An evaluation and analysis method for the contribution of co-authored papers based on the authors' academic background

Abstract

The present invention belongs to the technical field of machine learning applications, and specifically relates to a method for evaluating and analyzing the contribution of co-authored papers based on the authors' academic backgrounds. First, the subject keywords of the co-authored paper are extracted, and then the authors' published papers are used to extract the authors' academic keywords. The semantic similarity between the two is calculated based on the constructed research field similarity evaluation and analysis model to determine the correlation between the research field of the co-authored paper theme and the authors' academic research field. Finally, the correlation is used to evaluate the authors' contribution to the co-authored paper. This method comprehensively considers the impact of academic background factors such as the authors' research field and the authors' academic keywords on the degree of participation in the paper, making up for the shortcomings of the existing technology in inaccurate evaluation of the actual academic contribution of paper co-authors.

Description

Technical Field

[0001] The present invention belongs to the technical field of machine learning applications, and in particular relates to a method for evaluating and analyzing the contribution of authors of co-authored papers based on the authors' academic backgrounds. Background Art

[0002] With the continuous deepening of tobacco scientific research activities and the continuous expansion of research fields, the complexity, comprehensiveness and interdisciplinary nature of tobacco scientific research issues have become increasingly prominent. More and more tobacco researchers have begun to pay attention to the integration of interdisciplinary knowledge. Scientific research cooperation has become the main trend in tobacco scientific research, and the scale and coverage of cooperation are also increasing. Scientific research cooperation is an important way for researchers to exchange technology, integrate knowledge, and share resources. Scientific research cooperation between researchers through co-authoring papers is a very common form of output. Through co-authoring papers, it can not only improve the quality of scientific research, speed up academic output, and expand academic influence, but also accelerate the dissemination of academic knowledge and the promotion of results. In recent years, the rapid development of mobile Internet technology has made academic exchanges between researchers in different research fields in the tobacco industry and between researchers in and outside the tobacco industry more and more convenient and fast. Scientific research cooperation is no longer restricted by geographical, institutional, and disciplinary conditions. Cooperation between researchers is becoming more and more frequent, and the number of co-authored papers has also increased significantly. The tobacco industry is a field that spans multiple disciplines. The evaluation and analysis of the contribution of co-authored papers in the tobacco field can not only assess the true academic ability of tobacco researchers, but also explore the relationships and cooperation between scholars. It is also conducive to objectively evaluating the scientific research output performance of tobacco researchers and building a database of scholars and experts in the tobacco field.

[0003] There has been no good solution to the problem of author contribution distribution in co-authored papers that can satisfy scholars in all disciplines. The tobacco field also faces such a problem. In the era of big science with increasingly close scientific research cooperation, it is urgent to conduct reasonable academic contribution evaluation and honor distribution for paper co-authors, maintain a good scientific research ethical order, and avoid academic achievement copyright disputes caused by academic misconduct such as false signatures and interference in professional title assessment and evaluation. At present, the calculation methods of co-author contribution can be generally divided into the following three types:

[0004] (1) Direct calculation method. This algorithm assigns all contributions to the paper to the first author or corresponding author. No matter how many other co-authors there are, no rights are assigned. Only the first author and corresponding author are considered to be the co-contributors of the paper. This method is simple to operate, but it ignores the contributions of other co-authors.

[0005] (2) Academic contribution weighting algorithm based on the number of authors. This algorithm assumes that all authors have the same contribution, and evenly distributes the contribution value according to the number of authors, without distinguishing the order of authors. This algorithm invisibly magnifies the contribution of most authors and is unfair to the real contributors of the paper.

[0006] (3) Academic contribution weight algorithm based on author signature order. This algorithm assumes that the signatures of all co-authored papers are arranged in the order of author contribution, and calculates the weight of the author contribution based on the signature order. It comprehensively considers the impact of the number of authors in the paper and the order of author signature on the contribution weight, but cannot correctly reflect the important contribution made by the corresponding author who is signed later to the paper.

