Intelligent submission expert matching method and device and electronic equipment

Abstract

The invention discloses an intelligent submission expert matching method and device and electronic equipment, and relates to the field of data processing. The method comprises the following steps: obtaining a submission material submitted by a user; storing the submitted materials in a data queue, and asynchronously extracting the submitted materials from the data queue to obtain keywords; based on the keyword, using a TF-IDF algorithm to calculate the similarity between the submitted material and a preset expert database, and generating a first data pool, the first data pool comprising a plurality of first experts; based on the first data pool, analysis is carried out in combination with historical submission data, a second data pool is generated, and the second data pool comprises a plurality of second experts; performing intersection calculation on the first data pool and the second data pool to obtain a submission expert group, the submission expert group comprising a plurality of submission experts; generating a submission topology for the submission material based on the submission expert group; and submitting the submission material according to the submission topology. By implementing the technical scheme provided by the invention, the matching accuracy of the audit experts is improved.

Description

Technical Field

[0001] This application relates to the field of data processing, and particularly to an intelligent expert matching method, device and electronic device for submission for review. Background Art

[0002] With the rapid development of technology and the advent of the knowledge economy era, the demand for the review of scientific research achievements and various thesis documents has increased sharply. The review of these materials not only requires judging their innovation and effectiveness, but also ensuring a high degree of matching with experts in related fields for in-depth and accurate evaluation. This demand has promoted the development of intelligent submission for review systems, aiming to improve the efficiency and quality of the submission for review process through automated tools.

[0003] Currently, existing expert matching technologies mostly rely on simple keyword searches or manual selection of experts, and these methods have multiple defects. First, simple keyword matching cannot accurately capture the deep content and complexity of materials, often resulting in inaccurate matching. Second, manually selecting experts is time-consuming and laborious, and is easily affected by subjective biases, making it difficult to ensure diversity in selection, which in turn leads to the selection of experts with mismatched research directions, resulting in a decrease in the accuracy of matching. Therefore, related technologies have the problem of inaccurate matching of experts for submission for review.

[0004] Therefore, there is an urgent need for an intelligent expert matching method, device and electronic device for submission for review. Summary of the Invention

[0005] This application provides an intelligent expert matching method, device and electronic device for submission for review, which improves the accuracy of expert matching for submission for review.

[0006] In the first aspect of this application, an intelligent expert matching method for submission for review is provided. The method includes: obtaining the submission for review materials submitted by a user; storing the submission for review materials in a data queue, and asynchronously extracting the submission for review materials from the data queue to obtain keywords; based on the keywords, using the TF-IDF algorithm to calculate the similarity between the submission for review materials and a preset expert library, generating a first data pool, where the first data pool includes multiple first experts; based on the first data pool, analyzing in combination with historical submission for review data to generate a second data pool, where the second data pool includes multiple second experts; performing an intersection calculation on the first data pool and the second data pool to obtain a submission for review expert group, where the submission for review expert group includes multiple submission for review experts; based on the submission for review expert group, generating a submission for review topology for the submission for review materials; and performing submission for review on the submission for review materials according to the submission for review topology.

[0007] By adopting the above technical solution, by obtaining the submission materials submitted by the user, storing the submission materials in the data queue, and asynchronously extracting the submission materials from the data queue to obtain keywords, the efficient reception and processing of the submission materials can be realized, and the blocking and delay caused by excessive submission materials can be avoided. By using the keywords and the TF-IDF algorithm to calculate the similarity between the submission materials and the preset expert database, a first data pool is generated, and potential reviewers related to the research direction of the submission materials can be quickly matched, improving the pertinence and accuracy of the matching. On this basis, by combining the analysis of historical submission data to generate a second data pool, high-quality reviewers with rich experience and familiarity with the research direction of the materials can be selected. By calculating the intersection of the two data pools, the relevance of the expert's research background and the historical review performance can be taken into account at the same time, and the final review expert group can be selected to ensure the review quality. According to the review expert group, a submission topology can be automatically generated to optimize the review process, reasonably arrange the review order and cooperation mode of the experts, and improve the review efficiency. Finally, by using the submission topology to realize the automatic transfer and review of the materials among the experts, and intelligently controlling the process direction according to the review results, the manual intervention can be reduced, and the automation and intelligence of the entire submission process can be realized. By calculating the intersection of the first data pool and the second data pool, the final review expert group obtained will be those experts who are not only highly relevant to the submission materials but also have excellent performance in the historical submission process. This double screening mechanism significantly improves the accuracy and reliability of expert matching.

[0008] Optionally, calculating the similarity between the submission materials and the preset expert database using the TF-IDF algorithm based on the keywords to generate a first data pool specifically includes: performing word segmentation on the keywords to obtain a word segmentation result; calculating the TF value and IDF value of the keywords based on the word segmentation result; multiplying the TF value by the IDF value to obtain the TF-IDF value of the keywords; calculating the similarity between the submission materials and each expert in the preset expert database based on the TF-IDF value to obtain a similarity result; and taking the experts with a similarity greater than or equal to a preset similarity threshold as the first experts and adding the first experts to the first data pool.

[0009] By adopting the above technical solution, through word segmentation processing of the keywords in the submitted materials, the research topics and core contents of the submitted materials can be expressed as terms recognizable by a computer, providing a data basis for subsequent similarity calculation. By calculating the TF value and IDF value of the keywords using the word segmentation results, the importance of each keyword in the materials and the distinctiveness in the research directions of experts can be quantitatively characterized. Multiplying the TF value by the IDF value to obtain the TF-IDF value can comprehensively consider these two important attributes of the keywords and more accurately represent the association strength between the materials and experts. By calculating the similarity between the submitted materials and the research directions of each expert in the preset expert database based on the TF-IDF value, experts with the most relevant research backgrounds can be quickly found. Setting a preset similarity threshold to screen experts and adding the selected first experts to the first data pool can control the scale of the data pool on the basis of ensuring relevance, avoiding introducing too many irrelevant experts and affecting the subsequent analysis efficiency and matching accuracy.

