[Abstract] With the rise of the field of learning sciences, deep learning has gradually become a core research topic in education. By employing citation analysis and co-word clustering analysis methods, this study investigates the research status and hotspots in the field of deep learning abroad over the past decade, based on 459 articles collected from the Web of Science database from 2005 to 2015. Through visual knowledge mapping, it aims to provide insights for similar research in China.
[Keywords] Deep learning; Citation analysis; Co-word clustering analysis; Knowledge mapping; Research hotspots
1. Introduction
The development of information technology has not only changed human production, thinking, and learning methods but also accelerated humanity’s transition into a learning society. With the popularity of new learning methods supported by technology, there is increasing attention on whether rich technologies and resources can significantly enhance learning outcomes, whether learners can integrate information to construct meaningful knowledge and apply it flexibly to solve practical problems. As the field of learning sciences has developed, deep learning has gradually emerged as an important and effective learning concept in this context, attracting widespread attention from researchers and learners.
The New Media Consortium’s “Horizon Report” (2015 K-12 Edition) also points out that the impact of deep learning strategies on classroom teaching is becoming increasingly profound, which is an important trend driving schools to apply educational technology [1]. Deep learning has also been translated as “deep-level learning”. American scholars Ference Marton and Roger Saljc first proposed this concept in their 1976 publication “The Essential Differences of Learning: Outcomes and Processes” [2]. Chinese scholar Li Jiahou believes that deep learning refers to the ability of learners to critically learn new ideas and facts based on understanding learning, integrating them into their existing cognitive structures, making connections among various ideas, and transferring existing knowledge to new contexts to make decisions and solve problems [3].
This article analyzes the temporal and spatial distribution of deep learning research abroad, core authors, and core journals based on citation analysis and co-word clustering analysis, and visually organizes the themes and hotspots in the deep learning research field to grasp the current status and development trends of deep learning research abroad, providing references for similar research in China.
2. Research Methods and Data Analysis Process
(1) Research Data Source
The data for this study comes from the WOS (Web of Science) database’s sub-database: Web of Science™ Core Collection. The Web of Science™ Core Collection has been rigorously selected and includes over 10,000 world-renowned, high-impact academic journals and more than 110,000 academic conference proceedings. Search item: topic, search term: Deep Learning, search period: 2005-2015, a total of 482 records were downloaded, and after deleting irrelevant and duplicate records, 459 valid records were obtained. Each record mainly includes full records of authors (Author), title (Title), abstract (Abstract), and citations (Descriptors and Identifiers).
(2) Research Methods and Tools
Citation analysis refers to the use of various mathematical and statistical methods, as well as logical methods such as comparison, induction, abstraction, and generalization, to analyze the citation and being cited phenomena of research objects such as scientific journals, papers, and monographs, in order to reveal their quantity distribution characteristics and inherent correlation laws [4]. This paper uses HistCite software for citation analysis. HistCite is a citation chronicle visualization analysis software developed by Eugene Garfield, a renowned American information scientist known as the “father of SCI,” and his colleagues [5]. Through citation analysis of literature related to Deep Learning in the Web of Science database, information on the temporal and spatial distribution, core authors, core journals, and core articles of Deep Learning-related literature from 2005 to 2015 is statistically obtained, thus providing a comprehensive understanding of the research status of deep learning abroad.
Co-word clustering analysis is a commonly used method in co-word analysis, which simplifies the complex co-word network relationships among numerous analysis objects into a relatively small number of clusters based on the frequency of co-occurrences of words, and visually represents this clustering process [6]. The tools used for co-word analysis in this paper are the Statistical Analysis Toolkit for Informetrics (SATI) 3.2, Excel, and SPSS 20.0. The SATI tool was developed by Liu Qiyuan, aiming to process the bibliographic information of journal full-text databases using general metric analysis, co-occurrence analysis, clustering analysis, multidimensional scaling analysis, social network analysis, and other data analysis methods to mine and present visual data results [7]. First, the frequency analysis function of SATI is used to extract the frequency of all keywords in the literature, and a co-word matrix is further extracted. SPSS 20.0 is then used to perform multidimensional scaling clustering analysis on the co-word matrix, thus forming a knowledge map of the deep learning field.
(3) Research Process
This paper mainly uses HistCite tools for citation analysis, SATI 3.2, Excel, and SPSS 20.0 software for co-word clustering analysis. The research process is divided into three parts: data collection, citation analysis, and co-word clustering analysis, as shown in Figure 1.
