Evaluation expert extraction method based on multiple layers of disciplines and multiple precision

Abstract

The invention provides a multi-layer subject and multi-precision-based review expert extraction method, which comprises the following steps of: constructing a multi-layer subject knowledge graph G = (V, E), obtaining a project feature vector P = (p1, p2,... pn) of a project to be reviewed, and screening a subject knowledge unit set S related to the project to be reviewed according to the multi-layer subject knowledge graph G = (V, E) and correlation Rij; the precision threshold value theta is compared with the correlation degree Rij, and when the correlation degree Rij is larger than or equal to theta, the subject knowledge units are included into a subject knowledge unit set S; and in an expert database related to the subject knowledge unit set S, extracting a review expert according to the knowledge background of the expert and the characteristics of the project to be reviewed. According to the method, the evaluation experts are extracted through the correlation Rij calculation formula so as to improve the precision control of the evaluation of the interdisciplinary to-be-evaluated project.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular, to a method for extracting review experts based on multiple disciplines and multiple precisions. Background Art

[0002] In activities such as reviewing numerous scientific research projects to be reviewed, evaluating academic achievements, and selecting professional competitions, accurately extracting appropriate review experts is crucial. Most of the existing review expert extraction methods are only based on knowledge in a single discipline field and are difficult to meet the review requirements of interdisciplinary projects to be reviewed. For example, in the field of bioinformatics, which integrates knowledge from multiple disciplines such as biology, computer science, and mathematics, the existing single-discipline extraction methods cannot comprehensively consider the comprehensive application of different disciplinary knowledge in this field, resulting in the experts extracted being unable to conduct a comprehensive and accurate evaluation of the project to be reviewed.

[0003] In addition, the existing methods also have deficiencies in terms of precision. They often use simple keyword matching or general expert database screening and do not conduct refined screening according to the complexity and special requirements of the specific project to be reviewed. For example, for some cutting-edge and innovative projects, whose technical details and research directions are unique, the conventional extraction methods cannot accurately locate experts with in-depth research in this specific direction, thus affecting the accuracy and professionalism of the review results.

[0004] In summary, the technical problem actually solved by the present invention is how to improve the precision control of the review of interdisciplinary projects to be reviewed. Summary of the Invention

[0005] In order to overcome the above-mentioned deficiencies in the precision control of the review of interdisciplinary projects in the prior art, the purpose of the present invention is to provide a method for extracting review experts based on multiple disciplines and multiple precisions, which extracts review experts through the correlation degree R ij calculation formula to improve the precision control of the review of interdisciplinary projects to be reviewed.

[0006] The present invention discloses a method for extracting review experts based on multiple disciplines and multiple precisions, including the following steps:

[0007] Construct a multi-disciplinary knowledge graph G=(V, E), where V represents the set of nodes, each node represents a disciplinary knowledge unit, and E represents the set of edges, and the edges represent the association relationships between the disciplinary knowledge units;

[0008] Obtain the project feature vector P=(p1, p2,... p n ) of the project to be reviewed, where p i represents the value of the project to be reviewed on the i-th feature dimension;

[0009] Calculate the correlation degree R between the project feature vector P and each disciplinary knowledge unit node according to the multi-layer disciplinary knowledge graph G ij , and the calculation formula is:

[0010]

[0011] where w k represents the weight of the k-th feature dimension, m represents the total number of feature dimensions, u jk represents the value of the j-th disciplinary knowledge unit node on the k-th feature dimension, and f represents the function for calculating the correlation degree;

[0012] According to the correlation degree R ij screen the set S of disciplinary knowledge units related to the project to be reviewed;

[0013] Set the precision threshold θ according to the precision requirement, and compare the precision threshold θ with the correlation degree R ij . When the correlation degree R ij ≥θ, include the disciplinary knowledge unit in the set S of disciplinary knowledge units;

[0014] Extract review experts from the expert database related to the set S of disciplinary knowledge units according to the knowledge background of the experts and the characteristics of the project to be reviewed.

[0015] Preferably, in the step of constructing the multi-layer disciplinary knowledge graph G=(V, E), it specifically includes the following steps:

[0016] Collect literature materials in multiple disciplinary fields;

[0017] Perform text mining on the literature materials to extract disciplinary knowledge units;

[0018] Analyze the mutual relationship between disciplinary knowledge units to determine the edge set E.

