Quantitative matching-based solicited object recommendation method, system, device and medium

Abstract

The application relates to the technical field of academic journal intelligent management, and discloses a manuscript solicitation object recommendation method, system, device and medium based on quantitative matching. The method comprises the following steps: acquiring multi-source data of each candidate object, constructing research direction labels, academic influence labels and historical cooperation behavior labels of each candidate object according to the multi-source data; acquiring a manuscript solicitation demand text, and extracting a manuscript solicitation direction label according to the manuscript solicitation demand text; determining a research direction adaptation degree according to the research direction labels and the manuscript solicitation direction label, determining an academic authority degree according to the academic influence labels, determining a historical cooperation degree according to the historical cooperation behavior labels, and comprehensively calculating a comprehensive recommendation score according to the research direction adaptation degree, the academic authority degree and the historical cooperation degree; and preliminarily sorting the candidate objects according to the comprehensive recommendation score, detecting conflicts in the preliminary sorting, and generating a recommendation sorting after eliminating the conflict objects. The method realizes intelligent screening and sorting recommendation of manuscript solicitation objects.

Description

Technical Field

[0001] This application relates to the field of intelligent management technology for academic journals, specifically to a method, system, device, and medium for recommending manuscript solicitors based on quantitative matching. Background Technology

[0002] In the academic journal publishing process, commissioning manuscripts for specific topics is a crucial step in enhancing the journal's academic influence and industry recognition. High-quality commissioning of manuscripts for specific topics requires inviting experts and scholars with deep research experience, strong academic authority, and a high willingness to collaborate as contributors. However, in current technology, the selection of contributors for specific topics mainly relies on the editor's manual experience, which presents the following technical problems: First, manual screening is inefficient. Quickly identifying potential contributors is crucial, but with a vast pool of experts, editors struggle to cover all suitable candidates. The screening process can take days or even weeks, severely hindering the timeliness of commissioning articles for special issues.

[0003] Second, the matching accuracy is insufficient. Human judgment relies heavily on the editor's subjective impressions and limited knowledge, making it difficult to quantify the suitability between experts and the research topic. This frequently results in discrepancies between recommended research directions and the topic's needs, and a mismatch between the expert's academic achievements and the topic's theme. Consequently, the research background of the commissioned authors becomes disconnected from the topic's requirements, impacting manuscript quality.

[0004] Third, there is a lack of systematic evaluation indicators. Existing methods do not comprehensively consider multiple factors such as the academic authority of experts and the effectiveness of past collaborations, leading to unstable recommendation results and significant fluctuations in manuscript response rates and manuscript quality. Some highly influential experts may have extremely low response rates, while young scholars who respond actively may go unnoticed.

[0005] Therefore, there is an urgent need for a technical solution that can quantify multi-dimensional matching indicators and achieve intelligent and accurate recommendations in order to overcome the limitations of the existing manual recommendation model and improve the efficiency and quality of topic compilation. Summary of the Invention

[0006] To address the aforementioned issues, this application provides a method, system, device, and medium for recommending commissioned manuscripts based on quantitative matching. By constructing a multi-dimensional quantitative evaluation system and a precise matching model, it achieves intelligent screening and ranking recommendation of commissioned manuscripts.

[0007] The technical solution adopted in this application is as follows: Firstly, this application provides a method for recommending commissioned manuscripts based on quantitative matching, including: Object tagging: Obtain multi-source data for each candidate object, and construct research direction tags, academic influence tags, and historical collaboration behavior tags for each candidate object based on the multi-source data; Demand feature construction: Obtain the manuscript request text and extract manuscript direction tags based on the manuscript request text; Multi-dimensional matching: The research direction fit is determined based on the research direction tag and the manuscript solicitation direction tag; the academic authority is determined based on the academic influence tag; the historical cooperation behavior tag is determined based on the historical cooperation behavior tag; and the comprehensive recommendation score is determined by comprehensively calculating the research direction fit, academic authority and historical cooperation. Generate recommendation results: The candidate objects are initially ranked according to the comprehensive recommendation score. After conflict detection and removal of conflicting objects, the recommendation ranking is generated.