[0010] Optionally, based on the first data pool, analyze in combination with historical submitted data to generate a second data pool, where the second data pool includes multiple second experts, specifically including: obtaining historical submitted materials with the same research direction as the submitted materials; counting the historical submitted experts corresponding to the historical submitted materials to obtain the submission times corresponding to each historical submitted expert; using the historical submitted experts with submission times greater than or equal to the preset number threshold as the second experts, and adding the second experts to the second data pool.

[0011] By adopting the above technical solution, by obtaining historical submitted materials with the same research direction as the submitted materials, the most relevant reference samples can be quickly locked in the vast historical data, providing an empirical basis for expert recommendation. Counting the reviewing experts corresponding to the historical submitted materials and their submission times, quantitatively analyzing the reviewing experience and workload of each expert in the relevant research direction, and excavating high-quality expert resources with rich experience. Using the submission times to set a preset number threshold to screen experts and adding the selected second experts to the second data pool can control the scale of the data pool on the basis of ensuring the reviewing experience of experts, improving the efficiency and accuracy of recommendation.

[0012] Optionally, after calculating the intersection of the first data pool and the second data pool to obtain the submitted expert group, the method further includes: obtaining the professional fields, reviewing times, and reviewing quality of each submitted expert in the submitted expert group; based on the professional fields, the reviewing times, and the reviewing quality, using a weighted algorithm to calculate the comprehensive scores of each submitted expert, and sorting the submitted expert group according to the comprehensive scores to obtain the target submitted expert group.

[0013] By adopting the above technical solutions, by obtaining multi-dimensional information such as the professional fields, review times, and review qualities of each reviewer in the submission review expert group, the academic backgrounds, work efficiencies, and work attitudes of the reviewers can be comprehensively evaluated, providing a basis for the comprehensive ranking of the experts. Using a weighted algorithm to calculate the comprehensive scores of the reviewers, the weights can be reasonably allocated according to the importance of different indicators, and a quantitative evaluation index that takes into account various factors can be obtained. Ranking the reviewers according to the comprehensive scores can quickly screen out the best expert combinations with excellent performances in all indicators, further improving the overall quality and matching degree of the entire review expert group.

[0014] Optionally, based on the submission review expert group, generating a submission review topology for the submission materials specifically includes: dividing the target submission review expert group into multiple expert subsets, each of the expert subsets including at least one of the reviewers; based on the order of the expert subsets in the target submission review expert group, generating a directed five-ring graph as the submission review topology for the submission materials, where the nodes in the submission review topology represent the expert subsets, and the edges in the submission review topology represent the review transfer relationships.

[0015] By adopting the above technical solutions, by dividing the target submission review expert group into multiple expert subsets, a division and cooperation mechanism can be introduced into the submission review topology. Reasonably setting the number of experts in each subset can improve the parallelism of the review while ensuring the review intensity of each link. Connecting them in sequence based on the order of the expert subsets to generate a directed five-ring graph as the submission review topology can achieve the orderly transfer and iterative review of the submission materials among the reviewers, constructing a collaborative and efficient review network with a rigorous process. Using the expert subsets as the nodes of the topology can achieve preliminary parallel review within the nodes, improving the review efficiency of a single link; then, the serial transfer is carried out according to the direction of the topology edges between the nodes, which can achieve step-by-step control and continuous optimization of the review process. Mapping the division and cooperation relationship of the target submission review expert group to the topological structure can achieve a visual expression of organizational management and process control, thus providing a structured process template for intelligent submission review and improving the comprehensibility and controllability of the entire review process.

[0016] Optionally, before submitting the submission materials for review according to the submission review topology, the method further includes: determining the review duration of each node in the submission review topology; calculating the estimated total review duration of the submission materials according to the review duration of each node; and sending the estimated total review duration to the user, so that the user can determine whether to adjust the submission review expert group of the submission materials according to the estimated total review duration.

[0017] By adopting the above technical solution, by determining the review duration of each node in the submission topology, the time required for each expert subset to complete the review work can be estimated, and the time progress and constraint conditions of the review process can be grasped. On this basis, by summarizing the review durations of each node, the estimated total review duration of the entire submission materials can be accurately calculated, providing a time budget and progress reference for the review work. Feeding back the estimated total review duration to the user can help the user understand the expected cycle of material review and reasonably arrange subsequent work plans. The user can dynamically adjust the expert group according to their own time requirements and preferences to achieve optimized human-computer interaction.

[0018] Optionally, the submission of the submission materials according to the submission topology specifically includes: determining a target submission expert according to the submission topology, where the target submission expert is the first submission expert in the submission topology; sending the submission materials to the target submission expert for review and receiving the review result of the target submission expert; if it is determined that the review result is passed, sending the submission materials to the next submission expert for review according to the submission topology; if it is determined that the review result is not passed, feeding back the review result to the user and ending the submission process.

[0019] By adopting the above technical solution, determining the first submission expert as the target submission expert according to the submission topology can clarify the starting point of the transfer of the submission materials among the submission experts and provide a starting point for the submission materials to enter the formal review stage. Sending the submission materials to the target submission expert and receiving the review result can trigger the review process of the submission materials automatically and achieve human-computer interaction, improving the timeliness and accuracy of the review work. Based on the review result of the submission expert, intelligently judging and controlling the transfer direction of the materials can realize the adaptive adjustment and dynamic optimization of the review process. If the review is passed, the submission materials will be automatically sent to the next submission expert for continuous review to ensure the orderly progress and positive transfer of the review work; if the review is not passed, the result will be promptly fed back to the user and the subsequent links will be automatically terminated to avoid unnecessary time waste.

[0020] In a second aspect of the present application, an intelligent review expert matching device is provided. The device includes an acquisition module and a processing module, where: the acquisition module is used to acquire the review materials submitted by the user; the processing module is used to store the review materials in a data queue, and asynchronously extract the review materials from the data queue to obtain keywords; the processing module is further used to calculate the similarity between the review materials and a preset expert database based on the keywords using the TF-IDF algorithm to generate a first data pool, and the first data pool includes multiple first experts; the processing module is further used to analyze and generate a second data pool based on the first data pool in combination with historical review data, and the second data pool includes multiple second experts; the processing module is further used to perform an intersection calculation on the first data pool and the second data pool to obtain a review expert group, and the review expert group includes multiple review experts; the processing module is further used to generate a review topology for the review materials based on the review expert group; the processing module is further used to submit the review materials for review according to the review topology.