3. Research Results and Analysis
(1) Temporal and Spatial Distribution of Literature in the Field of Deep Learning Research
1. Temporal Distribution
From Figure 2, it can be seen that the number of international research on deep learning has shown an overall upward trend from 2005 to 2015. Before 2007, research on deep learning abroad was relatively scarce, and from 2007 to 2010, the number of studies increased slightly compared to before, but the annual growth rate was not significant, indicating that attention to deep learning research internationally was still low, and research progress was slow. In 2010, Sawyer (R.K. Sawyer) provided a more authoritative definition of learning sciences in the “Cambridge Handbook of Learning Sciences” [8]. Chinese scholar Jiao Jianli pointed out that the field of learning sciences aims to explore how deep learning occurs and to guide how to design deep learning, ultimately cultivating the skills necessary for students’ deep conceptual understanding [9]. With the continuous rise of international research in the field of learning sciences, it has also promoted the development of deep learning research to a certain extent. From Figure 2, it can be seen that in 2011, research on deep learning began to show significant growth, indicating that deep learning research has received widespread attention and that research results have been enriched year by year. In 2015, only the first eight months were counted, and the number of articles had already reached 102, indicating that international research on deep learning continues to deepen and maintain its heat.
2. Country (Region) Distribution
By statistically analyzing the countries (regions) of the literature, we can understand the attention and contributions of different countries to this research field. As shown in Figure 3, the top five countries in terms of total research literature are: the United States, China, Australia, the United Kingdom, and Canada. In addition to comparing the number (Recs), the Total Local Citation Score (TLCS) in HistCite citation analysis often better reflects the influence of literature in the research field.
From Figure 3, it can be seen that the top five countries in TLCS ranking are Canada, the United States, China, Australia, and Singapore. Among them, Canada, the United States, China, and Australia rank in the top five in both the number of publications and TLCS rankings, indicating that they not only have a large number of publications but also produce high-impact research results. Although Singapore has fewer research articles, its high TLCS ranking indicates that it has also conducted in-depth research on deep learning. The statistical results also show that China’s research on deep learning ranks highly in terms of both publication quantity and TLCS ranking, indicating that research in the field of deep learning has received widespread attention and has produced high-quality research results domestically.
(2) Author, Core Journals, and Core Literature Analysis in the Field of Deep Learning Research
1. Author Analysis
By analyzing the number of published papers and the frequency of citations of published papers, we can understand the contribution level of paper authors to a specific field of research. Authors with high contributions constitute the core author group in that field. Understanding and tracking the research themes and directions of these researchers can help us quickly grasp the research focus and development trends in a particular field [10]. From Table 1, it can be seen that the highest-ranked authors in terms of publication quantity are Professor Yoshua Bengio and Professor Geoffrey Hinton, who published six papers from 2010 to 2015. Through literature reading and data inquiry, it is found that Professors Yoshua Bengio, Geoffrey Hinton, and Yann LeCun are representative and influential researchers in the international deep learning field, with their latest research results published in the article titled “Deep Learning” in Nature magazine in May 2015 [11]. Among them, Professor Yoshua Bengio not only ranks first in publication quantity but also ranks first in citation frequency with 54 citations, indicating his important position in the field of deep learning.
2. Core Journal Analysis
This article determines the core journals in the field of deep learning based on the number of research literature and citation frequency. By continuously focusing on and studying the core journals, one can gain a comprehensive understanding of the research frontiers and trends in the field of deep learning. First, the top ten journals are identified based on TLCS citation frequency, which serves as the basic framework for core journals in the field of deep learning. Then, by comprehensively ranking the journals based on the number of published articles, the following ten journals are ultimately determined, with a total TLCS of 142, accounting for 71.5% of the total journal volume, as shown in Table 2. Among them, the top three journals by TLCS are the Journal of Machine Learning Research, IEEE Transactions on Pattern Analysis and Machine Intelligence, and Computers & Education, which also publish a large number of articles in the field of deep learning, indicating their important position in this field. The Journal of Machine Learning Research is published in the United States and mainly focuses on computer engineering technology and artificial intelligence. IEEE Transactions on Pattern Analysis and Machine Intelligence is an influential international academic journal in the field of artificial intelligence, also published in the United States. Computers & Education, published in the UK, primarily focuses on computer science and educational technology.
