Empirical research of emerging trends and patterns across the flipped classroom studies using topic modeling.

This study presents topic modeling based bibliometric characteristics of the articles related to the flipped classroom. The corpus of the study consists of 2959 articles published in the Scopus database as of the end of 2021. In addition to the bibliometric characteristics of the field, research interests and trends were also revealed with the study, which was based on the topic modeling-based bibliometric analysis method. According to the results of the study, an increase in the number of publications has been observed since 2015. Nearly one-third of the studies are of United States origin. According to the findings of the topic analysis in which the research interests and trends in the studies were revealed, the articles in this field were gathered under 16 topics. Considering the number of publications of the topic, it was seen that the three most voluminous topics were "Performance and perception", "Nursing education" and "Effectiveness and motivation", respectively. It is thought that the results of the study will provide a general perspective to the researchers in this field and provide important outputs in the context of monitoring the issues that may become prominent in the future.

Introduction

Today’s modern understanding of education has caused change and transformation in learning processes. As a result of the change and transformation, student-centered innovative learning approaches have become prominent. Developing information and communication technologies (ICT) have increased the demand for interactive active learning processes and ensured the inclusion of such environments in education processes (Alvarez, 2021; Latorre-Cosculluela et al., 2021; Singhal et al., 2020). Active learning increases students’ interest, motivation and learning efficiency, reduces learning gaps and thus increases student success (Freeman et al., 2014; Pyatt & Sims, 2012). The flipped classroom (FC) approach has an important place in the organization and implementation of student-centered and active learning environments (Chen et al., 2014; Davies et al., 2013; Karabatak & Polat, 2020; Vivek & Ramkumar, 2021).

Although the FC approach dates back to the 2000s (Bolat, 2016), it has become more convenient and effective in recent years with the development of ICT (Abeysekera & Dawson, 2015; Touchton, 2015). ICT plays a central role in the skill set that students are expected to master during their studies. The rapid development of ICT presents both opportunities for students and teachers (Sointu et al., 2019; Zheng et al., 2020). In recent years, several new technologies with different pedagogical approaches supported by ICT have attracted attention. One of the notable trends among the current trends is the flipped classroom (Adam Beckers, 2017). In other words, the flipped classroom approach is a current trend in educational technology that has the potential to integrate digital technology into teaching and learning processes (Ng, 2015). In this context, many studies can be mentioned that focus on changing pedagogical practices and integrating ICT by applying FC. From this point of view, FC is an example of a good integration of ICT into education and training processes (Sointu et al., 2019).

The basis of this approach, called flipped classroom or flipped learning, is based on the change of classroom and home roles in traditional learning (Lee & Martin, 2020; Thai et al., 2017). In this approach, the operations and processes carried out in the classroom environment in traditional learning were transferred to the home environment, and the operations processes performed in the home environment were transferred to the classroom environment (Bergmann & Sams, 2012; Filiz & Kurt, 2015; Turan & Goktas, 2015). The learning environment is moved from the classroom environment to the individual environment, especially with the help of technology-aided tools. The classroom, on the other hand, is transformed into an environment where previously learned information is reinforced, deficiencies are eliminated and various practices are made (Filiz & Kurt, 2015; Jacot et al., 2014; Sever, 2014; Turan & Goktas, 2015). In the FC process, the first stage of teaching shifts to individual learning instead of group learning in traditional classroom environments, and active learning comes to the fore. Thus, an important part of the processes such as giving information to students about the lesson, understanding and interpreting this information and making practices within the limited course hours are carried out of the classroom (Kong, 2014; Sever, 2014; Turan & Goktas, 2015). In this way, it paves the way for students to come to the classroom environment ready for learning experience (Kong, 2014; Sever, 2014). Besides, the studies carried out in the classroom environment, which is the second stage, turn into dynamic, interactive environments where reinforcement is realized. In this process, the teacher plays a role guiding students in learning new concepts and topics (Davies et al., 2013; Turan & Goktas, 2015).