[0019] Preferably, in the step of obtaining the project feature vector P of the project to be reviewed, it specifically includes the following steps:

[0020] Analyze and extract the project document to be reviewed to obtain processing information;

[0021] Convert the processing information into the project feature vector P, where each feature dimension corresponds to a key attribute of the project to be reviewed.

[0022] Preferably, the function for calculating the correlation degree R ij is the cosine similarity function, that is:

[0023]

[0024] Preferably. The setting method of the precision threshold θ includes the following steps:

[0025] Manually set the precision threshold according to the grading of the project to be reviewed, and / or, set the precision threshold θ through a machine learning algorithm according to the historical data of the project to be reviewed and the review results.

[0026] Preferably, in the expert database related to the set S of disciplinary knowledge units, the steps of extracting review experts according to the expert's knowledge background and the characteristics of the project to be reviewed specifically include the following steps:

[0027] Construct an expert knowledge graph to establish an association between experts and disciplinary knowledge units;

[0028] Calculate the matching degree between the expert and each disciplinary knowledge unit in the set S of disciplinary knowledge units;

[0029] Generate a matching table by sorting in descending order according to the matching degree, and sequentially select the experts in the matching table as review experts.

[0030] Preferably, the steps of constructing the expert knowledge graph specifically include the following steps:

[0031] Collect the academic information of experts;

[0032] Analyze the academic information to extract the research fields of experts and the disciplinary knowledge units involved;

[0033] Establish an association relationship between experts and disciplinary knowledge units.

[0034] Preferably, the steps of calculating the matching degree between the expert and each disciplinary knowledge unit in the set S of disciplinary knowledge units specifically include the following steps:

[0035] Based on the weighted average method, set the weights according to the research depth of the expert in each disciplinary knowledge unit;

[0036] Calculate the similarity between the research results of the expert in each disciplinary knowledge unit and the corresponding disciplinary knowledge unit in the set S of disciplinary knowledge units, and obtain the matching degree according to the weights.

[0037] Preferably, after the steps of extracting review experts according to the expert's knowledge background and the characteristics of the project to be reviewed in the expert database related to the set S of disciplinary knowledge units, the following steps are further included:

[0038] Check the conflict of interest information of the expert. If there is an interest association between the expert and the project to be reviewed, the expert is excluded, and / or, evaluate the time availability of the expert. If the expert cannot participate in the review during the review time period, the expert is excluded.

[0039] After adopting the above technical solution, compared with the prior art, an extraction method for review experts based on multiple disciplines and multiple precisions of the present invention, through the correlation degree R ijCalculation formula extraction of review experts to improve the precision control of interdisciplinary projects to be reviewed; specifically, by constructing a multi-layer disciplinary knowledge graph G=(V, E), which comprehensively covers multi-disciplinary knowledge units, can comprehensively consider the application of different disciplinary knowledge in the projects to be reviewed, avoids the problem of incomplete evaluation caused by extracting experts only based on single-disciplinary knowledge, and ensures that the extracted review experts have the ability to comprehensively evaluate interdisciplinary projects.

[0040] Abandon simple keyword matching or database screening methods, and calculate the correlation degree R between the project feature vector P of the project to be reviewed and the disciplinary knowledge unit nodes ij , and set a precision threshold θ according to different precision requirements to screen the relevant disciplinary knowledge unit set S. For projects to be reviewed with frontier, innovative, technical details and unique research directions, it can accurately locate experts with in-depth research in specific directions, thus significantly improving the accuracy and professionalism of the review results.

[0041] Provide two ways of manually setting and automatically determining the precision threshold θ by machine learning algorithms. For projects to be reviewed with high importance and large complexity such as national major scientific research projects to be reviewed, a higher precision threshold θ can be manually set to ensure a high matching degree between the extracted review experts and the projects to be reviewed; for regular small projects to be reviewed, use machine learning algorithms to automatically determine an appropriate precision threshold θ according to historical data.

[0042] When extracting experts, construct a multi-layer disciplinary knowledge graph G=(V, E) and calculate the matching degree between the experts and the relevant disciplinary knowledge unit set, and select experts according to the matching degree ranking, which ensures a high fit between the extracted experts and the relevant disciplinary knowledge units of the projects to be reviewed, and further improves the accuracy and professionalism of the review.