[0007] In summary, there are still many shortcomings in the current methods for calculating the academic contribution of authors of co-authored papers. The existing methods focus more on the number of authors and the order of signatures, but these factors cannot truly reflect the actual contribution of authors to the co-authored papers. Summary of the invention

[0008] The purpose of the present invention is to provide a method for evaluating and analyzing the contribution of co-authored papers based on the authors' academic backgrounds, so as to solve the problem that the existing methods for evaluating the contribution of co-authored papers cannot accurately evaluate the actual contribution of paper co-authors to the co-authored papers.

[0009] In order to solve the above technical problems, the technical solutions provided by the present invention and the corresponding beneficial effects of the technical solutions are as follows:

[0010] The present invention provides a method for evaluating and analyzing the contribution of co-authored papers based on the authors' academic backgrounds, comprising the following steps:

[0011] 1) Obtain the co-authored papers to be analyzed and identify the authors of the co-authored papers to be analyzed;

[0012] 2) Extract several subject keywords of the co-authored papers to be analyzed to determine the research field of the co-authored papers to be analyzed;

[0013] 3) For a certain author of the co-authored paper to be analyzed, obtain the published papers and extract several academic keywords of the author from them to determine the academic research field of the author;

[0014] 4) Calculate the semantic similarity between several subject keywords of the co-authored paper to be analyzed and several academic keywords of the author to determine the correlation between the research field of the co-authored paper to be analyzed and the academic research field of the author, and use the determined similarity value as the contribution weight based on the academic background to evaluate the author's contribution to the co-authored paper to be analyzed i ;

[0015] 5) Determine the contribution of the author to the co-authored paper to be analyzed based on the signature order of the author in the co-authored paper to be analyzed and the contribution weight of the author; wherein, the greater the contribution weight, the greater the contribution.

[0016] The beneficial effect of the above technical solution is that the present invention takes into account the influence of the author's research field, the author's academic keywords and other academic backgrounds on the degree of participation in the co-authored paper, and comprehensively evaluates the author's contribution to the co-authored paper in combination with the author's signature order in the paper. Specifically, the influence of the author's research field, the author's academic keywords and other academic backgrounds on the degree of participation in the co-authored paper is calculated in the following manner: first, the subject keywords of the co-authored paper are extracted, and then the author's published papers are used to extract the author's academic keywords, and finally the semantic similarity between the two is calculated to determine the correlation between the research field of the co-authored paper theme and the author's academic research field. Compared with the existing technical methods, this method improves the comprehensiveness and accuracy of the calculation and allocation of the actual academic contribution of the paper co-authors, and makes up for the shortcomings of the existing technology in the inaccurate assessment of the actual academic contribution of the paper co-authors; moreover, this method can also be used to evaluate the true academic ability of tobacco scientific researchers, and then serve the fair and objective scientific research performance evaluation of scientific researchers in various fields and the construction of scholar expert databases in various fields.

[0017] Furthermore, in step 5), the contribution of a certain author to the co-authored paper to be analyzed is calculated using the following formula:

[0018] λ i =k i *ω i

[0019] In the formula, λ i represents the contribution of the i-th author determined in the order of signature; k i Represents the signature order weight determined according to the signature order of the i-th author in the co-authored paper to be analyzed.

[0020] The beneficial effect of the above technical solution is: taking the product of the signature order weight and the contribution weight as the author's contribution can reflect the principle that the greater the signature order weight, the greater the author's contribution, and the greater the contribution weight, the greater the author's contribution, and the calculation is simple.

[0021] Furthermore, the earlier the author's signature order is, the greater the weight of the author's signature order is.

[0022] Furthermore, if the author is an author other than the first author and the corresponding author, the signature order weight k of the i-th author is i for:

[0023]

[0024] Where N represents the total number of all authors of the co-authored papers to be analyzed, i≥2;

[0025] If the author is the first author, the first author's signature order weight k 1 for:

[0026]

[0027] In the formula, α represents the additional weight of the first author, 0<α<1;

[0028] If the author is the corresponding author, the corresponding author's signature order weight k i for:

[0029]

[0030] Where β represents the additional weight of the first author, 0<β<1.

[0031] The beneficial effect of the above technical solution is: considering that the first author and the corresponding author are the core authors of the co-authored paper, and generally they will undertake more work in the process of co-authoring the paper and make greater contributions to the co-authoring of the paper, a weight is added to the first author and the corresponding author relative to other authors to improve the accuracy of author evaluation.