[0008] Secondly, this application also provides a quantitative matching-based recommendation system for commissioned manuscripts, including: The object tagging unit is used to obtain multi-source data for each candidate object and construct research direction tags, academic influence tags, and historical cooperation behavior tags for each candidate object based on the multi-source data. The demand feature construction unit is used to obtain the manuscript request text and extract the manuscript direction tags based on the manuscript request text; The multi-dimensional matching unit is used to determine the research direction suitability based on research direction tags and manuscript solicitation tags, the academic authority based on academic influence tags, the historical cooperation level based on historical cooperation behavior tags, and to calculate the comprehensive recommendation score based on the research direction suitability, academic authority, and historical cooperation level. The recommendation result generation unit is used to initially rank candidate objects based on the comprehensive recommendation score, perform conflict detection on the initial ranking and remove conflicting objects to generate the recommendation ranking.

[0009] Thirdly, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for recommending commissioned manuscripts based on quantitative matching.

[0010] Fourthly, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method for recommending commissioned manuscripts based on quantitative matching.

[0011] The above-mentioned technical solution adopted in this application can achieve the following beneficial effects: I. Improving Recommendation Accuracy. By constructing a multi-dimensional object tagging system (research direction, academic influence, historical collaboration behavior) and a quantitative matching model, the selection of commissioned manuscripts, which originally relied on subjective judgment, is transformed into a calculable and comparable recommendation score, avoiding the subjective bias of manual recommendations. Experimental results show that using the method described in this application reduces the matching error between commissioned manuscripts and thematic areas by more than 40%.

[0012] II. Improve manuscript solicitation efficiency. Automated screening and sorting of expert resources reduces the time required for traditional manual screening from days (or even weeks) to hours, supporting rapid retrieval and matching of large-scale expert databases. Editors only need to input the topic requirements text to obtain a high-quality recommendation list within minutes, significantly improving the response speed of topic solicitation.

[0013] III. Optimizing Manuscript Solicitation Effectiveness. Through quantitative evaluation of historical collaboration behavior tags, priority is given to recommending experts with high response rates, excellent manuscript quality, and good cooperation, resulting in a manuscript response rate increase of over 30% and an average citation count of accepted manuscripts increase of over 25%, significantly improving the quality of special topic manuscript solicitation and the journal's academic influence. Attached Figure Description

[0014] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart illustrating a method for recommending commissioned manuscripts based on quantitative matching according to an embodiment of this application is shown. Figure 2 A schematic diagram of the structure of a commissioned writing object recommendation system based on quantitative matching according to an embodiment of this application is shown; Figure 3 A schematic diagram of the structure of an electronic device according to an embodiment of this application is shown. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. 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.

[0016] Figure 1 A flowchart illustrating a quantitative matching-based manuscript recommendation method according to an embodiment of this application is shown below. Figure 1 As shown, this embodiment includes steps S110 to S140: Step S110, Object Tag Construction: Obtain multi-source data for each candidate object, and construct research direction tags, academic influence tags, and historical collaboration behavior tags for each candidate object based on the multi-source data.

[0017] For each candidate (i.e., a potential commissioning expert), the system collects the following multi-source data from public academic databases, research project management institutions, public patent databases, and historical collaboration records within journals.

[0018] Literature data: including the title, abstract, keywords, publication year, journal source, citation count, etc. of all publications of the candidate.

[0019] Patent data: including the name, abstract, classification number, publication year, etc. of all patents filed by the candidate.

[0020] Project data: including the name, level, and duration of the research projects led or participated in by the candidate.

[0021] Honors data: including the academic honors, award names and levels received by the candidates.

[0022] Historical collaboration data: including the candidate journal's historical manuscript solicitation records, response status, review cycle, and cooperation in revisions.

[0023] After data collection is completed, the system processes each type of data to generate labels in three dimensions.

[0024] (1) Research direction label.

[0025] For each candidate, the system extracts the title, abstract, and keywords of each of its documents, applies a trained topic assignment model to infer the probability distribution of the document in each research category.

[0026] Similarly, for each candidate, the system extracts the name, abstract, and classification number of each patent, and applies the same topic allocation model to infer the probability distribution of the patent in each research direction category.

[0027] Then, the probability values ​​of all documents and patents in the same research direction category are summed to obtain the cumulative intensity of the candidate object in each research direction category. A vector is constructed using the cumulative intensity values ​​across all categories, and this vector is the research direction label of the candidate object.