[0043] Conduct conflict-of-interest checks and time availability assessments on the extracted experts, exclude experts with interest associations with the applicants of the projects to be reviewed and those who are unable to participate in the review during the review period, ensure the fairness and feasibility of the review process, and guarantee the quality of the review. Brief Description of the Drawings

[0044] Figure 1 It is a schematic diagram of the steps of a method for extracting review experts based on multi-layer disciplines and multi-precision of the present invention. Detailed Embodiment

[0045] The advantages of the present invention are further elaborated below in conjunction with the drawings and specific embodiments.

[0046] Exemplary embodiments will be described in detail herein, and examples thereof are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0047] The terms used in the present disclosure are for the purpose of describing particular embodiments only and are not intended to limit the present disclosure. The singular forms "a", "the", and "said" used in the present disclosure and the appended claims are also intended to include the plural forms unless the context clearly dictates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0048] It should be understood that although the terms first, second, third, etc. may be used in the present disclosure to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, without departing from the scope of the present disclosure, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "when" or "upon" or "in response to determining".

[0049] In the description of the present invention, it should be understood that the orientation or positional relationships indicated by the terms "longitudinal", "transverse", "up", "down", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. are based on the orientation or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation of the present invention.

[0050] In the description of the present invention, unless otherwise specified and defined, it should be noted that the terms "mounted", "connected", and "coupled" should be understood in a broad sense. For example, it may be a mechanical connection or an electrical connection, or it may be the communication inside two elements. It may be directly connected or indirectly connected through an intermediate medium. For those of ordinary skill in the art, the specific meanings of the above terms can be understood according to specific circumstances.

[0051] In the subsequent description, the suffixes such as "module", "component", or "unit" used to represent elements are only for the convenience of describing the present invention, and they do not have a specific meaning in themselves. Therefore, "module" and "component" can be used interchangeably.

[0052] A method for extracting review experts based on multi-layer disciplines and multi-precision includes the following steps: constructing a multi-layer discipline knowledge graph G=(V, E), where V represents the set of nodes, and each node represents a discipline knowledge unit, and E represents the set of edges, and the edges represent the association relationships between discipline knowledge units; obtaining the project feature vector P=(p1, p2,... p n ) of the project to be reviewed, where p i represents the value of the project to be reviewed on the i-th feature dimension; according to the multi-layer discipline knowledge graph G=(V, E), calculating the correlation degree R ij between the project feature vector P and each discipline knowledge unit node, and the calculation formula is: where, w k represents the weight of the k-th feature dimension, m represents the total number of feature dimensions, u jk represents the value of the j-th discipline knowledge unit node on the k-th feature dimension, and f represents the function for calculating the correlation degree; screening the set S of discipline knowledge units related to the project to be reviewed according to the correlation degree R ij ; setting a precision threshold θ according to the precision requirement, and comparing the precision threshold θ with the correlation degree R ij , when the correlation degree R ij ≥θ, incorporating the discipline knowledge unit into the set S of discipline knowledge units; extracting review experts from the expert database related to the set S of discipline knowledge units according to the knowledge background of the experts and the characteristics of the project to be reviewed.

[0053] Refer to Figure 1 as shown, in this embodiment, a method for extracting review experts based on multi-layer disciplines and multi-precision will be described in detail, which includes the following steps:

[0054] Step S100: In this step, a multi-layer disciplinary knowledge graph G=(V, E) will be constructed. Collect materials such as professional literature, research reports, and academic works in multiple disciplinary fields. These materials cover the knowledge information of each discipline and are the basic data sources for constructing the knowledge graph. Use text mining techniques, such as named entity recognition and relation extraction in natural language processing, to extract disciplinary knowledge units from the collected materials. Disciplinary knowledge units can be specific disciplinary concepts, technical methods, research directions, etc. Analyze the semantic relationships (such as inclusion relationships, causal relationships, etc.) and citation relationships (such as citations between papers) between disciplinary knowledge units to determine the edge set E. Edges represent the association relationships between disciplinary knowledge units, and through these association relationships, a graph reflecting the multi-disciplinary knowledge structure can be constructed, comprehensively covering multi-disciplinary knowledge units and their association relationships. For example, when facing projects that integrate multi-disciplinary knowledge such as bioinformatics, it is possible to comprehensively consider the application of different disciplinary knowledge in the project to be reviewed, avoiding the problem of incomplete evaluation caused by extracting experts based only on knowledge in a single disciplinary field, and ensuring that the extracted experts have the ability to comprehensively evaluate interdisciplinary projects.

