1. bookAHEAD OF PRINT
Detalles de la revista
License
Formato
Revista
eISSN
2444-8656
Primera edición
01 Jan 2016
Calendario de la edición
2 veces al año
Idiomas
Inglés
Acceso abierto

Intelligent Matching System of Clauses in International Investment Arbitration Cases Based on Big Data Statistical Model

Publicado en línea: 15 Jul 2022
Volumen & Edición: AHEAD OF PRINT
Páginas: -
Recibido: 23 Apr 2022
Aceptado: 29 Jun 2022
Detalles de la revista
License
Formato
Revista
eISSN
2444-8656
Primera edición
01 Jan 2016
Calendario de la edición
2 veces al año
Idiomas
Inglés
Introduction

International investment arbitration has the advantages of flexibility, autonomy, efficiency, neutrality, and other advantages in various methods of resolving investor and host country treaty disputes. The acceptance of international arbitration to resolve investment disputes in bilateral investment treaties concluded by China. has dramatically increased, which seems to help Chinese enterprises to go global and bring in foreign investment. Many arbitration institutions exist in mainland China, but none have ever accepted arbitration between foreign investors and host countries. This is due to legislative reasons and international cooperation at the government level. The rapid spread of information in the Internet age has led to the emergence of news. At the same time, discussions on case news have also been launched on Weibo, WeChat, Zhihu, and other major social platforms [1]. Although news about legal cases emerges endlessly in new media, such news information lacks professional legal interpretation. Therefore, more and more scholars have paid attention to the research necessity of information search in the legal field.

On October 1, 2016, the “Regulations of the Supreme People's Court on People's Courts Publishing Judgments on the Internet” came into effect. As of August 30, 2020, the relevant search websites for judgment documents include China Judgment Documents Network, Peking University Magic Treasure Website, etc. All central-local courts have corresponding official websites for document inquiries. On the China Judgment Documents website alone, the total number of uploaded judgment documents has exceeded 100 million. Since then, this kind of judgment document retrieval system has provided users with many legal research cases. This provides data protection for basic legal case searches. Judgment documents record the process and results of the trial of the people's court, and it is the carrier of the results of the litigation activities. The content includes the public prosecution, the defendant, the plaintiff, the defender, the judiciary, the evidence information, the court's judgment basis, and the referenced laws and regulations. The practical value of adjudication documents is reflected in the fact that the adjudication documents have been reviewed many times, the words are used carefully, the language is well standardized, and the case explanation is detailed. It can provide non-professional users with customized services for pushing related cases. Users can obtain relevant judgment documents according to unique case conditions.

In this paper, news-like factual texts are used as queries. In this paper, the structured and standardized judgment documents are used as the full-text corpus. This paper generalizes the professional retrieval system by using the structure and content features of the judgment documents [2]. This meets the retrieval needs of non-professional users who lack legal knowledge. This paper proposes the “news-documents” automatic recommendation system framework. The framework can be implemented to recommend relevant judgment documents for news-like authentic texts.

Related Models

This paper integrates the structure information and content information of judgment documents into the information retrieval system of international investment judgment documents. This paper uses the BM25 model to calculate the similarity between feature words and documents [3]. And use the BM25 values of different feature words in factual texts as multi-dimensional features in the similarity algorithm. The overall similarity between the factual text and the international investment arbitration instrument is obtained.

The BM25 model is an algorithm for evaluating the relevance between search terms and documents. Its formula expression is simBM25(dj,q)~ki(dj,q)Bi,j×lg(Nni+0.5ni+0.5) {sim}_{BM\,25}\left({{d_j},\,q} \right)\sim\sum\limits_{{k_i}\left({{d_j},q} \right)} {{B_{i,j}} \times \lg \left({{{N - {n_i} + 0.5} \over {{n_i} + 0.5}}} \right)}

Where Bi,j=(K1+1)fi,jK1[(1b)+b×len(dj)avgdoclen]+fi,j {B_{i,j}} = {{\left({{K_1} + 1} \right){f_{i,j}}} \over {{K_1}\left[{\left({1 - b} \right) + b \times {{len\left({{d_j}} \right)} \over {{avg}_{doclen}}}} \right] + {f_{i,j}}}} ; K1 and b are empirically determined constants. fi,j represents the frequency of index item i in document j. It refers specifically to feature words in this paper. Feature words in factual texts contain a lot of semantic information. This paper extracts feature words in factual texts to calculate the correlation between feature words and international investment arbitration documents. This achieves the purpose of assessing the overall relevance between the factual text and the international investment arbitration instrument.