[0028] For example: There are three research directions, denoted as T1, T2, and T3; a candidate X has 3 publications, denoted as L1, L2, and L3, and 1 patent, denoted as P1; Based on inference, the probability distribution of each document and patent in terms of research direction category is as follows: L1: T1=0.6, T2=0.3, T3=0.1; L2: T1=0.9, T2=0.05, T3=0.05; L3: T1=0.2, T2=0.7, T3=0.1; P1: T1=0.1, T2=0.1, T3=0.8; Therefore, the cumulative intensity of T1 is 0.6 + 0.9 + 0.2 + 0.1 = 1.8; The cumulative intensity of T2 is 0.3 + 0.05 + 0.7 + 0.1 = 1.15; The cumulative intensity of T3 is 0.1 + 0.05 + 0.1 + 0.8 = 1.05; The research direction label vector of candidate object X is [1.8, 1.15, 1.05].

[0029] It should be noted that research direction categories can be predefined based on the energy discipline classification system (such as fossil energy, new energy, energy storage technology, power systems, etc.). For each energy discipline classification, it can be further subdivided into dozens or even hundreds of parallel research direction categories. For example, under the discipline classification of "power systems", it can be further subdivided into specific research direction categories such as "transient stability", "voltage stability", "frequency stability", "new energy grid connection", "electricity market", and "relay protection".

[0030] The topic allocation model in this application is applicable to the classification of research directions under any of the above-mentioned energy disciplines. However, in practice, it is usually carried out using several research direction categories under one energy discipline.

[0031] Furthermore, the topic assignment model used in this application employs a Labeled Latent Dirichlet Allocation (LDA) model, which has been trained and adapted for energy discipline classification. Labeled LDA is a supervised topic model that assumes each research direction category (label) corresponds to a fixed topic, and each topic is represented by a probability distribution over words. In this application, each research direction category predefined according to an energy discipline classification system is directly used as a label in the Labeled LDA, with each label uniquely corresponding to a topic. During training, documents labeled with research direction categories are used, and the model learns the word distribution under each label (research direction category). During inference, for a new document, the model outputs the probability distribution of the words in the document across each label.

[0032] In this application, a document may be: a text composed of the title, abstract, and keywords of a document; a text composed of the name, abstract, and classification number of a patent; or the core elements of a topic requirement text.

[0033] From publicly available academic databases, we collect the titles, abstracts, and keywords of documents under an energy discipline category that have been categorized by research direction. From publicly available patent databases, we collect the names, abstracts, and classification numbers of patents under the same energy discipline category that have been categorized by research direction. We also collect the core elements of manuscript requests from target journals over the past 10 years targeting the same energy discipline category and its categorized research direction. The categorization of research directions can be done manually by editors or semi-automatically (i.e., automatically categorized first, then manually verified).

[0034] During training, the Dirichlet prior parameter α for the document-tag distribution was set to 1, and the Dirichlet prior parameter β for the tag-word distribution was set to 0.01. Gibbs sampling was used for parameter estimation, with 2000 iterations. After training, the tag-word distribution matrix was obtained.

[0035] For a new document, with the label-word distribution matrix fixed, the probability distribution of the document in K research direction categories is obtained after 100 iterations of folded Gibbs sampling.

[0036] Although documents, patents, and commissioned manuscripts differ in subject matter and style, they all describe the same research direction category. Labeled LDA, as a supervised topic model, has the core advantage of mapping documents with different expressions to the same predefined label space, thus achieving semantic alignment. Therefore, despite the differences in vocabulary among the three types of documents, the "label-word" distribution matrix learned by Labeled LDA can cluster synonyms or near-synonyms from different documents under the same research direction label. Therefore, the same Labeled LDA model can be used to infer the relationship between the three types of documents.

[0037] When emerging research directions appear, they need to be added as new research direction categories to the labels (e.g., increasing K by 1). However, there are almost no relevant documents available for training Labeled LDA. In this case, a zero-shot learning method is adopted. Specifically, by utilizing the semantic relationship between emerging research directions and research direction categories in the knowledge graph, the semantic vector of the emerging research direction is matched with existing labels, thereby assigning corresponding labels to experts who have made preliminary achievements in that emerging research direction.

[0038] (2) Academic influence label.

[0039] For each candidate, the system calculates the following metrics based on its literature data.

[0040] H-index: The H-index is a comprehensive evaluation indicator that measures the quantity and impact of an expert's academic output. It is defined as follows: if an expert's H-index is h, it means that among the expert's published literature, h of them have been cited at least h times, and the remaining literature has been cited no more than h times.

[0041] Number of publications in core journals: This counts the number of articles published by the candidate in core journals.