[0021] In a third aspect of the present application, an electronic device is provided, which includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are both used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the method described in any one of the above.

[0022] In a fourth aspect of the present application, a computer-readable storage medium is provided. The computer-readable storage medium stores instructions, and when the instructions are executed, the method described in any one of the above is executed.

[0023] In summary, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. By obtaining the submission materials submitted by users, storing the submission materials in a data queue, and asynchronously extracting the submission materials from the data queue to obtain keywords, the efficient reception and processing of submission materials can be achieved, avoiding blockage and delay caused by excessive submission materials. By using the keywords and the TF-IDF algorithm to calculate the similarity between the submission materials and the preset expert database, a first data pool can be generated, which can quickly match potential reviewers related to the research direction of the submission materials, improving the pertinence and accuracy of the matching. On this basis, by combining the analysis of historical submission data to generate a second data pool, high-quality reviewers with rich experience and familiarity with the research direction of the materials can be selected. By calculating the intersection of the two data pools, both the relevance of the experts' research backgrounds and their historical review performances can be taken into account, and the final review expert group can be selected to ensure the review quality. According to the review expert group, a submission topology can be automatically generated, which can optimize the review process, reasonably arrange the review order and collaboration methods of the experts, and improve the review efficiency. Finally, by using the submission topology to realize the automatic transfer and review of materials among experts, and intelligently controlling the process direction according to the review results, manual intervention can be reduced, and the automation and intelligence of the entire submission process can be achieved.

[0024] 2. By performing word segmentation on the keywords of the submission materials, the research themes and core contents of the submission materials can be represented as computer-recognizable terms, providing a data basis for subsequent similarity calculations. By using the word segmentation results to calculate the TF values and IDF values of the keywords, the importance of each keyword in the materials and its discrimination in the research directions of the experts can be quantitatively characterized. By multiplying the TF value by the IDF value to obtain the TF-IDF value, these two important attributes of the keywords can be comprehensively considered, and the correlation strength between the materials and the experts can be more accurately characterized. By calculating the similarity between the submission materials and the research directions of each expert in the preset expert database based on the TF-IDF value, the experts with the most relevant research backgrounds can be quickly found. By setting a preset similarity threshold to screen the experts and adding the selected first experts to the first data pool, the scale of the data pool can be controlled on the basis of ensuring relevance, avoiding introducing too many irrelevant experts and affecting the subsequent analysis efficiency and matching accuracy.

[0025] 3. By obtaining historical submission materials with the same research direction as the submission materials, the most relevant reference samples can be quickly locked in the massive historical data, providing an empirical basis for expert recommendation. By counting the reviewers corresponding to the historical submission materials and their submission times, the review experience and workload of each expert in the relevant research direction can be quantitatively analyzed, and high-quality expert resources with rich experience can be mined. By using the submission times to set a preset times threshold to screen the experts and adding the selected second experts to the second data pool, the scale of the data pool can be controlled on the basis of ensuring the review experience of the experts, improving the efficiency and accuracy of the recommendation. Description of the Drawings

[0026] Figure 1It is a schematic flowchart of an intelligent submission expert matching method disclosed in an embodiment of the present application; Figure 2 It is a schematic module diagram of an intelligent submission expert matching device disclosed in an embodiment of the present application; Figure 3 It is a schematic structural diagram of an electronic device disclosed in an embodiment of the present application.

[0027] Explanation of reference numerals: 201, acquisition module; 202, processing module; 300, electronic device; 301, processor; 302, communication bus; 303, user interface; 304, network interface; 305, memory. Detailed implementation manners

[0028] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments.

[0029] In the description of the embodiments of the present application, words such as "for example" or "for illustration" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as "for example" or "for illustration" in the embodiments of the present application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, using words such as "for example" or "for illustration" aims to present relevant concepts in a specific manner.

[0030] In the description of the embodiments of the present application, the meaning of the term "a plurality" refers to two or more. For example, a plurality of systems refers to two or more systems, and a plurality of screen terminals refers to two or more screen terminals. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the technical features indicated. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. The terms "include", "comprise", "have" and their variants all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0031] The present application provides an intelligent submission expert matching method, referring to Figure 1 , Figure 1 It is a schematic flowchart of an intelligent submission expert matching method provided in an embodiment of the present application. This method is applied to a server, and the server is a server that executes an intelligent submission expert matching program. The server can be a single server, or a server cluster composed of multiple servers, or a cloud computing service center. This method includes steps S101 to S107, and the above steps are as follows: Step S101: Obtain the submission materials submitted by the user.

[0032] In step S101, the server obtains the submission materials submitted by the user through interaction with the user terminal device. This process can be achieved in various ways. First, the server provides a user-friendly interface for the user to upload and submit the submission materials. This interface can be a web application that the user can access through a browser, or it can be a mobile application that the user can use on a smartphone or tablet. On the user interface, the server provides a file upload control that allows the user to select a local file and upload it to the server. For the submission materials, the file formats include PDF, Word documents, and LaTeX source files. After the user fills in all the necessary information and submits the materials, the server processes and stores the uploaded submission materials.

[0033] For example, assume that a user is preparing an academic paper and hopes to find suitable reviewers. He logs in to the system's web interface and clicks the "Submit new material" button. The server pops up a form asking him to fill in basic information such as the title, abstract, and keywords of the paper, and upload the full PDF of the paper. After the user fills in all the information, he clicks the "Submit" button. The server stores the PDF file in the file system in the background, with the file name "paper- <uuid>.pdf" (where <uuid>The user's submission materials are submitted and the next step is to analyze and process the submission materials.

[0034] Step S102: storing the submitted materials in a data queue, and asynchronously extracting the submitted materials from the data queue to obtain keywords.