The above three journals mainly come from the field of computer science. In addition, deep learning has also received widespread attention in the fields of education and psychology, with the top ten journals including the American journals Contemporary Educational Psychology, Educational Psychology, Research in Higher Education, and the British journal British Journal of Educational Psychology. Deep learning has also received some attention in medical education, with a certain number of articles published in Advances in Health Sciences Education, Nurse Education Today, and Medical Teacher.
3. Core Literature Analysis
The frequency of citations of literature reflects the degree to which that literature has been noticed by other researchers and its influence on related research by other researchers [12]. Generally speaking, the higher the citation frequency of the literature, the greater its influence in the research field, and it can be identified as core literature. Reading core literature helps researchers to quickly and comprehensively understand the research focus and core connotations of the field, which is of great significance for further in-depth research. In this paper, the literature searched in HistCite software is arranged in descending order according to LCS, and the results are shown in Table 3.
(3) High-Frequency Words and Research Hotspots Analysis in Deep Learning
1. High-Frequency Keyword Analysis
In this study, the SATI tool for bibliographic information statistical analysis is mainly used to extract all keywords from the collected sample, manually merging synonyms and deleting irrelevant keywords. Then, SATI is used to count and rank the frequency of keywords. According to the Price formula, high-frequency keywords are determined based on a high-frequency citation threshold, with the calculation formula: M=0.749, where M is the high-frequency threshold, and Nmax represents the highest citation frequency of academic papers in the interval [14]. In the analysis of core journals mentioned above, it has been determined that the article with the highest citation frequency in the sample is a research paper by Erhan D and Bengio Y, which has been cited 20 times since its publication. Using the formula, this study selects 46 keywords with a frequency greater than 4, deletes duplicates and keywords with unclear meanings, and ultimately identifies the following 41 as high-frequency keywords in the field of deep learning, as detailed in Table 4.
From the content of Table 4, it can be seen that the top ten keywords in this research field, including deep learning, are: Deep learning (134), Education (51), Learning approaches (30), Assessment (15), Classification (12), Convolutional neural networks (11), Unsupervised learning (9), Neural networks (9), Learning strategies (9), Collaborative learning (8), Higher education (8). However, the frequency of keywords alone cannot reveal the internal connections among keywords, and thus cannot fully describe the research hotspots and trends in the deep learning field; co-word analysis is needed to further explore the deep connections among keywords.
2. High-Frequency Keyword Dissimilarity Matrix Analysis
In this article, the high-frequency keyword dissimilarity matrix is generated in the SATI software co-occurrence relationship matrix. Based on the 41 keywords identified above, the matrix row and column numbers are set to 41, and the dissimilarity matrix is obtained in Excel, resulting in a 41×41 dissimilarity matrix for deep learning keywords, as detailed in Table 5. The values in the dissimilarity matrix indicate the distance between the corresponding two keywords; the closer the value is to 1, the farther the keywords are from each other, indicating lower similarity. Conversely, the closer the value is to 0, the closer the keywords are to each other, indicating higher similarity [15].
As shown in part of the dissimilarity matrix in Table 5, the keywords that are increasingly closer to deep learning are: Learning approaches (0.9225), Education (0.9015), Assessment (0.8876), Unsupervised learning (0.8701), Classification (0.7496). From the complete keyword dissimilarity matrix, it can be seen that the keywords related to computer science technology, such as Neural networks and Support vector machine, have smaller values with Deep learning, while education-related keywords, such as Learning approaches and Student learning, have larger values with Deep learning, around 0.9. This indicates that compared to the education field, the computer science field has a higher degree of similarity with deep learning research, showing that research in computer science on deep learning is more extensive. The similarity values among keywords are relatively small, and some keywords are even unrelated, indicating that it is necessary to further explore through clustering analysis.
3. Keyword Clustering Diagram and Analysis
The keyword clustering results can further reflect the closeness relationships among keywords. In the clustering dendrogram, the closer the keywords are, the more similar they are. Through co-word clustering analysis, similar clusters of closely related thematic words can be formed, helping to initially clarify the research hotspots in the field of deep learning. According to research needs, the core keyword “Deep learning” is removed, and in SATI software, the row value is set to 42. The similarity matrix of deep learning keywords (42×42) is obtained in Excel. This similarity matrix is then imported into SPSS 20.0, and the “system clustering” method is used for clustering analysis, ultimately resulting in a co-word clustering dendrogram of deep learning keywords, as shown in Figure 4.