[0042] Total citations: The total number of citations received by all documents.

[0043] For each candidate, the system calculates the project level based on its project data: for example, national-level projects are assigned a value of 1.0, provincial and ministerial-level projects are assigned a value of 0.7, and other projects are assigned a value of 0.4. The project level of a candidate is obtained by summing up all the assigned values.

[0044] For each candidate, the system calculates the honor level based on its honor data: for example, national-level awards (first prize is assigned 1.0, second prize is assigned 0.8), provincial and ministerial-level awards (first prize is assigned 0.6, second prize is assigned 0.4), and other honors are assigned 0.2. All the assigned values ​​of a candidate are added together to obtain its honor level.

[0045] All the above elements are normalized separately (for example, by using Min-Max normalization to map the original values ​​to the [0,1] interval), and then combined into a multi-dimensional vector, which is the academic influence label.

[0046] For example: Candidate X's academic influence labels are [H-index normalized value 0.9, core journal publications normalized value 0.8, total citations normalized value 0.95, project level 0.8, honor level 0.6].

[0047] (3) Historical cooperative behavior tags.

[0048] For each candidate, the system determines the following indicators based on its historical cooperation data.

[0049] Manuscript response rate: The proportion of manuscripts that have been commissioned in the past that have been responded to (agreed to write). The formula can be the number of responses / the total number of manuscripts commissioned.

[0050] Manuscript acceptance rate: The percentage of manuscripts accepted in the past. The formula can be the number of accepted manuscripts / the number of responses.

[0051] Average peer review period: The average number of days from commissioning a manuscript to submitting it. This indicator needs to be reverse normalized (the shorter the period, the higher the score).

[0052] Compliance rating: The editor gives a score of 1-5 based on the degree of cooperation with the experts' revision suggestions.

[0053] The manuscript response rate, manuscript acceptance rate, and revision cooperation score were normalized respectively, and the average review cycle was reverse-normalized. Then, a four-dimensional vector was formed, which is the historical cooperation behavior label.

[0054] For example, the historical collaboration behavior labels for candidate X are [commission response rate 0.8, manuscript acceptance rate 0.9, average review cycle 0.7, revision cooperation score 0.9].

[0055] Through the above steps, each candidate object is represented as a three-dimensional label structure: research direction label (multi-dimensional vector), academic influence label (five-dimensional vector), and historical cooperation behavior label (four-dimensional vector).

[0056] Step S120, Demand Feature Construction: Obtain the manuscript request text and extract manuscript direction tags based on the manuscript request text.

[0057] The system receives the manuscript request text input by the editor. The manuscript request text usually includes core elements such as core direction, technical methods, and application scenarios.

[0058] Core direction: This indicates the most core research topic of the manuscript, such as "stability of new power systems".

[0059] Technical methods: This indicates the key technologies or methods involved in the manuscript, such as "transient stability analysis" or "voltage stability control".

[0060] Application scenario: Indicates the application field of the commissioned article, such as "new energy grid connection" or "power grid dispatch".

[0061] For each core element, the system uses the same topic assignment model as in object labeling to infer the probability distribution of that core element across various research direction categories.

[0062] It should be noted that, since the topic assignment model is the same, the research direction category used when inferring core elements (i.e., the various labels of Labeled LDA) is completely consistent with the research direction category in the research direction label, and the two are aligned in the same semantic space.

[0063] For example: There are three research areas, still referred to as T1, T2, and T3; Based on inference, the probability distribution of each core element in terms of research direction category is as follows: Core directions: T1=0.7, T2=0.2, T3=0.1; Technical method: T1=0.6, T2=0.3, T3=0.1; Application scenario: T1=0.3, T2=0.5, T3=0.2.

[0064] The system pre-sets basic weights for each core element. For example, the core direction has a weight of 0.4, the technical method has a weight of 0.3, and the application scenario has a weight of 0.3. These weights can be customized by editors according to actual needs (within ±0.2). For each research direction category, the probability values ​​of each core element in that category are weighted and summed according to the pre-set weights to obtain the comprehensive probability for that category. The comprehensive probabilities of all research direction categories constitute a vector, i.e., the manuscript direction label.

[0065] For example: The overall probability of T1 is: 0.7×0.4+0.6×0.3+0.3×0.3=0.55; The overall probability of T2 is: 0.2×0.4+0.3×0.3+0.5×0.3=0.32; The overall probability of T3 is: 0.1×0.4+0.1×0.3+0.2×0.3=0.13; The vector of call for papers is [0.55, 0.32, 0.13].