[0035] In step S102, the server stores the submitted materials in a data queue, extracts the submitted materials from the queue asynchronously, and processes them to extract keywords. The server selects a suitable message queue middleware, such as Apache Kafka and RabbitMQ. These middlewares provide reliable message storage and distribution functions, and can support high-concurrency producer-consumer models. The server deploys these middlewares at the infrastructure layer. When the user submits the submitted materials, the server does not process the submitted materials directly, but encapsulates them into a message and sends them to the specified message queue. This message contains metadata of the submitted materials (such as title, author, submission time, etc.) and a pointer to the actual material file (such as file path or URL). By sending the materials to the queue, the server can quickly respond to the user's submission request without having to wait for the actual processing of the submitted materials to be completed.

[0036] At the other end of the queue, the server deploys one or more consumer programs, which subscribe to messages in the queue and process the messages asynchronously. When a new message arrives in the queue, the consumer program automatically pulls the message and starts processing. In step S102, the main task of the consumer program is to obtain the submitted materials from the queue and extract keywords from them.

[0037] To extract keywords, the consumer program converts the original file of the submitted material (such as PDF) into a plain text format. This usually requires the use of some third-party libraries or tools, such as Apache PDFBox, iText, etc. The converted text can be further pre-processed, such as removing stop words, stemming, etc., to improve the quality of keyword extraction. Keyword extraction can use a variety of natural language processing techniques, such as TF-IDF, TextRank, Word2Vec, etc.

[0038] For example, when a PDF file of an academic paper is sent to the queue, an idle consumer program will pull the message and start processing. It calls the Apache PDFBox library to convert the PDF into plain text, then tokenizes the text and removes stop words. Next, it uses the TF-IDF model to extract keywords and obtains a list of 10 keywords. Finally, the consumer saves these keywords together with the metadata of the original paper to a downstream storage (such as a distributed cache or a database) for use in subsequent expert matching steps.

[0039] In summary, the use of a message queue and asynchronous consumers in step S102 can significantly improve the throughput and efficiency of the system in processing submission materials. At the same time, keyword extraction, as a crucial preprocessing step, lays an important foundation for subsequent intelligent expert matching.

[0040] Step S103: Based on the keywords, use the TF-IDF algorithm to calculate the similarity between the submission materials and a preset expert library, and generate a first data pool. The first data pool includes multiple first experts.

[0041] In step S103, based on the keywords, using the TF-IDF algorithm to calculate the similarity between the submission materials and a preset expert library, and generating a first data pool specifically includes: performing a word segmentation process on the keywords to obtain a word segmentation result; based on the word segmentation result, calculating the TF value and IDF value of the keywords; multiplying the TF value by the IDF value to obtain the TF-IDF value of the keywords; based on the TF-IDF value, calculating the similarity between the submission materials and each expert in the preset expert library to obtain a similarity result; according to the similarity result, taking the experts with a similarity greater than or equal to a preset similarity threshold as the first experts, and adding the first experts to the first data pool.

[0042] Specifically, the server uses the keywords extracted from the submission materials and calculates the similarity between the materials and each expert in the preset expert library through the TF-IDF algorithm, thereby generating a first data pool, which contains the first experts with a relatively high similarity to the submission materials. First, the server performs word segmentation on the keywords extracted from the submission materials. This step can use the same word segmentation tools and methods as those for keyword extraction, such as NLTK, Stanford Word Segmenter, etc. The purpose of word segmentation is to convert keywords into finer-grained lexical units for subsequent TF-IDF calculations. For example, a keyword "natural language processing" may be decomposed into three words: "natural", "language", and "processing". Next, the server calculates the TF value and IDF value of each keyword after word segmentation. TF (Term Frequency) represents the frequency of a word in the current material, which reflects the importance of this word to the current material. And IDF (Inverse Document Frequency) represents the reciprocal of the frequency of a word in the preset expert library, which reflects the importance of this word for differentiating different experts. The calculation formulas for TF and IDF are as follows: TF(t, d) = (the number of times the word t appears in the document d) / (the total number of words in the document d) IDF(t, D) = log((the total number of experts in the expert library D) / (the number of experts containing the word t + 1)) where t represents a keyword, d represents the current submission material, and D represents the entire preset expert library. The server can use the pre-established inverted index to quickly calculate the TF and IDF values of each keyword.

[0043] After obtaining the TF and IDF values, the server multiplies them to obtain the TF-IDF value of each keyword. The higher the TF-IDF value, the higher the discrimination degree of this word for the current material and experts. The server can form a vector of the TF-IDF values of all keywords as the feature representation of the current material.

[0044] Finally, the server uses the cosine similarity method to calculate the similarity between the feature vector of the submission material and the feature vectors of each expert in the expert library. The server sorts all the calculated similarity results and selects the experts with a similarity greater than the preset similarity threshold (such as 0.6) as the first experts to be added to the first data pool. These experts are the most relevant to the research fields and topics of the submission materials and are potential reviewers.

[0045] For example, assume a paper on "The Application of Deep Learning in Image Recognition" is submitted for review, and the extracted keywords include "deep learning", "convolutional neural network", "image recognition", etc. The server tokenizes these keywords and calculates their TF-IDF values to obtain the feature vector of the paper. Then, it calculates the similarity between this vector and the feature vectors of each expert in the expert database. Finally, the server selects 10 first experts with a similarity exceeding 0.6, and these first experts form the first data pool, providing high-quality candidates for the subsequent matching step.

[0046] Step S104: Based on the first data pool, analyze in combination with historical submission data to generate a second data pool, where the second data pool includes multiple second experts.

[0047] In step S104, based on the first data pool, analyze in combination with historical submission data to generate a second data pool, where the second data pool includes multiple second experts, specifically including: obtaining historical submission materials with the same research direction as the submission materials; counting the historical submission experts corresponding to the historical submission materials to obtain the submission times corresponding to each historical submission expert; taking the historical submission experts with submission times greater than or equal to the preset number threshold as the second experts, and adding the second experts to the second data pool.