Based on the closeness of the cluster connection distances in the clustering diagram, the deep learning keywords can be aggregated into the following four thematic categories, as detailed in Table 6. The analysis results from the keyword clustering diagram can be used to depict the subsequent multidimensional scaling analysis diagram, further obtaining the knowledge map of research hotspots in deep learning.
Category One: Includes seven keywords: online learning, peer assessment, shallow learning, learning styles, science, biochemistry, collaborative learning.
Category Two: Includes thirteen keywords: deep belief networks, models, unsupervised learning, speech recognition, machine learning, object recognition, autoencoders, convolutional neural networks, classification, feature extraction, image classification, support vector machines, restricted Boltzmann machines.
Category Three: Includes ten keywords: constructive alignment, experiential learning, assessment, critical thinking, education, problem-based learning, learning approaches, teaching, individuality, feature learning.
Category Four: Includes nine keywords: self-efficacy, academic achievement, learning strategies, achievement goals, higher education, student learning, geography, active learning, reflection.
4. Analysis of Research Hotspots in Deep Learning
Multidimensional scaling analysis reveals thematic structures by measuring the distances between thematic words [16]. It simplifies the thematic words in high-dimensional space to low-dimensional space for analysis. Compared to the clustering dendrogram, multidimensional scaling analysis can intuitively judge the position of a research field within the discipline in a lower-dimensional space [17]. The 41×41 keyword dissimilarity matrix is imported into SPSS 20.0, and multidimensional scaling analysis ALSCAI is conducted on the dissimilarity matrix to obtain the research hotspots knowledge map in the field of deep learning, as shown in Figure 5.
By combining the results from the deep learning keyword clustering dendrogram in Figure 4 with the Euclidean distance model scatter plot obtained from the multidimensional scaling analysis in Figure 5, the research hotspots in the field of deep learning abroad can be summarized into the following four aspects:
Field One: Research on deep learning in E-learning environments. From the keywords “Online learning,” “Peer assessment,” and “Collaborative learning,” it can be seen that as research on deep learning continues to deepen and information technology rapidly develops, international researchers are beginning to focus on deep learning research based on E-learning environments, gradually exploring deep learning supported by information technology, which includes online courses, online learning communities, educational games, and SNS (Social Networking Services) platforms and tools [18].
Helen Barren (2006) explored the application effects of a technology-supported deep learning model based on digital storytelling techniques, showing that digital storytelling projects can enhance students’ understanding of course content and improve their problem-solving abilities [19]. Akyol Zehra (2011) conducted empirical research on cognitive presence in online blended learning, demonstrating that cognitive presence in online learning communities is closely related to perceived and actual learning outcomes [20]. Van der Spek, Erik D (2012) experimentally studied activities in serious games that can promote deep learning among students. The results indicated no significant difference between the experimental and control groups in terms of shallow learning participation, but the experimental group performed better in knowledge structure, suggesting that surprising events in serious games can enhance students’ deep learning [21]. Nienke Vos (2011) analyzed the impact of educational games on student motivation and deep learning strategies through empirical research, finding that using educational games can effectively enhance student motivation and deep learning strategies [22]. Pegrum Mark (2015) applied creative podcasts to first-year STEM learning content, analyzing whether creative podcasts could promote students’ deep learning through empirical research. After some time, the enhanced learning results showed that creative podcasts could effectively foster students’ deep learning [23].
From the content analysis of the literature, the research themes in this field specifically include the design and application of deep learning models in E-learning environments, the processes and activities of deep learning in E-learning environments, and the effectiveness of technology-supported deep learning. International researchers mostly adopt empirical research paradigms, using survey and experimental research methods, focusing on designing specific experiments to reasonably apply information technology in deep learning activities during the experimental process to verify practical effects or proposed hypotheses. The specific research subjects include learners in basic education, higher education, and adult education, with a predominance of applications in higher education across medical, mathematical, physical, and scientific disciplines. Overall, international scholars have conducted rich empirical research on deep learning in E-learning environments, often leaning towards using certain technologies to promote deep learning, with comprehensive research considering the overall application of information technology being relatively rare, and most studies habitually approaching research from an educational perspective rather than a technological one.
Field Two: Deep learning research in the computer field. Through the keywords “Machine learning” and “Convolutional neural networks,” it is evident that the concept of deep learning in computer science originates from research on artificial neural networks, representing a new area in machine learning research. Its motivation is to establish and simulate neural networks that analyze and learn like the human brain, mimicking the brain’s mechanisms to interpret data such as images, sounds, and texts [25]. Common models and methods of deep learning include AutoEncoders, Restricted Boltzmann Machines (RBM), and Convolutional Neural Networks.