[0066] Step S130, multi-dimensional matching: Determine the research direction fit based on research direction tags and commissioned manuscript direction tags, determine the academic authority based on academic influence tags, determine the historical cooperation degree based on historical cooperation behavior tags, and calculate the comprehensive recommendation score based on the research direction fit, academic authority, and historical cooperation degree.

[0067] For each candidate, the system performs the following four steps: (1) Determine the suitability of the research direction.

[0068] For each candidate, the system first calculates the cosine similarity between the candidate's research direction label and the commissioned manuscript direction label.

[0069] The formula for calculating cosine similarity is: , formula (1); in, Indicates the index of the candidate object. A research direction label (vector) representing a candidate object. Vector labels indicating the direction of commissioned articles This represents the modulus. The closer the cosine similarity value is to 1, the more similar the two samples are.

[0070] For example, the cosine similarity between the research direction label of candidate X and the manuscript direction label is approximately 0.967.

[0071] The system then calculates the degree of thematic overlap between each candidate and the commissioned text.

[0072] The research direction category with the highest probability value for the solicitation direction label is selected as the solicitation topic category. For example, continuing with the above example, the solicitation topic category is determined to be T1.

[0073] The statistics include all literature and patents of the candidate within the most recent preset time period (e.g., the last 5 years). For example, the three literatures of a candidate X are denoted as L1, L2, and L3, all within the last year, and the one patent is denoted as P1, which is within the last 4 years.

[0074] For each of these documents and patents, extract its probability value in the commissioned topic category. For example, the probability value of L1 in T1 is 0.6, the probability value of L2 in T1 is 0.9, the probability value of L3 in T1 is 0.2, and the probability value of P1 in T1 is 0.1.

[0075] A time decay factor is introduced: the time decay factor for the past 3 years is 1.0, the time decay factor for 3-5 years is 0.7, and the time decay factor for more than 5 years is 0.3. For example, the time decay factor for L1, L2, and L3 is 1.0, and the time decay factor for P1 is 0.7.

[0076] Multiply the probability value of each article by the corresponding time decay factor and sum them to obtain the first result. For example, the first result = 0.6×1 + 0.9×1 + 0.2×1 + 0.1×0.7 = 1.77.

[0077] The second result is obtained by summing the time decay factors of these documents and patents. For example, the second result = 1.0 + 1.0 + 1.0 + 0.7 = 3.7.

[0078] Divide the first result by the second result to obtain the topic overlap rate. For example, topic overlap rate = 1.77 ÷ 3.7 ≈ 0.478. The topic overlap rate reflects the depth of recent research conducted by the candidate on the commissioned topic.

[0079] For each candidate, the system finally sums the cosine similarity and topic overlap according to preset weights (e.g., 0.5 each) to obtain the research direction suitability.

[0080] For example: The research direction fit of candidate X is 0.967 × 0.5 + 0.478 × 0.5 = 0.7225. Converted to percentages, this is 72.25 points.

[0081] (2) Determine academic authority.

[0082] For each candidate, the system calculates the weighted sum of the elements of the academic influence label to obtain the academic authority level.

[0083] Specifically, the weights can be preset as follows: H-index 0.3, number of publications in core journals 0.25, total citations of research results 0.2, project level 0.1, and honor level 0.15. These weights can be customized by editors according to actual needs (adjustment range ±0.2).

[0084] For example, if candidate X's academic influence label is [0.9, 0.8, 0.95, 0.8, 0.6], then the academic authority score is 0.9 × 0.3 + 0.8 × 0.25 + 0.95 × 0.2 + 0.8 × 0.1 + 0.6 × 0.15 = 0.83. This is equivalent to 83 points on a percentage scale.

[0085] (3) Determine the degree of historical cooperation.

[0086] For each candidate, the system calculates the weighted sum of each element of the historical cooperation behavior label to obtain the historical cooperation degree.

[0087] Specifically, the weights can be preset as follows: manuscript response rate 0.4, manuscript acceptance rate 0.3, average review period 0.15, and revision cooperation score 0.15. These weights can be customized by editors according to actual needs (adjustment range ±0.2).

[0088] For example, if candidate X's historical cooperative behavior labels are [0.8, 0.9, 0.7, 0.9], then the historical cooperation degree = 0.8 × 0.4 + 0.9 × 0.3 + 0.7 × 0.15 + 0.9 × 0.15 = 0.83. This is equivalent to 83 points on a percentage scale.