[0055] Step S200: In this step, calculate the project feature vector P=(p1, p2,... p n ) of the project to be reviewed. Extract keywords from the project documents to be reviewed (such as project applications, technical solutions, research reports, etc.), and use methods such as word frequency statistics and TF-IDF to find the key information in the project to be reviewed. Adopt techniques such as topic model analysis (such as LDA model) to mine the topic information of the project to be reviewed, and further clarify the core content and research direction of the project to be reviewed. Convert the extracted keywords and topic information into the project feature vector P, where each feature dimension p i corresponds to a key attribute of the project to be reviewed, such as technological innovation, application field, research difficulty, etc.

[0056] Step S300: Calculate the correlation degree R ij between the project feature vector P of the project to be reviewed and each disciplinary knowledge unit node. Specifically, for each feature dimension k, assign a weight w k The determination of the weight can be carried out according to methods such as expert experience and historical data statistics to reflect the importance of different feature dimensions in the evaluation of the project to be reviewed; for each disciplinary knowledge unit node j, obtain its value u jk on the k-th feature dimension; use the formula to calculate the correlation degree R ij , where m is the total number of feature dimensions and f is the function for calculating the correlation degree.

[0057] Step S400: Screen the set S of disciplinary knowledge units related to the project to be reviewed according to the correlation degree R ij .

[0058] Step S500: Set the precision threshold θ according to different precision requirements. The higher the precision requirement, the larger the set precision threshold θ; traverse all disciplinary knowledge unit nodes, and when the correlation degree R ij ≥θ, incorporate the corresponding disciplinary knowledge units into the disciplinary knowledge unit set S.

[0059] Step S600: In this step, extract review experts from the database related to the disciplinary knowledge unit set S. Specifically, establish the association between experts and disciplinary knowledge units based on information such as experts' academic papers, scientific research projects, and professional skills, and construct an expert knowledge graph; calculate the matching degree between experts and each disciplinary knowledge unit in the disciplinary knowledge unit set S. The calculation of the matching degree comprehensively considers factors such as the research depth and the number of published achievements of experts in relevant disciplinary knowledge units; sort the experts from high to low according to the matching degree, and select the experts with the top rankings as review experts.

[0060] In the steps of constructing the multi-layer disciplinary knowledge graph G=(V, E), it specifically includes the following steps: collect literature materials in multiple disciplinary fields; perform text mining on the literature materials to extract disciplinary knowledge units; analyze the mutual relationships between disciplinary knowledge units to determine the edge set E.

[0061] In the above step S100, it specifically further includes the following steps:

[0062] Step S110: Collect materials such as professional literature, research reports, and academic conference papers in multiple disciplinary fields from multiple channels. These materials should cover the knowledge systems of relevant disciplines as comprehensively as possible to ensure that the scientific knowledge graph G=(V, E) can accurately reflect the associations between disciplines.

[0063] Step S120: Use natural language processing tools and technologies to preprocess the collected materials, such as word segmentation, part-of-speech tagging, etc.; then apply a named entity recognition algorithm to identify the disciplinary knowledge units in the materials, such as disciplinary terms, technology names, etc.

[0064] Step S130: Determine the edge set E by analyzing the semantic relationships and citation relationships between disciplinary knowledge units. For semantic relationships, technologies such as semantic networks and ontologies can be used for analysis; for citation relationships, information such as the number of citations and citation directions between papers can be statistically analyzed. According to the analysis results, establish corresponding association edges between disciplinary knowledge units to form the edge set E of the multi-layer disciplinary knowledge graph G=(V, E).

[0065] In the step of obtaining the project feature vector P of the project to be reviewed, the following steps are specifically included: analyzing and extracting the project document to be reviewed to obtain processing information; converting the processing information into the project feature vector P, where each feature dimension corresponds to a key attribute of the project to be reviewed.

[0066] In the above step S200, the following steps are specifically further included:

[0067] Step S210: Clean the project document to be reviewed, removing useless symbols, stop words, etc.; then use methods such as term frequency - inverse document frequency (TF - IDF) statistics to find the keywords with higher occurrence frequencies and representativeness in the project document to be reviewed. These keywords can reflect the main content and key information of the project to be reviewed.