[0032] Furthermore, in step 4), the means used to calculate the semantic similarity between several subject keywords of the co-authored paper to be analyzed and several academic keywords of the author are: based on several subject keywords of the co-authored paper to be analyzed, constructing the co-authored paper subject keyword word vector; based on several academic keywords of the author, constructing the author academic keyword word vector; using the cosine similarity method to calculate the semantic similarity between the co-authored paper subject keyword word vector and the author academic keyword word vector, and the obtained similarity calculation result is the semantic similarity between several subject keywords of the co-authored paper to be analyzed and several academic keywords of the author.

[0033] The beneficial effect of the above technical solution is: using the semantic similarity calculation method to evaluate the correlation between the research field of the topic of the co-authored paper to be analyzed and the academic research field of the author, this method is simple and effective.

[0034] Furthermore, in step 3), the means used to extract several academic keywords of the author are: performing word segmentation processing on the author's published papers; extracting subject keywords based on the word segmentation processing results to generate a subject keyword matrix; extracting paper keywords based on the word segmentation processing results to obtain paper keyword word vectors; calculating the semantic similarity between the subject keywords and all paper keywords under the same topic, obtaining the similarity ranking results between the subject keywords and paper keywords under each topic, and selecting several subject keywords and paper keywords with higher similarity as the author's academic keywords.

[0035] Furthermore, the LDA topic model is used to extract topic keywords, and the Word2Vec word vector model is used to extract paper keywords.

[0036] Furthermore, in step 2), the means for extracting several subject keywords of the co-authored paper to be analyzed is: performing word segmentation on the co-authored paper to be analyzed, and counting the attribute information of the words after word segmentation; inputting the attribute information of the words after word segmentation into the word classification model to obtain the classification result; using the classification result to determine the weight of each word, and using it as the initial weight of the TextRank algorithm, and using the TextRank algorithm to process to screen out several subject keywords of the co-authored paper to be analyzed.

[0037] Furthermore, the word classification model is a support vector machine model.

[0038] Furthermore, the attribute information includes at least one of semantics, word frequency, part of speech, length, number of documents in which the word appears, and position in the document.

[0039] The beneficial effect of the above technical solution is: various types of attribute information are selected to cluster and classify words, making the classification effect more accurate, thereby ensuring the accuracy of the author's academic keyword screening. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is an overall flow chart of the method for evaluating and analyzing the contribution of co-authored papers based on the authors' academic backgrounds of the present invention;

[0041] Figure 2 This is a flowchart of step 3 of the method for evaluating and analyzing the contribution of co-authored papers based on the authors' academic backgrounds of the present invention;

[0042] Figure 3 This is a flowchart of step 4 of the method for evaluating and analyzing the contribution of authors of co-authored papers based on the authors' academic backgrounds of the present invention. DETAILED DESCRIPTION

[0043] The present invention takes into account the academic background of the co-authored paper author to analyze the actual contribution of the co-authored paper author to the co-authored paper. The following is a detailed description of a co-authored paper author contribution evaluation and analysis method based on the author's academic background in conjunction with the accompanying drawings and embodiments.

[0044] Method Example:

[0045] This embodiment takes a co-authored paper in the field of tobacco as an example to specifically introduce the co-authored paper author contribution evaluation and analysis method based on the author's academic background of the present invention. The overall process is as follows: Figure 1 As shown, the process is as follows:

[0046] Step 1: Obtain the literature information of papers in the tobacco field and analyze the basic information such as paper title, author name, author signature order, unit name, paper keywords, subject classification number, etc. Assume that the authors in the paper to be analyzed are A, B, C, D, and E in signature order.

[0047] Step 2: Mark the authors in the order of their signatures, and identify the first author and corresponding author, A and E respectively.

[0048] Step 3: For the co-authored paper to be analyzed, the TextRank keyword extraction algorithm based on part-of-speech features is used, and the integrity constraint is used to screen the relevant keyword word set of the research topic of the co-authored paper to be analyzed (i.e. the theme keywords of the co-authored paper) to determine the research field of the co-authored paper topic. Figure 2 As shown, the specific implementation method is as follows:

[0049] 1) Download the required co-authored paper data from the literature database, pre-process the original paper data, and perform part-of-speech tagging. Combined with the domain dictionary, perform text segmentation and stop word removal to obtain the segmentation processing result.