In the context of big data development, high-tech companies such as Google, Microsoft, IBM, and Baidu have invested substantial resources in deep learning technology, achieving significant progress in fields like speech, image, natural language, and online advertising [26]. Microsoft researchers utilized deep belief networks to directly model thousands of Senones, proposing a context-sensitive deep neural network successfully applied in large vocabulary speech recognition systems—hidden Markov mixed models [27]. D. Bahdanau and others proposed the RNNsearch model, which can predict the target word based on its position and previously translated words during translation, achieving higher scores than traditional models in machine word translation [28]. Apple’s iPhone Siri voice recognition system employs “deep learning” technology, and Baidu has also launched its first voice search service based on deep learning [29]. Deep learning has tremendous development prospects in the field of artificial intelligence. From a research perspective, computer science research on deep learning often adopts design and development methods, innovatively proposing or gradually overcoming technical challenges based on existing deep learning models through computer program algorithms, thereby promoting continuous development in various fields such as speech recognition, image recognition, video classification, and behavior analysis.
Field Three: Teaching applications of deep learning research in the perspective of learning sciences. Based on the keywords “Learning approaches,” “Problem-based learning,” “Teaching,” and “Critical thinking,” the application of deep learning in teaching abroad mainly revolves around deep learning research in classroom teaching and studies on learning methods that promote deep learning. Research indicates that new teaching methods, such as problem-based learning and project-based learning, are more effective in promoting students’ deep learning. Furthermore, Yang Nanchang and others conducted content analysis on the “Journal of Learning Sciences” (1991-2009), revealing that problem-based learning research occupies a core position in learning sciences [30]. Kek, Megan Yih Chyn A (2011) analyzed the significant role of problem-based learning in enhancing students’ critical thinking abilities in digital learning environments [31]. Phan, Huy P (2011) conducted empirical research over two years on deep learning methods and critical thinking development courses, further confirming previous findings that critical thinking can serve as an important information source for students’ participation in deep learning [32]. Loyens, Sofie M. (2013) conducted empirical research on students’ learning methods and academic achievements within PBL environments, indicating that PBL groups could utilize deeper learning methods, effectively promoting students’ deep learning [33]. Mayhew, Matthew J (2012) explored how deep learning approaches influence moral reasoning development among first-year college students, revealing a positive correlation between deep learning approaches and students’ moral reasoning development [34].
Through content analysis of the literature, the main research themes in this field include deep learning research supported by new teaching models, research on cultivating higher-order thinking skills during deep learning processes, and research on deep learning teaching strategies. Analyzing from a research paradigm perspective, researchers tend to adopt empirical research and mixed methods, often utilizing mixed research methods, surveys, and experimental methods in specific teaching practices.
In recent years, researchers have increasingly focused on practical applications and empirical research, widely applying deep learning theories to educational teaching, social work, and other practices to guide teaching activities in fields such as accounting, medicine, mathematics, physics, geography, and biology [35]. The research subjects encompass learners in basic education, higher education, and teacher education across various identities. Deep learning in the perspective of learning sciences emphasizes cultivating higher-order thinking skills and focuses on problem-solving during the learning process, requiring learners to possess high metacognitive abilities. According to existing literature, researchers have conducted practical studies on cultivating learners’ critical thinking, while research on metacognitive abilities and problem-solving processes remains relatively scarce.
Field Four: Research on deep learning processes and outcomes. From the keywords “Higher education” and “Academic achievement,” it can be seen that there is considerable research on the application of deep learning in higher education abroad, primarily focusing on the processes and outcomes of deep learning, including factors influencing deep learning and studies on its effectiveness. Abbas Sadeghi (2012) and others used survey methods to explore the factors influencing deep learning among college students, analyzing that factors affecting college students’ deep learning include learning goals, academic activities, teacher-student characteristics, and learning strategies [36]. Roziana Shaari (2012) conducted a survey on graduate students at Malaysia’s University of Technology, focusing on the relationship between demographic factors and deep learning [37]. Heijne-Penninga (2010) determined learners’ learning levels through moderate information processing tests and used the results from cognitive demand scales to explore the relationship between deep learning, cognitive demand, and test scores [38]. Marlies Baeten et al. (2010) conducted qualitative research on the literature, summarizing various factors affecting deep learning in student-centered learning environments and further analyzing how these factors influence deep learning [39].