[0048] Specifically, based on the first data pool, the server further analyzes in combination with historical submission materials to generate a second data pool. The server obtains historical submission materials with the same research direction as the current submission materials from the historical database. The server classifies and labels the research directions of the historical submission materials and the current materials. One method is to use a topic model, such as LDA (Latent Dirichlet Allocation). The server pre-trains an LDA model to map both the historical submission materials and the current submission materials to a set of topics. Then, the server calculates the topic distribution of the current submission materials and finds a batch of historical submission materials most similar to it as the reference set. For example, if the research direction of the current submission materials is identified as "computer vision", the server will search the historical database for all materials with the "computer vision" label and use them as the historical submission materials.

[0049] Next, the server counts the review experts for each historical submission material in the reference set and calculates the number of times each expert has conducted reviews. The server can use query languages such as SQL to obtain the number of times each expert has conducted reviews. For example, the server discovers that in the "computer vision" field, expert A has reviewed 10 materials, expert B has reviewed 8 materials, and expert C has reviewed 15 materials. Then, the server selects the experts whose review times are greater than or equal to a preset threshold (such as 5 times) as the second-level experts and adds them to the second data pool. Continuing with the above example, if the preset threshold is 5 times, then experts A, B, and C will all be selected into the second data pool and become the second-level experts.

[0050] Step S105: Calculate the intersection of the first data pool and the second data pool to obtain a review expert group, which includes multiple review experts.

[0051] In step S105, the server calculates the intersection of the first data pool and the second data pool to obtain the final review expert group. The review expert group includes multiple review experts. Specifically, the server compares the expert IDs in the first data pool and the second data pool to find the experts who appear in both data pools. For example, the first data pool contains experts A, B, C, D, and E, and the second data pool contains experts B, C, F, and G. The server compares the expert IDs in the two data pools and finds that experts B and C appear in both pools, so they are selected into the review expert group.

[0052] After step S105, the method further includes: obtaining the professional fields, review times, and review qualities of the various review experts in the review expert group; calculating the comprehensive scores of the various review experts using a weighted algorithm based on the professional fields, review times, and review qualities, and sorting the review expert group according to the comprehensive scores to obtain the target review expert group.

[0053] Specifically, the server obtains information such as the professional fields, review times, and review qualities of each expert in the review expert group. This information can be extracted from data sources such as the experts' personal profiles and review records. The professional field reflects the research direction and academic background of the expert, and the higher the degree of match with the submission material, the higher the score should be. The review time reflects the work efficiency and response speed of the expert, and the shorter the time, the higher the score should be. The review quality reflects the work attitude and review ability of the expert, and the higher the quality, the higher the score should be. The server can use technologies such as natural language processing and data mining to automatically extract and quantify this information.

[0054] For example, for Expert A, the server discovers that his research field is "Artificial Intelligence", with a matching degree of 90% to the material; the average review time is 10 days, and the review quality score is 4.5 (out of 5). Next, the server uses a weighted algorithm to calculate the comprehensive score of each expert. The core idea of the weighted algorithm is to assign different weights according to the importance of each dimension, and then add up the weighted scores of all dimensions to obtain the final comprehensive score.

[0055] Comprehensive score = Score of professional field × Field weight + Score of review time × Time weight + Score of review quality × Quality weight; among them, the scores of each dimension need to be normalized and converted into values between 0 and 1.

[0056] The score of the professional field reflects the matching degree between the research direction of the expert and the material to be reviewed. The calculation steps include: extracting the keyword of the expert's research field and constructing the expert field vector. From the information such as the expert's personal profile and research results, natural language processing technology can be used to automatically extract keywords and count the word frequencies to generate an n-dimensional vector. Extract the theme keywords of the material to be reviewed and construct the material theme vector. Using the same method, an n-dimensional vector can be generated from the information such as the title, abstract, and keywords of the material. Calculate the similarity between the expert field vector and the material theme vector as the score of the professional field. Common similarity measurement methods such as cosine similarity and Jaccard similarity can be used. Normalize the similarity to the interval [0, 1] as the final score of the professional field. The maximum-minimum normalization formula can be used: Normalized score = (Original score - Minimum score) / (Maximum score - Minimum score); for example, the research field vector of Expert A is (0.3, 0.5, 0.2), and the theme vector of the material M to be reviewed is (0.4, 0.4, 0.2), and the cosine similarity of the two vectors is 0.88. Assuming that among all experts, the minimum similarity is 0.4 and the maximum similarity is 0.95, then the score of Expert A's professional field is: (0.88 - 0.4) / (0.95 - 0.4) = 0.87 The review time score reflects the work efficiency and response speed of experts. The calculation steps include: counting the average review time of experts in the past year, with the unit being days. The start date and end date of each review can be automatically extracted from the records of the review system, and the average value of the time difference is calculated. Set a standard review time as the benchmark value. This time can be set according to the requirements of the journal or the conventions of the field, such as 30 days. Use the following formula to calculate the review time score: review time score = standard review time / actual average review time; limit the review time score within the range of [0, 1]. If the score is greater than 1, take 1; if the score is less than 0, take 0. For example, the average review time of expert B in the past year is 25 days. Assuming the standard review time is 30 days, then the review time score of B is: 30 / 25 = 1.2, which exceeds 1, so take 1.

[0057] The review quality score reflects the work attitude and review ability of experts. The calculation steps include: collecting all the review records of experts in the past year, including indicators such as the number of words in the review comments, the number of modification suggestions, and the average quality score. Normalize each indicator and convert it into a value within the range of [0, 1]. Z-score normalization or min-max normalization can be used. Perform a weighted average on the normalized indicators to obtain a comprehensive review quality score. The weights of each indicator can be set according to their importance. For example, expert C has reviewed 10 times in the past year. The average number of words in the comments each time is 800 words, the number of modification suggestions is 8, and the comprehensive quality score is 4.2 points (out of 5 points). After normalization and weighted average, the review quality score of C is 0.85.

[0058] The server performs the same calculations for all experts in the review expert group to be sent for review to obtain their comprehensive scores. Then, the server sorts the experts in descending order of the comprehensive scores to obtain an ordered target review expert group. The experts ranked higher indicate that they perform well in all dimensions.

[0059] Step S106: Generate a review topology for the materials to be sent for review based on the review expert group.