[0050] 2) Statistical attribute information of each word after word segmentation, mainly including the word's semantics, frequency, part of speech, length, number of documents in which it appears, and position.

[0051] 3) Input the attribute information of each word into the support vector machine model to obtain the classification result. The weights assigned to the words in a classification are consistent, and the assigned weights are used as the initial weights of TextRank. The TextRank algorithm is used for processing to obtain the ranking of all words.

[0052] 4) Sort the candidate words in descending order, and screen several topic keywords of the co-authored papers to be analyzed based on a given threshold and using integrity constraints. The formed topic keyword set is Ω = {crops, pests and diseases, image recognition, deep learning, rice, smart agriculture, target detection, Android, data screening, feature selection}.

[0053] Step 4: Obtain the published papers of all authors of the co-authored papers to be analyzed, and use an algorithm that combines the LDA topic model with the keyword extraction algorithm of the Word2Vec word vector model to obtain the author's subject keywords and paper keywords from the text abstract, citation, and full text, and then obtain the academic keywords of each author to determine the academic research field of each author. For co-authors A, B, C, D, E, Figure 3 As shown, the specific implementation method is as follows:

[0054] 1) Download the required paper data from the literature database (i.e., all published papers of the author. If some papers are not available, it is also acceptable to obtain only some of the published papers). Preprocess the original data, perform part-of-speech tagging, and perform text segmentation and stop word removal in combination with the domain dictionary.

[0055] 2) Use the LDA topic model to preliminarily extract the topics of the data set and generate a topic-word (m×n) matrix. Here, each topic corresponds to several topic keywords.

[0056] 3) According to the word segmentation results obtained in step 1), the Word2vec word vector model is used for training and learning to generate a word vector model, thereby obtaining the keyword word vector of the author's paper. The Word2Vec word vector model is a simplified neural network that uses one-hot to encode sentences, represents each word that appears in the sentence with a vector of a specified length, converts the keywords used into word feature vectors at the semantic level, and then calculates the semantic similarity between the keywords. In this embodiment, the training parameters of the Word2Vec embedding model are: the most similar word dimension topNSize=40, the context window size parameter Window=5, the configuration threshold of the random downsampling of high-frequency words is 1e-3, and the CBOW algorithm model is used to obtain the keyword vector and the Softmax method is used for optimization.

[0057] 4) Based on the word vector model generated in step 3), use the cosine similarity method to calculate the similarity between words under a certain topic in step 2) as the weight between the two points. Set a threshold to filter out word relationships with lower weights, build a keyword relationship network, iterate using the PageRank method, and finally output the TopN1 words with the highest PR value as the keywords under the topic. For other topics, use the same method to filter out the TopN2 words with the highest PR value as the keywords under the topic. Merge the keywords filtered out under each topic. TopN1 and TopN2 can be equal or different.

[0058] 5) After merging, a keyword network is formed, which is filtered to remove some interference or useless words, so as to finally obtain several academic keywords of the author.

[0059] In this embodiment, the academic keyword set of each author obtained according to the above method is:

[0060] A = {big data, neural network, deep learning, artificial intelligence, sensor, image recognition, smart agriculture, crops, pests and diseases, feature selection};

[0061] B = {enterprise management, management innovation, management model, human cost, employees, rules and regulations, assessment mechanism, performance management, organizational structure, resource allocation};

[0062] C = {smart agriculture, big data, crops, rice, tobacco bacterial wilt, image recognition, pesticides, pests and diseases, agricultural mechanization, chemical control};

[0063] D = {porous material, nanofiber, impedance matching, graphene, absorbing material, attenuation characteristics, structural design, multiple reflections, organic polymer material, curing reaction};

[0064] E = {big data, Internet of Things, data collection, data correlation analysis, machine learning, data set, system integration, data governance, digital twin, intelligent manufacturing}.

[0065] In this step, subject keywords and paper keywords may be repeated, and several academic keywords finally screened out may be subject keywords or paper keywords.