[0068] Step S220: Use topic model algorithms such as LDA to model the project document to be reviewed; by analyzing the co - occurrence relationship of words in the project document to be reviewed, excavate the research direction and core content of the project to be reviewed; then determine the feature dimensions of the project feature vector P and the values of each feature dimension according to the extracted keywords and topic information. In some embodiments, the occurrence frequency of keywords, the probability distribution of topics, etc. can be used as elements of the project feature vector P, so as to convert the text information of the project to be reviewed into a digital project feature vector P.

[0069] Calculate the correlation degree R ij The function for calculating it is the cosine similarity function, that is:

[0070] In this embodiment, a detailed description will be given to the function for calculating the above - mentioned correlation degree R ij The function for calculating it is the cosine similarity function. For the element p i of the project feature vector P in the k - th feature dimension and the element u jk of the disciplinary knowledge unit node j in this feature dimension, calculate the sum of their products Respectively calculate the square root of the sum of the squares of the elements of the project feature vector P in the k - th feature dimension and the square root of the sum of the squares of the elements of the disciplinary knowledge unit node j in the special feature dimension Then multiply these two square roots to get the denominator; finally, divide the numerator by the denominator to obtain the cosine similarity f(p k , u jk ) between the project feature vector P and the scientific knowledge unit node j in the k - th feature dimension.

[0071] The method for setting the precision threshold θ includes the following steps: manually setting the precision threshold according to the grading of the project to be reviewed, and / or setting the precision threshold θ through machine learning algorithms according to historical data and review results of the projects to be reviewed.

[0072] In the expert database related to the set S of disciplinary knowledge units, the steps of extracting review experts according to the knowledge backgrounds of experts and the characteristics of projects to be reviewed specifically include the following steps: constructing an expert knowledge graph to establish an association between experts and disciplinary knowledge units; calculating the matching degrees between experts and each disciplinary knowledge unit in the set S of disciplinary knowledge units; generating a matching table by sorting the matching degrees in descending order, and successively selecting the experts in the matching table as review experts.

[0073] The method of setting the precision threshold θ in the above step S500 is described in detail. The method of setting the precision threshold θ mainly includes manual and machine learning algorithm settings. The manual setting includes, but is not limited to, manual judgment and setting by domain experts or review organizers according to factors such as the complexity, importance, and review requirements of the project to be reviewed. When setting by machine learning algorithms, the machine learning algorithms collect historical project data and corresponding review results, and use these data as the training set. Select appropriate machine learning algorithms, such as decision trees, neural networks, etc., to train the training set. During the training process, the algorithm will learn the relationship between project characteristics and review effects, so as to automatically determine an appropriate precision threshold θ. In practical applications, the characteristics of the project to be reviewed are input into the trained model, and the model can output the corresponding precision threshold θ.

[0074] In the expert database related to the set S of disciplinary knowledge units, the steps of extracting review experts according to the knowledge backgrounds of experts and the characteristics of projects to be reviewed specifically include the following steps: constructing an expert knowledge graph to establish an association between the experts and the disciplinary knowledge units; calculating the matching degrees between the experts and each disciplinary knowledge unit in the set S of disciplinary knowledge units; generating a matching table by sorting the matching degrees in descending order, and successively selecting the experts in the matching table as review experts.

[0075] The following steps are specifically included in the above step S600:

[0076] Step S610: Collect information such as the academic papers, research projects, and professional skills of experts, and organize and analyze this information. Use knowledge graph construction technology to establish an association between experts and disciplinary knowledge units to form an expert knowledge graph. In the graph, there is a connection relationship between each expert node and the disciplinary knowledge unit nodes they are involved in, reflecting the knowledge background and research fields of the experts.

[0077] Step S620: For each disciplinary knowledge unit in the disciplinary knowledge unit set S, calculate the matching degree between the expert and the disciplinary knowledge unit. The calculation of the matching degree can comprehensively consider factors such as the research depth of the expert in this knowledge unit (such as the number of published papers, citation times, etc.), the relevance of research results, etc. Methods such as weighted average can be used to comprehensively calculate multiple factors to obtain the overall matching degree between the expert and each disciplinary knowledge unit in the disciplinary knowledge unit set S.

[0078] Step S630: According to the calculated matching degree, sort the experts from high to low. Select the experts with the top rankings as the review experts according to the number of experts required for the review.