This paper uses the SVM Rank and Lambda MART algorithms to fit the BM25 values of multi-dimensional feature words [4]. In this way, the overall similarity calculation between news corpus and international investment arbitration documents is realized.

Based on SVM, some scholars have proposed a ranking learning algorithm, SVM Rank. They transformed the ranking problem into a binary classification problem. The basic idea is to give a dataset {xi, yi}. where yi ∈ {1, ⋯, R}. There exists a function h(x) that satisfies h(xi) > h(xj) ⇔ yi > yj. Therefore, we set the relevant international investment arbitration document set of factual text as {xi, yi}. The relevant international investment arbitration document association pair {xi, yi} and its correlation label y constitute training data xi, yi meaning that the international investment arbitration document and the factual text document are opposite relationships. Assuming m = |ρ|, the optimization problem of SVM Rank can be transformed into the mathematical form: minw,ζi,j012wTw+Cm(i,j)ρξi,j \mathop {\min\,}\limits_{w,\,{\zeta _{i,j}} \ge 0} {1 \over 2}{w^T}w + {C \over m}\sum\limits_{\left({i,j} \right) \in \rho} {{\xi _{i,j}}} s.t.(i,j)ρ:(wTxi)(wTxi)+1ξi,j s.t.\forall \left({i,j} \right) \in \rho :\left({{w^T}\,{x_i}} \right) \ge \left({{w^T}{x_i}} \right) + 1 - {\xi _{i,j}}

We find a linear function h(x) such that the training corpus has a corresponding order. This is an orderly return. The algorithm can be integrated into the multi-dimensional vector represented by the BM25 value of the news corpus feature word. It effectively improves the calculation effect of document similarity [5]. Therefore, we can calculate the overall similarity between the feature words in the news corpus and the international investment arbitration documents.

The Lambda MART algorithm has achieved good results in information retrieval. Its algorithmic nature can be widely used in sorting tasks. The content includes advertisement recommendations, automatic scoring, etc. The basic idea of the RankNet algorithm is to provide a scoring function si = f(xi). xi represents the feature word vector extracted from the factual text. Then we calculate the probability value that IIA document i ranks ahead of IIA document j. Its calculation formula is P(UiUj)=11+eσ(sisj) P\left({{U_i} \triangleright {U_j}} \right) = {1 \over {1 + {e^{- \sigma \left({{s_i} - {s_j}} \right)}}}} . Its loss function is C=Pi,j¯lg(Pi,j)(1Pi,j¯)lg(1Pi,j) C = - \overline {{P_{i,j}}} \lg \left({{P_{i,j}}} \right) - \left({1 - \overline {{P_{i,j}}}} \right)\lg \left({1 - {P_{i,j}}} \right) . Where Pi,j¯ \overline {{P_{i,j}}} refers to the actual probability that the International Investment Arbitration Instrument Document i precedes the International Investment Arbitration Instrument Document j. Pi,j is the predicted probability that the International Investment Arbitration Instrument Document i precedes the International Investment Arbitration Instrument Document j.

Framework for an automatic recommendation of international investment arbitration documents based on structural content characteristics
Feature Words of Legal Text Corpus

The normative texts of international investment arbitration instruments have solid qualitative expressions, such as the charges' description, the case's characterization, etc. However, the news corpus has a relatively strong event statement nature, and international investment arbitration documents only have event statements contained in the structure of the trial process [6]. Therefore, this paper explores the idea of extracting keywords from such structural content to enhance the calculation effect of text similarity.

At present, the more common feature word extraction algorithms are the TF-IDF algorithm, mutual information, and information get on. TF-IDF is a classic text keyword extraction algorithm. It mainly calculates the relative weight of the feature word to the document from two aspects: the number of times the feature word appears in all documents and the number of times the feature word appears in this document. The main idea is to weaken the influence of high-frequency words and stop words in literature. Information gain is obtained by calculating whether the feature word t is the probability that an article appears in the category c. Mutual information is obtained by calculating the amount of information that feature word t can provide for the category c. In this paper, TF-IDF is selected as the feature word selection algorithm. We extract feature words highly correlated with crime types from international investment arbitration documents. On this basis, we use feature words to represent factual texts semantically. We calculate the BM25 value separately between the feature word and the international investment arbitration instrument. We calculate the overall similarity between the international investment arbitration instrument and the factual text.