[0060] In step S106, generating a review topology for the materials to be sent for review based on the review expert group specifically includes: dividing the target review expert group into multiple expert subsets, with each expert subset including at least one review expert; generating a directed five-ring graph based on the order of the expert subsets in the target review expert group as the review topology of the materials to be sent for review. The nodes in the review topology represent the expert subsets, and the edges in the review topology represent the review transfer relationship.

[0061] Specifically, based on the target review expert group, the server generates a submission topology for the submission materials. The submission topology is a directed graph that represents the review transfer relationship and order among experts. The server can divide the expert group into multiple subsets and generate a reasonably structured and highly efficient review network according to the order of the subsets, thereby optimizing the entire review process.

[0062] First, the server divides the target review expert group into multiple expert subsets. An expert subset is a subset of the expert group, and each subset contains at least one expert. The division of subsets can be carried out according to factors such as the research fields of experts, review experience, and working regions. The goal is to make the characteristics of experts within each subset similar and the characteristics of experts between subsets complementary. Specific division methods can use clustering algorithms such as K-means and hierarchical clustering. The server can set parameters such as the number of subsets and the division granularity to control the scale and complexity of the submission topology.

[0063] For example, the server divides a target review expert group of 10 experts into 3 subsets: Subset 1 includes experts A, B, and C, mainly researching the field of artificial intelligence; Subset 2 includes experts D, E, and F, mainly researching the field of big data; Subset 3 includes experts G, H, I, and J, mainly researching the field of cloud computing.

[0064] Next, based on the order of the expert subsets, the server generates a directed five-ring graph as the submission topology of the submission materials. The directed five-ring graph is a special directed graph, consisting of 5 nodes forming a ring. Each node represents an expert subset, and each directed edge represents a review transfer relationship. Specifically, the server connects them in sequence into a ring according to the order of the subsets in the target review expert group. Each subset points to the next subset, and the last subset points to the first subset. In this way, a closed review transfer loop is formed, and the materials can circulate on the loop until all review experts have completed the review.

[0065] Continuing with the above example, the server generates a directed five-ring graph according to the order of Subset 1, Subset 2, and Subset 3. Subset 1 points to Subset 2, Subset 2 points to Subset 3, and Subset 3 points to Subset 1, forming a triangular loop. A sub-topology can also be generated inside each node to represent the review transfer relationship among experts within the subset.

[0066] Step S107: Conduct the submission of the submission materials according to the submission topology.

[0067] In a possible implementation, before step S107, determine the review duration of each node in the submission review topology; calculate the estimated total review duration of the submission materials based on the review duration of each node; send the estimated total review duration to the user so that the user can determine whether to adjust the review expert group of the submission materials according to the estimated total review duration.

[0068] Specifically, the server determines the review duration of each node in the submission review topology. The review duration refers to the time required for each expert subset to complete the review, which depends on factors such as the number of experts in the subset, their professional level, and work saturation. The server can analyze the historical review data of experts to establish a review duration prediction model, and predict the review duration of each node according to the characteristics of the current materials and the composition of the expert subset. The prediction methods include regression analysis and time series analysis.

[0069] Next, the server calculates the estimated total review duration of the submission materials based on the review duration of each node. The estimated total review duration refers to the total time required for the materials to complete the entire review process, which is equal to the sum of the review durations of all nodes in the submission review topology. The server can use methods such as the critical path algorithm and network flow algorithm to calculate the estimated total review duration of the submission materials on the submission review topology. The estimated total review duration represents the shortest time required for the submission materials to be finalized from submission.

[0070] Finally, the server sends the estimated total review duration to the user so that the user can determine whether to adjust the review expert group of the submission materials according to their time requirements. The server can push the estimated total review duration to the user through methods such as email, text message, and system notification.

[0071] In step S107, submitting the submission materials for review according to the submission review topology specifically includes: determining the target review expert according to the submission review topology, where the target review expert is the first review expert in the submission review topology; sending the submission materials to the target review expert for review and receiving the review result of the target review expert; if it is determined that the review result is passed, sending the submission materials to the next review expert for review according to the submission review topology; if it is determined that the review result is not passed, feedback the review result to the user and end the submission review process.

[0072] Specifically, the server formally submits the materials to be reviewed according to the review topology. The purpose of this step is to send the materials to be reviewed to each reviewer in turn according to the expert order and transfer relationship set by the review topology, and collect their review results until the materials complete the entire review process. First, the server determines the first reviewer as the target reviewer according to the review topology. Specifically, the server finds the node with an in-degree of 0 in the review topology, that is, the node that no other node points to, as the starting node. Then, the server selects the first expert among the starting nodes as the target reviewer. This expert will be responsible for the initial review of the materials and give the first-round review opinions.

[0073] Next, the server sends the materials to be reviewed to the target reviewer for review and receives the review results of the target reviewer. The server can send the electronic version of the materials to be reviewed to the target reviewer by means of email, text message, system notification, etc. At the same time, the server also provides an online review platform for experts to submit review opinions and results. The review results usually include "pass" and "fail".

[0074] Then, the server decides the next transfer of the materials to be reviewed according to the review results of the target reviewer. If the review result is "pass", the server will automatically send the materials to be reviewed to the next node in the review topology for the next expert to continue the review. The server sends the materials to be reviewed to each reviewer in turn according to the connection relationship of the nodes in the review topology until all reviewers have completed the review. This process is similar to a workflow, and the materials to be reviewed are transferred orderly among experts, and the review results of each expert will affect the subsequent direction of the materials.

[0075] If the review result is "fail", the server will immediately terminate the review process of the materials and feedback the result to the user. The server will generate a comprehensive review report, summarize the review opinions and results of all experts, and send them to the user by means of email, text message, etc. At the same time, the server will also update the status of the materials to "not passed" in the user's personal center or workbench.