[0066] Step 5: Based on the author's academic keywords selected in step 4 and the subject keywords of the co-authored paper selected in step 3, construct the paper keyword vector and the subject keyword vector of the co-authored paper, construct a research field similarity evaluation and analysis model, and use the cosine similarity method to calculate the semantic similarity between the subject keyword vector of the co-authored paper and the author's academic keyword vector. The similarity calculation result cos(P,Q) is obtained, as shown in formula (1). The correlation between the author's academic research field and the research field of the co-authored paper theme can be obtained. The larger the value of cos(P,Q), the higher the similarity between the author's academic research field and the research field of the co-authored paper theme, which means that the author's academic participation in the co-authored paper is higher, that is, the greater the academic contribution of the author to the co-authored paper.

[0067]

[0068] In the formula, P = [p 1 ,p 2 ,…,p n ] represents the keyword vector of the co-authored paper, Q = [q 1 ,q 2 ,…,q n ] represents the author’s academic keyword vector.

[0069] According to the academic keyword set of each author in this embodiment, the contribution weights of N authors are obtained by using formula (1): ω A =0.732,ω B =0.126,ω C =0.622,ω D =0.186,ω E =0.536.

[0070] Step 6: Based on the order of authors’ signatures in the co-authored paper and the correlation between the author’s academic research field and the research field of the co-authored paper calculated in step 5, the contribution weight ω based on academic background is used to evaluate the author’s contribution to the co-authored paper to be analyzed. i , by combining the academic relationship information of the paper authors, the actual academic contribution of the paper co-authors is calculated and evaluated. Specifically:

[0071] 1) For a co-authored paper with N authors, the academic contribution weight algorithm based on the author signature order can obtain the signature order weight k of the i-th author. i , k i It is the normalized value, which includes the following three cases:

[0072] ① If the author is an author other than the first author and the corresponding author, the signature order weight k of the i-th author is ifor:

[0073]

[0074] In the formula, k i It represents the signature order weight determined according to the signature order of the i-th author in the co-authored paper to be analyzed, i ≥ 2. For example, when the author is in the second place in the entire signature order, i = 2, and when the author is in the fifth place in the entire signature order, i = 5.

[0075] ② If the author is the first author, the signature order weight k of the first author is 1 for:

[0076]

[0077] In the formula, α represents the additional weight of the first author, 0<α<1. As can be seen from the above formula, a weight α is added to the first author alone, considering that the first author has a greater contribution to the co-authored paper.

[0078] ③ If the author is the corresponding author, the corresponding author's position in the entire signature order is uncertain, it may be the second or the last, then the corresponding author's signature order weight k i for:

[0079]

[0080] In the formula, β represents the additional weight of the first author, 0<β<1. Similarly, giving the corresponding author an additional weight β also takes into account that the first author has made a greater contribution to the co-authored paper.

[0081] The additional weights α and β are obtained by iterating repeatedly based on the calculation of the algorithm model on the test data set and the order of the authors’ signatures in the co-authored papers to be analyzed. That is, the values ​​of α and β are constantly adjusted during the iteration process to obtain more appropriate values. This is done because sometimes the corresponding author is at the end of the entire signature order, and the signature order weight obtained based only on the signature order will be smaller, lower than the signature order weights of the third author, fourth author, and other authors who are ranked lower, but this is not reasonable, so a weight will be added to make the corresponding author’s final signature order weight higher, but this value must ensure that k after the weight is added iIt cannot be greater than 1, so the value needs to be set reasonably. The final values ​​of α and β may be the same or different, generally α>β. According to the algorithm iteration calculation, α=0.535 and β=0.465 are generally taken. Of course, if an author is both the corresponding author and the first author, the weight added to the author can be the larger value of α and β, or a value larger than α can be added, all of which can be assigned according to the actual situation. However, the main idea of ​​this treatment method is to consider that the first author and the corresponding author have a greater contribution to the co-authored paper, and add a weight to them, for example, set it to α=0.615 and β=0.385.

[0082] In this embodiment, α=0.535, β=0.465 are set and substituted into formulas (3) and (4), and the author signature order weights can be obtained as follows: A =0.973, k B =0.219, k C =0.146, k D =0.110, k E =0.553.