[0089] (4) Calculation of comprehensive recommendation score.

[0090] The comprehensive recommendation score is obtained by weighting and summing the research direction suitability, academic authority, and historical cooperation.

[0091] Specifically, the basic weights can be configured as follows: research direction fit 0.5, academic authority 0.3, and historical collaboration 0.2. Each basic weight can be customized by the editor according to actual needs (adjustment range ±0.2). For example, for emerging research directions, the research direction fit weight can be increased to 0.6, the academic authority weight decreased to 0.25, and the historical collaboration weight decreased to 0.15.

[0092] For example: The overall recommendation score for candidate X is 0.7225 × 0.5 + 0.83 × 0.3 + 0.83 × 0.2 = 0.776. Converted to a percentage, this is 77.6 points.

[0093] Step S140: The candidate objects are initially ranked according to the comprehensive recommendation score. After conflict detection and removal of conflicting objects, a recommended ranking is generated.

[0094] The system sorts all candidates in descending order based on their comprehensive recommendation scores, generating an initial sorted list.

[0095] Conflict detection is performed based on the initial sorting list to exclude candidates who have already accepted commissions within a preset time window (e.g., the past 6 months), thus avoiding resource waste or a decline in manuscript quality caused by frequent commissions for the same candidate.

[0096] Additionally, overlapping candidates from the same research team can be excluded: for example, by analyzing the institutions to which the candidates belong, candidates belonging to the same research team can be identified. For multiple candidates belonging to the same research team, only the candidate with the highest overall recommendation score is retained, while the others are eliminated.

[0097] The retained candidates are arranged into a recommended ranking list. The system outputs a final recommendation report, which includes the recommended ranking list, as well as basic information about each recommended candidate and a detailed comprehensive recommendation score (including research direction suitability, academic authority, and historical collaboration).

[0098] Furthermore, the system supports manual intervention. Editors can manually adjust the recommendation ranking list, add custom objects to the ranking list, or exclude unsuitable objects from the recommendation results interface. The system will record these manual interventions and use them as feedback data for subsequent model optimization training and weight adjustment, achieving human-machine collaborative optimization.

[0099] Figure 2 A schematic diagram of a commissioned writing object recommendation system based on quantitative matching according to an embodiment of this application is shown. (Refer to...) Figure 2 As shown, the quantitative matching-based commissioned article recommendation system 200 includes: The object label construction unit 210 is used to obtain multi-source data of each candidate object and construct research direction labels, academic influence labels and historical cooperation behavior labels of each candidate object based on the multi-source data. The demand feature construction unit 220 is used to obtain the manuscript request text and extract the manuscript direction tags based on the manuscript request text. The multi-dimensional matching unit 230 is used to determine the research direction suitability based on research direction tags and manuscript solicitation direction tags, the academic authority based on academic influence tags, the historical cooperation level based on historical cooperation behavior tags, and to calculate the comprehensive recommendation score based on the research direction suitability, academic authority, and historical cooperation level. The recommendation result generation unit 240 is used to perform preliminary ranking of candidate objects based on the comprehensive recommendation score, perform conflict detection on the preliminary ranking and remove conflicting objects to generate a recommendation ranking.

[0100] In some optional implementations, in the above system, the object labeling unit 210 is used to: for each candidate object, acquire the candidate object's literature data, patent data, project data, honor data, and historical cooperation data; extract the title, abstract, and keywords of each document, apply a trained topic allocation model to infer the probability distribution of each document in each research direction category; extract the name, abstract, and classification number of each patent, apply a trained topic allocation model to infer the probability distribution of each patent in each research direction category; sum the probabilities of all documents and patents in each research direction category to obtain the candidate object. The cumulative intensity of an object across various research categories is used as the research direction label, with the cumulative intensity vector serving as the research direction label. Based on the candidate object's literature data, the H-index, number of publications in core journals, and total citations are determined. Based on the candidate object's project data, the project level is determined. Based on the candidate object's honor data, the honor level is determined, and these are normalized to form an academic influence label. Based on the candidate object's historical collaboration data, the manuscript response rate, manuscript acceptance rate, average review period, and revision cooperation score are determined. The manuscript response rate, manuscript acceptance rate, and revision cooperation score are normalized, and the average review period is inversely normalized to form a historical collaboration behavior label.