“News-International Investment Arbitration Documents” Recommendation System Framework

The recommendation system framework of this paper is mainly divided into structured index construction of international investment arbitration documents and feature extraction of news corpus texts. The primary purpose of constructing the structure index of international investment arbitration documents is to realize the structuring of the text of international investment arbitration documents. It indexes different structural information of international investment arbitration instruments [7]. The primary purpose of news corpus text feature extraction is to represent the news corpus semantically. The algorithm extracts feature words to enhance the retrieval effect of long texts. Finally, we perform iterative learning according to the similarity ranking algorithm and output the relevant international investment arbitration document set. The specific framework is shown in Figure 1.

Figure 1

“News-International Investment Arbitration Documents” recommendation system framework

This paper uses the corresponding text similarity algorithm, Lambda MART algorithm, and Svm Rank algorithm to calculate the semantic similarity between news corpus and international investment arbitration documents. Users can sort the recommended referee documents according to their algorithm in descending order of relevance. According to the needs, this article refers to and acquires knowledge of highly relevant international investment arbitration documents [8]. The content includes, but is not limited to, lawyer recommendations, references to legal provisions, and evidence fixation. It meets the retrieval needs of obtaining relevant legal information using factual text content like news.

Experiment Design and Implementation
Establishment of experimental corpus

The corpus used in this paper is the structured corpus provided by open law. Cn as the full-text corpus of legal documents. It is used to provide a candidate set of relevant legal instruments. The corpus is provided by open law. Cn not only contains the corresponding crime of each case but also provides information on the structure of legal documents. This experiment focuses on the criminal case literature. The logic of this type of legal documentation is relatively straightforward [9]. Among them, there are 7320 criminal case documents. The statistics on the distribution characteristics of the types of international investment arbitration document cases are shown in Table 1.

Statistics of crime types in criminal cases

Type of criminal offense Frequency
theft 1854
Crime of smuggling, selling, transporting, and manufacturing drugs 904
Intentional injury 694
Fraud 544
Dangerous Driving Offense 454
Crime of picking quarrels and provoking trouble 433
Offenses of illegal manufacture, sale, transportation, mailing, storage of firearms, ammunition, explosives 419
Robbery 411
Crime of opening a casino 336
Credit card fraud 193

From Table 1, it can be seen that the crime type of theft is significantly higher than other crimes. Starting with the robbery at No. 8, there is a clear downward trend in the number of paperwork for other crime types. And according to statistics, there are 175 types of crimes involved in international investment arbitration documents [10]. We can find that the categories of crimes and punishments in international investment arbitration documents are widely distributed, and the traditional classification retrieval system increases the difficulty of information retrieval for users.

In addition, the results of the statistics of the types of crimes involved in the cases recorded in the international investment arbitration documents are shown in Table 2.

Number of offenses covered by legal instruments

The number of crimes involved in the case Frequency Distribution percentage/%
1 5512 75.57
2 1355 18.51
3 219 2.99
4 75 1.04
5 35 0.48
5 12 0.15
7 5 0.07
≥8 5 0.08

It can be seen from Table 2 that the number of crimes involved in legal documents is distributed in steps. The number of documents involving one count accounted for 75.57%, and the number of documents involving two counts was 18.51%. The number of documents with no less than three charges is about 5%. The nature of international investment arbitration cases is mainly based on a single crime, but about 25% of international investment arbitration documents still contain two or more crimes. Therefore, the scope of sentencing involved in international investment arbitration instruments and the legal provisions vary significantly. It's unique [11]. According to the crime classification and indexing system, this method is not conducive to the indexing and retrieval of traditional search engines. This statistical result is in line with the content characteristics of the international investment arbitration documents summarized above and also confirms the importance of this experimental study from the side.

The similarity score uses the crime type of the news corpus and the crime type of the legal document for matching evaluation. The main algorithm is Hamming distance. The relevance of other corpora defaults to irrelevant literature.