[0076] Refer to Figure 2 , this application also provides an intelligent review expert matching device, which is a server. The server includes an acquisition module 201 and a processing module 202, where: The acquisition module 201 is used to acquire the review materials submitted by the user; The processing module 202 is used to store the review materials in the data queue, and asynchronously extract the review materials from the data queue to obtain keywords; The processing module 202 is also used to calculate the similarity between the review materials and the preset expert library based on the keywords using the TF-IDF algorithm, and generate a first data pool. The first data pool includes multiple first experts; The processing module 202 is also used to analyze based on the first data pool in combination with historical review data to generate a second data pool. The second data pool includes multiple second experts; The processing module 202 is also used to calculate the intersection of the first data pool and the second data pool to obtain a review expert group. The review expert group includes multiple review experts; The processing module 202 is also used to generate a review topology for the review materials based on the review expert group; The processing module 202 is also used to conduct the review of the review materials according to the review topology.

[0077] In a possible implementation manner, the processing module 202 calculates the similarity between the review materials and the preset expert library based on the keywords using the TF-IDF algorithm, and generates a first data pool, which specifically includes: The processing module 202 performs word segmentation on the keywords to obtain a word segmentation result; The processing module 202 calculates the TF value and IDF value of the keywords based on the word segmentation result; The processing module 202 multiplies the TF value by the IDF value to obtain the TF-IDF value of the keywords; The processing module 202 calculates the similarity between the review materials and each expert in the preset expert library based on the TF-IDF value to obtain a similarity result; The processing module 202 takes the experts with a similarity greater than or equal to the preset similarity threshold as the first experts according to the similarity result, and adds the first experts to the first data pool.

[0078] In a possible implementation manner, the processing module 202 analyzes based on the first data pool in combination with historical review data to generate a second data pool. The second data pool includes multiple second experts, which specifically includes: The acquisition module 201 acquires historical review materials with the same research direction as the review materials; The processing module 202 counts the historical review experts corresponding to the historical review materials to obtain the review times corresponding to each historical review expert; The processing module 202 takes the historical review experts with a review time greater than or equal to the preset number threshold as the second experts, and adds the second experts to the second data pool.

[0079] In a possible implementation, after the processing module 202 calculates the intersection of the first data pool and the second data pool to obtain the review expert group to be submitted, the method further includes: the acquisition module 201 acquires the professional fields, review times, and review qualities of each review expert in the review expert group to be submitted; the processing module 202 calculates the comprehensive scores of each review expert using a weighted algorithm based on the professional fields, review times, and review qualities, and sorts the review expert group to be submitted according to the comprehensive scores to obtain the target review expert group.

[0080] In a possible implementation, the processing module 202 generates a submission topology for the submission materials based on the review expert group to be submitted, specifically including: the processing module 202 divides the target review expert group into multiple expert subsets, and each expert subset includes at least one review expert; the processing module 202 generates a directed five-ring graph based on the order of the expert subsets in the target review expert group as the submission topology of the submission materials, where the nodes in the submission topology represent the expert subsets, and the edges in the submission topology represent the review transfer relationships.

[0081] In a possible implementation, before the processing module 202 submits the submission materials according to the submission topology, the method further includes: the processing module 202 determines the review duration of each node in the submission topology; the processing module 202 calculates the estimated total review duration of the submission materials according to the review duration of each node; the processing module 202 sends the estimated total review duration to the user, so that the user can determine whether to adjust the review expert group of the submission materials according to the estimated total review duration.

[0082] In a possible implementation, the processing module 202 submits the submission materials according to the submission topology, specifically including: the processing module 202 determines the target review expert according to the submission topology, and the target review expert is the first review expert in the submission topology; the processing module 202 sends the submission materials to the target review expert for review and receives the review result of the target review expert; if the processing module 202 determines that the review result is passed, it sends the submission materials to the next review expert for review according to the submission topology; if the processing module 202 determines that the review result is not passed, it feedbacks the review result to the user and ends the submission process.

[0083] It should be noted that when the device provided in the above embodiments realizes its functions, only the above-mentioned division of each functional module is used for illustration. In practical applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process is detailed in the method embodiments, which will not be repeated here.

[0084] This application also provides an electronic device. Refer to Figure 3 , Figure 3 is a schematic structural diagram of an electronic device provided by an embodiment of the present application. The electronic device 300 may include: at least one processor 301, at least one network interface 304, a user interface 303, a memory 305, and at least one communication bus 302.

[0085] Among them, the communication bus 302 is used to realize the connection and communication between these components.

[0086] Among them, the user interface 303 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 303 may further include a standard wired interface and a wireless interface.

[0087] Among them, the network interface 304 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface).

[0088] Among them, the processor 301 may include one or more processing cores. The processor 301 uses various interfaces and lines to connect various parts within the entire server. By running or executing instructions, programs, code sets, or instruction sets stored in the memory 305, and by calling data stored in the memory 305, the processor 301 executes various functions of the server and processes data. Optionally, the processor 301 may be implemented in at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), or programmable logic array (PLA). The processor 301 may integrate one or several combinations of a central processing unit (CPU), a graphics processing unit (GPU), and a modem, etc. Among them, the CPU mainly processes the operating system, user interface, application programs, etc.; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communication. It can be understood that the above modem may not be integrated into the processor 301 and may be implemented separately by a single chip.

[0089] Among them, the memory 305 may include a Random Access Memory (RAM), or may also include a Read-Only Memory. Optionally, the memory 305 includes a non-transitory computer-readable storage medium. The memory 305 can be used to store instructions, programs, codes, code sets, or instruction sets. The memory 305 may include a program storage area and a data storage area. Among them, the program storage area can store instructions for implementing the operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned method embodiments, etc.; the data storage area can store the data involved in the above-mentioned method embodiments. Optionally, the memory 305 may also be at least one storage device located far from the aforementioned processor 301. Refer to Figure 3 In the memory 305, as a computer storage medium, it may include an operating system, a network communication module, a user interface module, and an application program of an intelligent submission expert matching method.

[0090] In Figure 3 In the electronic device 300 shown, the user interface 303 is mainly used to provide an input interface for the user to obtain the data input by the user; and the processor 301 can be used to call the application program of an intelligent submission expert matching method stored in the memory 305. When executed by one or more processors 301, the electronic device 300 is caused to execute one or more of the methods as described in the above embodiments. It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present application is not limited by the described action sequence, because according to the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present application.