[0083] 2) The calculated semantic similarity between several subject keywords of the co-authored paper to be analyzed and several academic keywords of the author is used as the contribution weight ω based on the academic background to evaluate the author's contribution to the co-authored paper to be analyzed i ,ω i The same is the normalized value. The contribution weight of N authors is W = {ω A ,ω B ,ω C ,ω D ,ω E}.

[0084] 3) Based on the author's signature order weight and contribution weight, the academic contribution of the i-th author is calculated:

[0085] λ i =k i *ω i

[0086] Combining the above data, the academic contribution of each co-authored paper author obtained in this embodiment is: A =0.712,λ B =0.028,λ C =0.091,λ D =0.020,λ E =0.296.

[0087] 4) According to λ i Evaluate each author's contribution to the co-authored paper.

[0088] In summary, the method of the present invention comprehensively considers the influence of the author's research field, the author's academic keywords and other academic backgrounds on the degree of participation in co-authored papers, and makes up for the shortcomings of the existing technology in the inaccurate assessment of the actual academic contribution of paper co-authors. By extracting and analyzing the academic background of the participating authors, the actual academic contribution of paper co-authors is calculated and evaluated in a way of comprehensive weighting of academic related information in combination with the number of authors of co-authored papers and the order of signatures, which improves the comprehensiveness and accuracy of the calculation and allocation of the actual academic contribution of paper co-authors in the tobacco field. Moreover, it can also be used to evaluate the true academic ability of tobacco researchers, and thus serve the fair and objective scientific research performance evaluation of tobacco researchers and the construction of a database of scholars and experts in the tobacco field.

[0089] This embodiment is aimed at the co-authored paper in the field of tobacco. Correspondingly, most of the published papers of the co-authors are also papers in the field of tobacco. Of course, not all the published papers of the co-authors are necessarily papers in the field of tobacco. Because some technologies have technical overlaps, the co-authors may have some papers in other fields. All of them can be processed according to the method introduced in this embodiment. Of course, the co-authored paper can also be a paper in other fields.

[0090] In order to evaluate the academic contribution of co-authors of a paper, two parameters are needed: the signature order weight k determined by the signature order of the co-authors of the paper and i and the contribution weight ω determined by the correlation between the research field of the co-authored paper to be analyzed and the academic research field of the co-authors of the paper i These two parameters are related to the contribution of the co-authors to the co-authored papers analyzed. i The relationship is: k i The larger the i The larger the i The larger the i In order to show this relationship, in this embodiment, k i and ω i The product of is taken as the contribution of the co-authors to the co-authored paper i , that is, i =k i *ω i As another implementation, k i and ω i The contribution of the co-authors to the co-authored papers is i , or k i and ω i The weighted sum of the results is used as the contribution of the co-authors to the co-authored paper i, are both possible. The distribution of the weights of the two parameters can be considered according to the actual situation. For example, ω can be set i >k i Of course, other methods can be set, but what needs to be reflected is k i The larger the i The larger the i The larger the i The bigger.

[0091] Moreover, in this embodiment, considering that the first author and the corresponding author undertake more work in the process of co-authoring a paper, a weight is added to the first author and the corresponding author accordingly. As another implementation method, the first author and the corresponding author may not be given a weight, and all authors calculate the signature order weight k according to formula (2) i That's it.

[0092] In this embodiment, more machine learning models are used, including the support vector machine model in step 3, the LDA topic model and the Word2Vec word vector model in step 4, etc., all of which need to be trained. Since the specific training process and method are prior art, no detailed introduction is given in this embodiment. Of course, in order to achieve the corresponding purpose, it is not limited to using these models, and other models in the prior art that can achieve the corresponding purpose can be used. For example, the role of the support vector machine model is clustering. In order to achieve the purpose of clustering, other clustering models in the prior art can be used to achieve this purpose.