[0079] The steps in constructing the expert knowledge graph specifically include the following steps: Collect the academic information of experts; analyze the academic information to extract the research fields of experts and the disciplinary knowledge units involved; establish the association relationship between experts and disciplinary knowledge units.

[0080] In the above step S610, it specifically includes the following steps:

[0081] Step S611: Collect information such as the academic papers, research projects, patents, professional skill certificates, etc. of experts from multiple channels. This information can come from academic databases, scientific research institution websites, experts' personal home pages, etc.

[0082] Step S612: Conduct text analysis on the collected expert information, and use technologies such as keyword extraction and topic model analysis to determine the research fields of experts and the disciplinary knowledge units involved. In some embodiments, by analyzing the contents such as the titles and abstracts of experts' papers, relevant disciplinary terms and research directions are extracted.

[0083] Step S613: According to the analysis results, establish the association relationship between experts and disciplinary knowledge units. Tools such as graph databases can be used to store and manage these association relationships to form the expert knowledge graph. In the expert knowledge graph, the connection between the expert node and the disciplinary knowledge unit node indicates that the expert has certain research or professional knowledge in this disciplinary knowledge unit.

[0084] The steps in calculating the matching degree between the expert and each disciplinary knowledge unit in the disciplinary knowledge unit set S specifically include the following steps: Based on the weighted average method, set weights according to the research depth of the expert in each disciplinary knowledge unit; calculate the similarity between the research results of the expert in each disciplinary knowledge unit and the corresponding disciplinary knowledge unit in the disciplinary knowledge unit set S, and obtain the matching degree according to the weights.

[0085] In the above step S620, it specifically includes the following steps:

[0086] Step S621: Set corresponding weights for each disciplinary knowledge unit according to the research depth of experts in each disciplinary knowledge unit. The research depth can be measured by indicators such as the number of published papers, citation times, and the scale of research projects undertaken. The greater the research depth, the higher the weight.

[0087] Step S622: For each disciplinary knowledge unit in the disciplinary knowledge unit set S, calculate the similarity between the research results of the expert in this knowledge unit and this knowledge unit. The calculation of similarity can adopt text similarity algorithms (such as cosine similarity, edit distance, etc.), and compare the similarity between the content of the expert's paper abstract, research report, etc. and the description of the disciplinary knowledge unit.

[0088] Step S623: Multiply the similarity of each disciplinary knowledge unit by the corresponding weight, and then sum them up to obtain the matching degree between the expert and each disciplinary knowledge unit in the disciplinary knowledge unit set S.

[0089] After the step of extracting review experts from the expert database related to the disciplinary knowledge unit set S according to the expert's knowledge background and the characteristics of the project to be reviewed, the following steps are also included: Check the conflict of interest information of the expert. If the expert has an interest connection with the project to be reviewed, then exclude the expert, and / or, evaluate the time availability of the expert. If the expert cannot participate in the review during the review time period, then exclude the expert.

[0090] After the above step S600, the following steps are also included:

[0091] Step S700: Check the conflict of interest between the extracted expert and the project to be reviewed, including but not limited to collecting relevant information of the expert and the applicant of the project to be reviewed, including cooperation relationships, economic interest exchanges, kinship, etc. Verify this information by querying enterprise databases, academic cooperation records, etc. If it is found that the extracted expert has an interest connection with the applicant of the project to be reviewed, such as the enterprise where the expert is located has a cooperation project with the applicant of the project to be reviewed, the expert holds shares in the applicant of the project to be reviewed, etc., then exclude this expert, and / or, check the time availability of the extracted expert for the project to be reviewed, including but not limited to communicating with the extracted expert to understand their work arrangements and time plans during the review time period. Their time arrangements can be directly obtained by sending questionnaires, making phone calls, etc. If the extracted expert has important meetings, business trips, etc. during the time period of the project to be reviewed, then exclude this expert.

[0092] It should be noted that the embodiments of the present invention have better implementability and do not impose any form of limitation on the present invention. Any person skilled in the art may use the technical content disclosed above to modify or transform it into equivalent effective embodiments. However, as long as the content of the technical solution of the present invention is not departed from, any modification, equivalent change or modification made to the above embodiments based on the technical essence of the present invention still falls within the scope of the technical solution of the present invention.