Combining the results of the literature content analysis, the research themes in this field can be subdivided into studies on deep learning processes, analyses of deep learning effectiveness, and studies on factors influencing deep learning. From a methodological perspective, empirical research and qualitative research, as well as mixed-method studies, are involved. The exploration of deep learning processes often employs experimental and survey research methods, while the analysis of influencing factors on deep learning primarily utilizes survey research and literature research methods. Overall, researchers’ attention to deep learning processes is still insufficient, lacking representative research results, while research on influencing factors is comparatively abundant, mostly conducted in traditional learning environments. As information technology continues to impact deep learning, focusing on the influencing factors of deep learning in E-learning environments holds significant importance for teaching practice.
4. Research Conclusions and Insights
This paper clearly and intuitively organizes the current research status and hotspots of deep learning internationally through citation analysis and co-word clustering methods. In terms of time, the number of deep learning studies has shown an overall upward trend in the past decade. After 2010, with the rise of research in the field of learning sciences, the number of deep learning studies has significantly increased, greatly enhancing the importance of deep learning in educational research.
From the analysis of the research distribution by country, North America, the UK, and Australia are the main contributors to international deep learning research, with Asian countries represented by China and Singapore. China ranks importantly in the field of deep learning research, both in terms of research quantity and citation index. According to the statistical results of the literature review by Fan Yaqin et al. on domestic deep learning research, a total of 213 articles were published from 2005 to 2014 [40], indicating that deep learning has made significant progress domestically. The distribution of core journals in the field of deep learning is also predominantly in Europe and the United States, covering fields such as computer science, education, and medicine. According to the statistical results of high-frequency words in the literature, the research hotspots in the field of deep learning include education, learning approaches, assessment, neural networks, learning strategies, collaborative learning, etc.
As time progresses, the focus of deep learning research has gradually shifted from machine learning in the computer field to the educational field within the perspective of learning sciences. With the development of E-learning, online learning and technology-supported deep learning have received further widespread attention, with research focusing on various learning methods that promote deep learning, the assessment of deep learning, and the influencing factors of deep learning. Analyzing the research status and hotspots in the field of deep learning abroad brings the following insights for related research in China:
First, emphasize interdisciplinary collaboration. The research on deep learning abroad comes from multiple fields, including computer science, education, and psychology, with numerous interdisciplinary studies on deep learning. The application of deep learning in education needs to draw on ideas and methods from artificial intelligence and computer science based on big data analysis, integrating brain science and educational psychology to study the processes and behaviors of deep learning. This not only broadens the application fields of deep learning but also effectively promotes the occurrence of deep learning in E-learning environments, providing theoretical support for technological developments in the field of machine learning. In the perspective of learning sciences, domestic deep learning research should also focus on interdisciplinary collaboration, utilizing big data deep learning analysis technologies to enhance the effectiveness of deep learning research.
Second, pay attention to the processes and evaluation of deep learning. The process of deep learning includes both explicit learning behaviors and internal cognitive processes [41]. The analysis of core keywords and representative literature shows that international research often focuses on empirically analyzing the learning processes and methods that promote deep learning, as well as the evaluation of deep learning. In contrast, domestic research pays less attention to the processes and evaluation of deep learning and lacks representative empirical research results. Researchers should further focus on the processes and evaluation of deep learning based on research on deep learning environment design. Research on deep learning processes should be more comprehensive, delving deeper into learners’ internal cognitive processes while analyzing their explicit learning behaviors, drawing on research designs from abroad, and emphasizing the analysis of learners’ metacognitive and higher-order thinking.
Finally, focus on technology-supported deep learning research. In the context of the booming development of big data, abundant research results on deep learning in E-learning environments have emerged abroad, exploring the effects of deep learning supported by various technologies through empirical research, such as online learning communities, educational games, and podcasts. Currently, domestic research on deep learning is primarily from the perspectives of education and learning sciences, lacking research on deep learning supported by technology in E-learning environments [42]. Domestic researchers should strengthen exchanges and collaborations with international research teams, attempting to localize and apply the research results of deep learning supported by technology abroad, further enhancing China’s academic position in the field of deep learning.
Authors: Zhang Siqi, Zhang Wenlan, Li Bao (College of Education, Shaanxi Normal University, Xi’an, Shaanxi 710062)
This article is from the Journal of Distance Education
[References] omitted