Claims

1. A method for evaluating and analyzing the contribution of co-authored papers based on the authors' academic background, characterized in that: The steps include: 1) Obtain the co-authored papers to be analyzed and identify the authors of the co-authored papers to be analyzed; 2) Extract several subject keywords of the co-authored papers to be analyzed to determine the research field of the co-authored papers to be analyzed; 3) For a certain author of the co-authored paper to be analyzed, obtain the published papers and extract several academic keywords of the author from them to determine the academic research field of the author; 4) Calculate the semantic similarity between several subject keywords of the co-authored paper to be analyzed and several academic keywords of the author to determine the correlation between the research field of the co-authored paper to be analyzed and the academic research field of the author, and use the determined similarity value as the contribution weight based on the academic background to evaluate the author's contribution to the co-authored paper to be analyzed i ; 5) Determine the contribution of the author to the co-authored paper to be analyzed based on the signature order weight determined by the signature order of the author in the co-authored paper to be analyzed and the contribution weight of the author; wherein, the greater the contribution weight, the greater the contribution; the greater the signature order weight, the greater the contribution.

2. The method for evaluating and analyzing the contribution of co-authored papers based on the authors' academic background according to claim 1 is characterized in that: In step 5), the contribution of a certain author to the co-authored paper to be analyzed is calculated using the following formula: l i =k i *oh i In the formula, λ i represents the contribution of the i-th author determined in the order of signature; k i Represents the signature order weight determined according to the signature order of the i-th author in the co-authored paper to be analyzed.

3. The method for evaluating and analyzing the contribution of co-authored papers based on the authors' academic background according to claim 2 is characterized in that: The earlier the author's signature order is, the greater the weight of the author's signature order is.

4. The method for evaluating and analyzing the contribution of co-authored papers based on the authors' academic background according to claim 2 is characterized in that: If the author is not the first author or the corresponding author, the signature order weight k of the i-th author is i for: Where N represents the total number of all authors of the co-authored papers to be analyzed, i≥2; If the author is the first author, the signature order weight k1 of the first author is: In the formula, α represents the additional weight of the first author, 0<α<1; If the author is the corresponding author, the corresponding author's signature order weight k i for: Where β represents the additional weight of the first author, 0<β<1.

5. The method for evaluating and analyzing the contribution of co-authored papers based on the authors' academic background according to claim 1 is characterized in that: In step 4), the means used to calculate the semantic similarity between several subject keywords of the co-authored paper to be analyzed and several academic keywords of the author are: Based on several subject keywords of the co-authored papers to be analyzed, construct the subject keyword word vector of the co-authored papers; Based on the author's academic keywords, construct the author's academic keyword word vector; The cosine similarity method is used to calculate the semantic similarity between the word vectors of the subject keywords of the co-authored paper and the word vectors of the author's academic keywords. The similarity calculation result is the semantic similarity between several subject keywords of the co-authored paper to be analyzed and several academic keywords of the author.

6. The method for evaluating and analyzing the contribution of co-authored papers based on the authors' academic background according to claim 1, characterized in that: In step 3), the method used to extract several academic keywords of the author is: Perform word segmentation on the author's published papers; According to the word segmentation processing results, the topic keywords are extracted to generate a topic keyword matrix; According to the word segmentation results, the paper keywords are extracted to obtain the paper keyword word vector; Calculate the semantic similarity between the subject keywords and all paper keywords under the same topic, obtain the similarity ranking results between the subject keywords and paper keywords under each topic, and select several subject keywords and paper keywords with higher similarity as the author's academic keywords.

7. The method for evaluating and analyzing the contribution of co-authored papers based on the authors' academic background according to claim 6 is characterized in that: The LDA topic model is used to extract topic keywords, and the Word2Vec word vector model is used to extract paper keywords.

8. The method for evaluating and analyzing the contribution of co-authored papers based on the authors' academic background according to claim 1 is characterized in that: In step 2), the means of extracting several subject keywords of the co-authored papers to be analyzed is: Perform word segmentation on the co-authored papers to be analyzed and count the attribute information of the words after word segmentation; Input the attribute information of the words after word segmentation into the word classification model to obtain the classification result; The classification results are used to determine the weight of each word and used as the initial weight of the TextRank algorithm. The TextRank algorithm is used to screen out several subject keywords of the co-authored papers to be analyzed.

9. The method for evaluating and analyzing the contribution of co-authored papers based on the authors' academic background according to claim 8, characterized in that: The word classification model is a support vector machine model.

10. The method for evaluating and analyzing the contribution of co-authored papers based on the authors' academic background according to claim 8, characterized in that: The attribute information includes at least one of semantics, word frequency, part of speech, length, number of documents in which it appears, and position in a document.

Images