Evaluation of experimental results
Evaluation indicators of the experiment

This evaluation experiment is to improve the retrieval effectiveness of the retrieval system. We use NDCG to evaluate this experiment. NDCG is gradually improved according to the evaluation methods of cumulative gain (CG) and impaired cumulative gain (DCG). The cumulative gain method is the sum of the correlations at the specified location. The CG calculation formula for the specified position p is: CGp=i=1preli {CG}_p = \sum\limits_{i = 1}^p {{rel}_i}

Where rel represents the relevance of the document at the position i. The impairment cumulative gain method adds the ranking information of the retrieval results to the evaluation of the retrieval results. The DCG calculation formula at its position p is DCGP=rel1+i=2prelilog2(i+1) {DCG}_P = {rel}_1 + \sum\limits_{i = 2}^p {{{{rel}_i} \over {{{\log}_2}\left({i + 1} \right)}}}

The evaluation index NDCG compares the predicted results with the ideal predicted results. We normalize the predicted retrieval results. The evaluation formula for the first p search results is: NDCGp=DCGpIDCGP {NDCG}_p = {{{DCG}_p} \over {{IDCG}_P}}

IDCG represents the ideal retrieval result, and we sort related documents in descending order according to their relevance.

NDCG can well reflect the difference between the similarity calculated by the model and the ideal similarity. The larger the NDCG value, the better the effect of model similarity estimation. In this paper, NDCG(1) and NDCG(5) will be selected to reflect the ranking of the most relevant documents in the system. NDCG(10) and NDCG(20) indicate the recommendation of related documents when the system returns more documents. The evaluation result of this experiment is to take the average value of multiple query NDCGs in the test set as the final evaluation index.

Comparison of experimental results

This experiment firstly uses the BM25 algorithm to verify the effect of the feature improvement algorithm. A bag-of-words model represents the news corpus. We use the known international investment arbitration documents to index the news corpus to extract relevant keywords—search based on indexing results. The experimental results are shown in Table 3.

The performance of feature words in the BM25 algorithm

Experimental results
NDCG(1) NDCG(5) NDCG(10) NDCG(20)
Don't use features 0 0.024 0.023 0.03
TF-IDF 0.1 0.142 0.151 0.166

Our experimental results without features are worse than the text representation after feature words are extracted with TF-IDF. The text of the BM25 algorithm news corpus is long, and there are too many potential query words. This increases the difficulty of calculating the similarity of the algorithm. Lots of meaningless words in the text. This does not contribute to the similarity calculation. Therefore, we select non-low-frequency words with rich connotations after using the feature word extraction algorithm to construct the factual text's query formula. We will use the TF-IDF algorithm to extract text keywords in subsequent experiments. At the same time, we index the text to reduce the computational complexity and improve the recommendation effect of the model. At the same time, this paper also incorporates the text structure of international investment arbitration documents into considering the text matching model and designs experiments. The results are shown in Table 4.

Model results on different structures

Text structure Model Experimental results
NDCG(1) NDCG(5) NDCG(10) NDCG(20)
Court Opinion Section BM25 0.1 0.142 0.151 0.166
Svm Rank 0.125 0.169 0.19 0.199
LTR lambda MART 0.366 0.305 0.269 0.24
Part of the trial process BM25 0.269 0.196 0.16 0.161
Svm Rank 0.114 0.196 0.226 0.249
LTR lambda MART 0.3 0.294 0.299 0.294
Full text BM25 0.211 0.21 0.169 0.169
Svm Rank 0.114 0.196 0.223 0.249
LTR lambda MART 0.266 0.232 0.255 0.254

From the performance results of different text matching models, the BM25 model has the worst performance, and the SVM Rank model and the Lambda MART model have improved to varying degrees. The retrieval performance of the BM25 model is higher than that of the SVM Rank model only in the evaluations of NDCG(1) and NDCG(5) when using the adjudication process and the structure-content features full-text results. In the evaluation of NDCG (10) and NDCG (20), the effect of the SVM Rank is significantly higher than that of BM25. Analysis of the reason may be that when the content of international investment arbitration documents increases, we use the method of keyword matching to facilitate the correlation of words in documents. This puts the most relevant documents at the top. The SVM Rank model is to find an optimal bound for ordinal regression in global data. The model ranks highly relevant documents ahead of less relevant documents as much as possible. Therefore, its NDCG performance results in NDCG(10) and NDCG(20) are still better than the BM25 algorithm. The Lambda MART model outperforms other retrieval models under different text structures.