Claims

1. A method for extracting review experts based on multiple disciplines and multiple precisions, characterized in that Including the following steps: Construct a multi-layer disciplinary knowledge graph G=(V, E), where V represents the set of nodes, each of the nodes represents a disciplinary knowledge unit, E represents the set of edges, and the edges represent the association relationships between the disciplinary knowledge units; Obtain the project feature vector P=(p1, p2, … p n ), where p i represents the value of the to-be-reviewed project on the i-th feature dimension; According to the multi-layer disciplinary knowledge graph G=(V, E), calculate the association degree R between the project feature vector P and each disciplinary knowledge unit node ij , and the calculation formula is: Among them, w k represents the weight of the k-th feature dimension, m represents the total number of feature dimensions, and u jk represents the value of the j-th subject knowledge unit node on the k-th feature dimension, and f represents a function for calculating the correlation degree; According to the relevance degree R ij Screen the set S of disciplinary knowledge units related to the project to be reviewed; Set the precision threshold θ according to the precision requirement, and compare the precision threshold θ with the correlation degree R ij When the correlation degree R ij ≥ θ, incorporate the disciplinary knowledge unit into the disciplinary knowledge unit set S; In the expert database related to the disciplinary knowledge unit set S, extract review experts according to the knowledge backgrounds of the experts and the characteristics of the project to be reviewed.

2. The method for extracting review experts based on multiple disciplines and multiple precisions according to claim 1, wherein, The step of constructing the multi-layer disciplinary knowledge graph G=(V, E) specifically includes the following steps: Collect literature materials in multiple disciplinary fields; Perform text mining on the literature materials to extract the disciplinary knowledge units; Analyze the mutual relationships between the disciplinary knowledge units to determine the edge set E.

3. The method for extracting review experts based on multiple disciplines and multiple precisions according to claim 1, wherein The step of obtaining the project feature vector P of the project to be reviewed specifically includes the following steps: Analyze and extract the project to be reviewed to obtain processing information; Convert the processing information into the project feature vector P, where each feature dimension corresponds to a key attribute of the project to be reviewed.

4. The method for extracting review experts based on multiple disciplines and multiple precisions according to claim 1, wherein The calculated correlation degree R ij uses the cosine similarity function, which is:

5. The method for extracting review experts based on multi-layer disciplines and multi-precision according to claim 1, characterized in that The method for setting the accuracy threshold θ includes the following steps: Manually set the accuracy threshold according to the classification of the project to be reviewed, and / or, set the accuracy threshold θ through a machine learning algorithm according to the historical data and review results of the project to be reviewed.

6. The method for extracting review experts based on multiple disciplines and multiple precisions according to claim 1, wherein The step of extracting review experts according to the knowledge backgrounds of the experts and the characteristics of the project to be reviewed in the expert database related to the disciplinary knowledge unit set S specifically includes the following steps: Construct an expert knowledge graph and establish an association between the experts and the disciplinary knowledge units; Calculate the matching degrees between the experts and each disciplinary knowledge unit in the disciplinary knowledge unit set S; Generate a matching table by sorting the matching degrees in descending order, and sequentially select the experts in the matching table as review experts.

7. The method for extracting review experts based on multiple disciplines and multiple precisions according to claim 6, characterized in that The step of constructing the expert knowledge graph specifically includes the following steps: Collect the academic information of the experts; Analyze the academic information to extract the research fields of the experts and the disciplinary knowledge units involved; Establish an association relationship between the experts and the disciplinary knowledge units.

8. The method for extracting review experts based on multiple disciplines and multiple precisions according to claim 6, characterized in that The step of calculating the matching degrees between the experts and each disciplinary knowledge unit in the disciplinary knowledge unit set S specifically includes the following steps: Based on the weighted average method, set weights according to the research depth of the experts in each disciplinary knowledge unit; Calculate the similarity between the research achievements of the experts in each disciplinary knowledge unit and the corresponding disciplinary knowledge unit in the disciplinary knowledge unit set S, and obtain the matching degree according to the weights.

9. The method for extracting review experts based on multiple disciplines and multiple precisions according to claim 1, wherein After the step of extracting review experts according to the knowledge backgrounds of the experts and the characteristics of the project to be reviewed in the expert database related to the disciplinary knowledge unit set S, the following steps are further included: Check the conflict of interest information of the experts. If there is an interest association between the expert and the project to be reviewed, then exclude the expert, and / or, evaluate the time availability of the expert. If the expert cannot participate in the review during the review time period, then exclude the expert.

Images