[0101] In some optional implementations, in the above system, the demand feature construction unit 220 is used to: obtain the manuscript request text and extract core elements; wherein, the core elements include: core direction, technical method and application scenario; apply the trained topic allocation model to make inferences respectively, and obtain the probability distribution of each core element in each research direction category; based on the preset weight of each core element, sum the probability values ​​of each core element under each research direction category according to the preset weight to obtain the manuscript request direction label.

[0102] In some optional implementations, in the above system, the multi-dimensional matching unit 230 is used to: calculate the cosine similarity between the research direction label and the solicitation direction label of each candidate object; take the research direction category with the highest probability value of the solicitation direction label as the solicitation topic category; count all the literature and patents of the candidate object in the most recent preset time period, sum the probability values ​​of the literature and patents in the solicitation topic category in the preset time period according to a preset time decay factor to obtain a first result, sum the time decay factors of all the literature and patents in the preset time period to obtain a second result, divide the first result by the second result to obtain the topic overlap; and sum the cosine similarity and the topic overlap to obtain the research direction fit of the candidate object.

[0103] In some optional implementations, in the above system, the multi-dimensional matching unit 230 is used to: sum the elements of the academic influence label of each candidate object with weights to obtain the academic authority; and sum the elements of the historical cooperation behavior label of each candidate object with weights to obtain the historical cooperation degree.

[0104] In some alternative implementations, in the above system, the multi-dimensional matching unit 230 is used to: weight and sum the research direction suitability, academic authority, and historical cooperation to obtain a comprehensive recommendation score.

[0105] In some optional implementations, in the above system, the recommendation result generation unit 240 is used to: sort each candidate object in descending order according to the comprehensive recommendation score to generate an initial sorting list; exclude candidate objects that have accepted commissions within a preset time window from the initial sorting list; and arrange the remaining candidate objects into a recommendation sorting list.

[0106] It should be noted that the aforementioned quantitative matching-based manuscript recommendation system 200 can implement the aforementioned quantitative matching-based manuscript recommendation method, which will not be elaborated further.

[0107] Figure 3 This invention illustrates a schematic diagram of the structure of an electronic device according to an embodiment of the present application. Figure 3 As shown, the electronic device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used for communication with external devices via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a quantitative matching-based manuscript recommendation method.

[0108] In one embodiment, the electronic device provided in this application includes a memory and a processor. The memory stores a database and a computer program that can run on the processor. When the processor executes the computer program, it implements the steps of the aforementioned method for recommending commissioned manuscripts based on quantitative matching.

[0109] The above is as stated in this application. Figure 2The method for implementing a quantitative matching-based commissioning object recommendation system disclosed in the illustrated embodiments can be applied to a processor or implemented by a processor. During implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by software instructions. The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The steps of the method disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0110] In one embodiment, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the aforementioned method for recommending commissioned manuscripts based on quantitative matching.

[0111] It should be noted that the functions or steps that the above-mentioned electronic devices or computer-readable storage media can achieve can be referred to the relevant descriptions in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0112] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), Synchlink, DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0113] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above.

[0114] It should be noted that any AI models, software tools, or components not belonging to this company appearing in the embodiments of this application are merely illustrative examples and do not represent actual use. All user personal information involved in the embodiments of this application has been authorized (with the knowledge and consent) by the relevant parties or has been fully authorized by all parties, and the executing entity may obtain it through various legal and compliant means. The collection, storage, use, processing, transmission, provision, and disclosure of the information, data, and signals involved all comply with relevant laws and regulations and do not violate public order and good morals.

[0115] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. 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 of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for recommending commissioned manuscripts based on quantitative matching, characterized in that, include: Object tagging: Obtain multi-source data for each candidate object, and construct research direction tags, academic influence tags, and historical collaboration behavior tags for each candidate object based on the multi-source data; Demand feature construction: Obtain the manuscript request text and extract manuscript direction tags based on the manuscript request text; Multi-dimensional matching: The research direction fit is determined based on the research direction tag and the manuscript solicitation direction tag; the academic authority is determined based on the academic influence tag; the historical cooperation behavior tag is determined based on the historical cooperation behavior tag; and the comprehensive recommendation score is determined by comprehensively calculating the research direction fit, academic authority and historical cooperation. Generate recommendation results: The candidate objects are initially ranked according to the comprehensive recommendation score. After conflict detection and removal of conflicting objects, the recommendation ranking is generated.