The LambdaMART model performs better in evaluating NDCG(1) and NDCG(5) using only the court opinion structure content feature. The best performance in evaluating NDCG (10) and NDCG (20) is to use only the trial process structure content feature. When we combine the two, the model's performance decreases to a certain extent. The reason may be that the text of the trial process contains a lot of multi-angle statements and unclear facts. To a certain extent, the court did not recognize its content, which caused the deviation of text similarity calculation. But that content is suitable for expanding potentially related documents. Therefore, the relevance ranking of the first few results returned by the Lambda MART model is more relevant when only the structural content features of court opinions are used. The model can return more relevant text when we use the structure content feature of the trial process. When using the full-text matching method, the model effect tends to be average. This is not conducive to calculating the similarity between factual texts and international investment arbitration instruments. The BM25 algorithm can effectively improve the query effect by using the structure content features of the trial process after the feature extraction of the query formula. The possible reason is that the BM25 algorithm can obtain a more significant matching probability after using the structure content feature of the trial process. The SVM Rank algorithm has similar performance fluctuations to the Lambda MART model under different structural content characteristics.

Therefore, the Lambda MART model uses the feature word extraction algorithm to construct the document relevance matrix. Its text structure features can effectively improve the retrieval effect of only using full-text content. This experiment uses a limited corpus dataset. Given the large number of datasets currently formed by the International Investment Arbitration Documents Network, we will get better results by using the court opinions to do text recommendation work.

Conclusion

This paper proposes an automatic recommendation framework for international investment arbitration documents based on the characteristics of structure and content by using the characteristic words and structural content characteristics of international investment arbitration documents. The algorithm improves the shortcomings of the traditional full-text retrieval model BM25 in retrieving factual texts using a news-like corpus. On this basis, this paper uses the SVM Rank algorithm and the Lambda MART algorithm. This algorithm improves the effect of information retrieval based on factual texts of news-like corpora. This better fulfills the legal information retrieval needs of non-professional users.

Figure 1

“News-International Investment Arbitration Documents” recommendation system framework
“News-International Investment Arbitration Documents” recommendation system framework

Model results on different structures

Text structure Model Experimental results
NDCG(1) NDCG(5) NDCG(10) NDCG(20)
Court Opinion Section BM25 0.1 0.142 0.151 0.166
Svm Rank 0.125 0.169 0.19 0.199
LTR lambda MART 0.366 0.305 0.269 0.24
Part of the trial process BM25 0.269 0.196 0.16 0.161
Svm Rank 0.114 0.196 0.226 0.249
LTR lambda MART 0.3 0.294 0.299 0.294
Full text BM25 0.211 0.21 0.169 0.169
Svm Rank 0.114 0.196 0.223 0.249
LTR lambda MART 0.266 0.232 0.255 0.254

Statistics of crime types in criminal cases

Type of criminal offense Frequency
theft 1854
Crime of smuggling, selling, transporting, and manufacturing drugs 904
Intentional injury 694
Fraud 544
Dangerous Driving Offense 454
Crime of picking quarrels and provoking trouble 433
Offenses of illegal manufacture, sale, transportation, mailing, storage of firearms, ammunition, explosives 419
Robbery 411
Crime of opening a casino 336
Credit card fraud 193

The performance of feature words in the BM25 algorithm

Experimental results
NDCG(1) NDCG(5) NDCG(10) NDCG(20)
Don't use features 0 0.024 0.023 0.03
TF-IDF 0.1 0.142 0.151 0.166

Number of offenses covered by legal instruments

The number of crimes involved in the case Frequency Distribution percentage/%
1 5512 75.57
2 1355 18.51
3 219 2.99
4 75 1.04
5 35 0.48
5 12 0.15
7 5 0.07
≥8 5 0.08

Yin, Q. Forced technology transfer performance requirement in international investment agreements—a Chinese perspective. Journal of Intellectual Property Law and Practice., 2022; 17(2): 114–131 YinQ. Forced technology transfer performance requirement in international investment agreements—a Chinese perspective Journal of Intellectual Property Law and Practice 2022 17 2 114 131 10.1093/jiplp/jpab176 Search in Google Scholar