[0091] The present application also provides a computer-readable storage medium storing instructions. When executed by one or more processors 301, the electronic device 300 is caused to execute one or more of the methods as described in the above embodiments.

[0092] In the above embodiments, the descriptions of the respective embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0093] In several implementation manners provided by this application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of units is only a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling, direct coupling, or communication connection to each other can be through some service interfaces. The indirect coupling or communication connection of devices or units can be in electrical or other forms.

[0094] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0095] In addition, each functional unit in various embodiments of this application can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0096] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on such an understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in various embodiments of this application. And the aforementioned memory includes: various media such as USB flash drives, mobile hard disks, magnetic disks, or optical discs that can store program codes.

[0097] The above are only exemplary embodiments of the present disclosure and should not be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made in accordance with the teachings of the present disclosure still fall within the scope covered by the present disclosure. Those skilled in the art will easily think of other implementation schemes of the present disclosure after considering the specification and the disclosure of the practical truth.

[0098] This application aims to cover any variations, uses, or adaptive changes of the present disclosure. These variations, uses, or adaptive changes follow the general principles of the present disclosure and include common general knowledge or conventional technical means in the technical field not recorded in the present disclosure. The specification and the embodiments are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.< / uuid> < / uuid>

Claims

1. An intelligent review expert matching method, characterized in that: The method comprises: Obtain review materials submitted by users; The submitted materials are stored in a data queue, and the submitted materials are asynchronously extracted from the data queue to obtain keywords; Based on the keywords, the similarity between the submitted materials and the preset expert database is calculated using the TF-IDF algorithm to generate a first data pool, where the first data pool includes a plurality of first experts; Based on the first data pool, combined with historical review data, analysis is performed to generate a second data pool, wherein the second data pool includes a plurality of second experts; Performing intersection calculation on the first data pool and the second data pool to obtain a review expert group, wherein the review expert group includes multiple review experts; Based on the review expert group, generating a review topology for the review materials; The submission materials are submitted for review according to the submission topology.

2. The method according to claim 1, characterized in that Based on the keywords, the similarity between the submitted materials and the preset expert database is calculated using the TF-IDF algorithm to generate the first data pool, which specifically includes: Performing word segmentation processing on the keywords to obtain word segmentation results; Based on the word segmentation result, calculate the TF value and IDF value of the keyword; Multiplying the TF value by the IDF value to obtain the TF-IDF value of the keyword; Based on the TF-IDF value, the similarity between the submitted material and each expert in the preset expert database is calculated to obtain a similarity result; According to the similarity result, an expert whose similarity is greater than or equal to a preset similarity threshold is taken as a first expert, and the first expert is added to the first data pool.

3. The method according to claim 1, characterized in that The first data pool is analyzed in combination with historical review data to generate a second data pool, wherein the second data pool includes a plurality of second experts, specifically including: Obtain historical submissions with the same research direction as the submissions in question; Count the historical review experts corresponding to the historical review materials, and obtain the number of submissions corresponding to each of the historical review experts; The historical review expert whose number of submissions is greater than or equal to a preset number threshold is used as the second expert, and the second expert is added to the second data pool.

4. The method according to claim 1, characterized in that: After calculating the intersection of the first data pool and the second data pool to obtain the review expert group, the method further includes: Obtain the professional field, review time and review quality of each review expert in the review expert group; Based on the professional field, the review time and the review quality, a weighted algorithm is used to calculate the comprehensive score of each review expert, and the review expert group is sorted according to the comprehensive score to obtain the target review expert group.

5. The method according to claim 4, characterized in that The generating of a review topology for the review materials based on the review expert group specifically includes: Dividing the target review expert group into a plurality of expert subsets, each of the expert subsets including at least one review expert; Based on the order of the expert subsets in the target review expert group, a directed five-ring graph is generated as the review topology of the review materials. The nodes in the review topology represent the expert subsets, and the edges in the review topology represent the review flow relationship.

6. The method according to claim 5, characterized in that Before submitting the submission materials for review according to the submission topology, the method further includes: Determine the review time of each node in the review topology; Calculate the estimated total review time of the submitted materials based on the review time of each node; The estimated total review time is sent to the user, so that the user can determine whether to adjust the review expert group of the submitted materials according to the estimated total review time.

7. The method according to claim 1, characterized in that The submitting the submitted materials for review according to the submitted materials for review topology specifically includes: Determine a target review expert according to the review topology, where the target review expert is the first review expert in the review topology; Send the submitted materials to the target submission expert for review, and receive the review results of the target submission expert; If the review result is determined to be passed, the review materials are sent to the next review expert for review according to the review topology; If it is determined that the review result is not passed, the review result will be fed back to the user and the review process will be terminated.

8. An intelligent review expert matching device, characterized in that: The device comprises an acquisition module (201) and a processing module (202), wherein: The acquisition module (201) is used to acquire the review materials submitted by the user; The processing module (202) is used to store the submitted materials in a data queue and asynchronously extract the submitted materials from the data queue to obtain keywords; The processing module (202) is further used to calculate the similarity between the submitted materials and the preset expert database based on the keywords using the TF-IDF algorithm to generate a first data pool, wherein the first data pool includes a plurality of first experts; The processing module (202) is further used to generate a second data pool based on the first data pool and in combination with historical review data, wherein the second data pool includes a plurality of second experts; The processing module (202) is further used to perform intersection calculation on the first data pool and the second data pool to obtain a review expert group, wherein the review expert group includes a plurality of review experts; The processing module (202) is further used to generate a review topology for the review materials based on the review expert group; The processing module (202) is also used to submit the submission materials for review according to the submission topology.

9. An electronic device, characterized in that: The electronic device (300) comprises a processor (301), a memory (305), a user interface (303) and a network interface (304), wherein the memory (305) is used to store instructions, the user interface (303) and the network interface (304) are used to communicate with other devices, and the processor (301) is used to execute the instructions stored in the memory (305) so that the electronic device (300) executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed, the method according to any one of claims 1 to 7 is performed.

Images