameter; Retrieve key fields from the new manuscript; Compare the key fields in the new manuscript with the local dynamic knowledge graph to obtain the key fields existing in the local knowledge graph; Retrieve the weight coefficient between the key fields in the local knowledge graph and the candidate reviewers as the second weight parameter; The implicit knowledge relationship strength between each candidate reviewer and the new manuscript is obtained according to the potential correlation value in combination with the first weight parameter and the second weight parameter.

8. The method for automatically optimizing a journal review process according to claim 1, characterized in that: The action space of each candidate reviewer is set based on their historical review data, and the state space of the reviewer is set using the implicit knowledge relationship strength corresponding to the candidate reviewer. The initial reviewer set is optimized based on the matching degree between the state space and the action space to obtain the target reviewer set, and journal review recommendations are made for the target reviewer set, including: Set the action space of each candidate reviewer based on their historical review data; Setting the reviewer's state space using the implicit knowledge relationship strength between each candidate reviewer and the new manuscript; Determining whether the action space of the candidate reviewer and the state space of the reviewer meet the matching requirements; The candidate reviewer whose action space and state space of the candidate reviewer meet the matching requirements is selected as the target reviewer; Target reviewers are screened to generate a target reviewer set, and the target reviewer set is used to optimize the initial reviewer set, and the target reviewer set is recommended for journal review.

9. The method for automatically optimizing a journal review process according to claim 8, characterized in that: Determining whether the candidate reviewer's action space and the reviewer's state space meet matching requirements includes: Retrieve each element contained in the action space of the candidate reviewer; Retrieving each element contained in the state space of the candidate reviewer; When each element contained in the candidate reviewer's action space and each element contained in the reviewer's state space meet the preset constraint conditions, it is determined that the candidate reviewer's action space and the reviewer's state space meet the matching requirements.

10. An automatic optimization system for journal review process, characterized by: include: The automatic journal matching and initial reviewer set acquisition module is used to automatically match the journal corresponding to the submission information of the new manuscript based on the submission information of the new manuscript when receiving the new manuscript, and send the new manuscript to the journal terminal corresponding to the journal; and the journal terminal retrieves candidate reviewers based on the journal field to which the new manuscript belongs to obtain the initial reviewer set; a potential correlation value acquisition module, configured to obtain a potential correlation value between each candidate reviewer and the new manuscript based on the manuscript features of the new manuscript and the interest correlation strength and relationship matching strength between each candidate reviewer and the new manuscript; wherein the first observation manuscript and the second observation manuscript are screened out using the difference between the L2 norm between the review feature vector of each candidate reviewer and the manuscript feature vector of the new manuscript and the L2 norm corresponding to the manuscripts that each candidate reviewer has completed reviewing, and the relationship matching strength is set using the cosine similarity relationship between the manuscript feature vectors of the first observation manuscript and the second observation manuscript and the manuscript feature vector of the new manuscript; An implicit knowledge relationship strength acquisition module is used to acquire the implicit knowledge relationship strength between each candidate reviewer and the new manuscript based on the potential correlation value between each candidate reviewer and the new manuscript; The optimization matching module is used to set the action space of candidate reviewers based on their historical review data, and to set the state space of reviewers using the implicit knowledge relationship strength corresponding to the candidate reviewers. The initial reviewer set is optimized according to the matching degree between the state space and the action space, and the target reviewer set is obtained. The target reviewer set is then recommended for journal review.

Images