Boue, J. C. Much More than a Footnote (or Three): Frank C. Hendryx and an Untold Story of Petroleum Concessions and the Genesis of ICSID. Journal of International Dispute Settlement., 2022; 13(1): 1–40 BoueJ. C. Much More than a Footnote (or Three): Frank C. Hendryx and an Untold Story of Petroleum Concessions and the Genesis of ICSID Journal of International Dispute Settlement 2022 13 1 1 40 10.1093/jnlids/idab031 Search in Google Scholar

Fei, L. Innovation and Development of the China International Commercial Court. The Chinese Journal of Comparative Law., 2020; 8(1): 40–44 FeiL. Innovation and Development of the China International Commercial Court The Chinese Journal of Comparative Law 2020 8 1 40 44 10.1093/cjcl/cxaa014 Search in Google Scholar

Weiß, W., & Furculita, C. The EU in search for stronger enforcement rules: assessing the proposed amendments to trade enforcement regulation 654/2014. Journal of international economic law., 2020; 23(4): 865–884 WeißW. FurculitaC. The EU in search for stronger enforcement rules: assessing the proposed amendments to trade enforcement regulation 654/2014 Journal of international economic law 2020 23 4 865 884 10.1093/jiel/jgaa033 Search in Google Scholar

Treleaven, P., Barnett, J., Knight, A., & Serrano, W. Real estate data marketplace. AI and Ethics., 2021;1(4): 445–462 TreleavenP. BarnettJ. KnightA. SerranoW. Real estate data marketplace AI and Ethics 2021 1 4 445 462 10.1007/s43681-021-00053-4 Search in Google Scholar

Andrusiv, U., Kinash, I., Cherchata, A., Polyanska, A., Dzoba, O., Tarasova, T., & Lysak, H. Experience and prospects of innovation development venture capital financing. Management Science Letters., 2020; 10(4): 781–788 AndrusivU. KinashI. CherchataA. PolyanskaA. DzobaO. TarasovaT. LysakH. Experience and prospects of innovation development venture capital financing Management Science Letters 2020 10 4 781 788 10.5267/j.msl.2019.10.019 Search in Google Scholar

Schultz, U. Gender and potential impacts on decision-making in arbitration: experiences from Germany in an international field. International Journal of the Legal Profession., 2022; 29(1): 53–74 SchultzU. Gender and potential impacts on decision-making in arbitration: experiences from Germany in an international field International Journal of the Legal Profession 2022 29 1 53 74 10.1080/09695958.2021.1953502 Search in Google Scholar

Drummer, D., & Neumann, D. Is code law? Current legal and technical adoption issues and remedies for blockchain-enabled smart contracts. Journal of Information Technology., 2020; 35(4): 337–360 DrummerD. NeumannD. Is code law? Current legal and technical adoption issues and remedies for blockchain-enabled smart contracts Journal of Information Technology 2020 35 4 337 360 10.1177/0268396220924669 Search in Google Scholar

Parmentola, A., Petrillo, A., Tutore, I., & De Felice, F. Is blockchain able to enhance environmental sustainability? A systematic review and research agenda from the perspective of Sustainable Development Goals (SDGs). Business Strategy and the Environment., 2022;31(1): 194–217 ParmentolaA. PetrilloA. TutoreI. De FeliceF. Is blockchain able to enhance environmental sustainability? A systematic review and research agenda from the perspective of Sustainable Development Goals (SDGs) Business Strategy and the Environment 2022 31 1 194 217 10.1002/bse.2882 Search in Google Scholar

Aghili, A. Complete Solution For The Time Fractional Diffusion Problem With Mixed Boundary Conditions by Operational Method. Applied Mathematics and Nonlinear Sciences., 2021; 6(1): 9–20 AghiliA. Complete Solution For The Time Fractional Diffusion Problem With Mixed Boundary Conditions by Operational Method Applied Mathematics and Nonlinear Sciences 2021 6 1 9 20 10.2478/amns.2020.2.00002 Search in Google Scholar

Wang, Y. & Chen, Y. Evaluation Method of Traffic Safety Maintenance of High-Grade Highway. Applied Mathematics and Nonlinear Sciences., 2021;6(1): 65–80 WangY. ChenY. Evaluation Method of Traffic Safety Maintenance of High-Grade Highway Applied Mathematics and Nonlinear Sciences 2021 6 1 65 80 10.2478/amns.2021.1.00007 Search in Google Scholar

Artículos recomendados de Trend MD

Planifique su conferencia remota con Sciendo