2. The method for recommending commissioned manuscripts based on quantitative matching according to claim 1, characterized in that, The object tag construction includes: For each candidate, obtain its literature data, patent data, project data, honors data, and historical cooperation data; The title, abstract, and keywords of each document are extracted, and a trained topic assignment model is applied to infer the probability distribution of each document in each research direction category. Extract the name, abstract and classification number of each patent, and apply the trained topic assignment model to infer the probability distribution of each patent in each research direction category; The probability of all documents and patents in each research direction category is summed to obtain the cumulative intensity of the candidate object in each research direction category, and the cumulative intensity vector is used as the research direction label. The H-index, number of publications in core journals, and total citations of the results are determined based on the candidate's literature data. The project level is determined based on the candidate's project data. The honor level is determined based on the candidate's honor data. After normalization, these are used to form an academic influence label. Based on the candidate's historical collaboration data, the manuscript response rate, manuscript acceptance rate, average review period, and revision cooperation score are determined. The manuscript response rate, manuscript acceptance rate, and revision cooperation score are normalized respectively, and the average review period is reverse normalized to form a historical collaboration behavior label.

3. The method for recommending commissioned manuscripts based on quantitative matching according to claim 1, characterized in that, The construction of the demand features includes: Obtain the article request text and extract its core elements; these core elements include: core direction, technical methods, and application scenarios. The trained topic assignment model is used to make inferences to obtain the probability distribution of each core element in each research direction category; Based on the preset weights of each core element, the probability values ​​of each core element under each research direction category are summed according to the preset weights to obtain the manuscript direction label.

4. The method for recommending commissioned manuscripts based on quantitative matching according to claim 1, characterized in that, The determination of research direction suitability based on research direction tags and commissioned manuscript direction tags includes: For each candidate, calculate the cosine similarity between the candidate's research direction label and the commissioned manuscript direction label; The research direction category with the highest probability value of the solicitation direction tag is used as the solicitation topic category; The first result is obtained by summing the probability values ​​of the literature and patents of the candidate within the most recent preset time period, weighted by a preset time decay factor, and the second result is obtained by summing the time decay factors of all literature and patents within the preset time period. The first result is divided by the second result to obtain the topic overlap. The research direction fit of the candidate object is obtained by weighting and summing the cosine similarity and topic overlap.

5. The method for recommending commissioned manuscripts based on quantitative matching according to claim 1, characterized in that, The determination of academic authority based on academic influence tags and the determination of historical cooperation based on historical cooperation behavior tags include: The academic authority is obtained by weighting and summing the elements of the academic influence labels of each candidate. The historical cooperation degree is obtained by weighting and summing the elements of the historical cooperation behavior labels of each candidate.

6. The method for recommending commissioned manuscripts based on quantitative matching according to claim 1, characterized in that, The comprehensive recommendation score is determined by comprehensively calculating the relevance of research direction, academic authority, and historical collaboration, including: The comprehensive recommendation score is obtained by weighting and summing the research direction suitability, academic authority, and historical cooperation.

7. The method for recommending commissioned manuscripts based on quantitative matching according to claim 1, characterized in that, The generated recommendation results include: Based on the comprehensive recommendation score, each candidate object is sorted in descending order to generate an initial sorted list; In the initial sorting list, candidates who have already accepted commissions within the preset time window are excluded; Arrange the retained candidate objects into a recommended sort list.

8. A commissioned article recommendation system based on quantitative matching, characterized in that, include: The object tagging unit is used to obtain multi-source data for each candidate object and construct research direction tags, academic influence tags, and historical cooperation behavior tags for each candidate object based on the multi-source data. The demand feature construction unit is used to obtain the manuscript request text and extract the manuscript direction tags based on the manuscript request text; The multi-dimensional matching unit is used to determine the research direction suitability based on research direction tags and manuscript solicitation tags, the academic authority based on academic influence tags, the historical cooperation level based on historical cooperation behavior tags, and to calculate the comprehensive recommendation score based on the research direction suitability, academic authority, and historical cooperation level. The recommendation result generation unit is used to initially rank candidate objects based on the comprehensive recommendation score, perform conflict detection on the initial ranking and remove conflicting objects to generate the recommendation ranking.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method for recommending commissioned manuscripts based on quantitative matching as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the quantitative matching-based manuscript recommendation method as described in any one of claims 1 to 7.

Images