There are many studies in the literature based on the FC approach. These studies carried out at almost every level and in every field of education, have made significant contributions to the field. In these studies, the effects of the FC approach on the learning experience were investigated from different aspects. While studies have shown mostly positive results, there are a limited number of studies with negative results. The aspects that are generally evaluated as positive in the studies can be listed as follows: learning outcomes and increasing success (Baepler et al., 2014; D’Souza & Rodrigues, 2015; Gaughan, 2014), students’ taking an active role and responsibility in the learning process (Hwang et al., 2015; Mok, 2014; Özyurt & Özyurt, 2017), coming to class in a more prepared manner (Gaughan, 2014; Özyurt & Özyurt, 2017), easing the learning and providing permanent learning (Kocabatmaz, 2016; Mok, 2014; Özyurt & Özyurt, 2017; Turan & Goktas, 2015), increasing interest, motivation and attitude towards the lesson by reducing anxiety (Abeysekera & Dawson, 2015; Butt, 2014; Gaughan, 2014; Kong, 2014; Mok, 2014; Özyurt & Özyurt, 2017; Sever, 2014). In addition to these, the negative aspects, even if they are limited, are as follows: coming to the lesson without making any preparation (Bristol, 2014), mislearning can be failed to notice (Kardaş & Yeşilyaprak, 2015), spending a lot of time for studying and not getting used to this method (Kocabatmaz, 2016; Musib, 2014).

The Covid-19 outbreak has caused significant change and transformation in the field of education as in every field (Kaisara & Bwalya, 2020; Paudel, 2020). With the pandemic, face-to-face lessons have been rapidly transformed into distance and online lessons (Adarkwah, 2021; Paudel, 2020; Rizun & Strzelecki, 2020). In this change and transformation, the FC approach has also been one of the frequently applied methods. During the pandemic period, studies on FC were carried out in different areas (Elkhatat & Al-Muhtaseb, 2021; Haqiyah et al., 2021; Nerantzi, 2020; Pozo-Sánchez et al., 2021; Yen & TF, 2020). The results of these studies are generally very similar to the results of previous studies. In addition to these, it is recommended that new policies should be implemented to extend such practices (Yen, 2020).

Frequent use of the FC approach makes it necessary and important to conduct descriptive studies, which enable a holistic view of the studies in this field, As a matter of fact, there are such studies in the literature using FC approach. Such studies have been conducted frequently as content analysis, systematic review and bibliometric analysis studies. In one study, content analysis was conducted with 20 articles about FC published between 2013 and 2015 (Zainuddin & Halili, 2016). Analysis results showed that FC has positive effects on students’ learning activities such as achievement, motivation, participation and interaction. In addition, difficulties such as poor quality of FC-related videos and not being suitable for untrained instructors have been reported (Zainuddin & Halili, 2016). Another study is a systematic review study with 71 articles (Akçayır & Akçayır, 2018). In this study, it was concluded that the most frequently reported advantage of FC was the improvement of student’s learning performance. Difficulties with this method are related to activities outside the classroom, such as inadequate student preparation before the lesson (Akçayır & Akçayır, 2018). Another systematic review study examining the trends and main findings of studies on FC in the field of English language teaching was conducted with 43 articles (Turan & Akdag-Cimen, 2020). A systematic review study was conducted with 24 articles on the FC approach published in nursing education. (Özbay & Çınar, 2021). In one of the bibliometric analysis studies in this area, the researchers performed bibliometric analysis with 233 FC articles. The study emphasized that FC can provide students with a more interactive environment that will provide higher learning achievement and better preparation for 21st century learning and working environments (Al-Shabibi & Al-Ayasra, 2019). Another study is the study with 346 articles. In the study, it was emphasized that the FC approach is a popular education model among researchers today (Julia et al., 2020). One of the current studies in this field is the study that makes a quantitative summary and qualitative synthesis of the FC literature from 2013 to 2020 (Al Mamun et al., 2021). With this article focusing on 106 FC studies in engineering education, a bibliometric mapping of the field was made. Finally, the bibliometric analysis study conducted with 1557 documents is considered to be the most comprehensive study in this field (Kushairi & Ahmi, 2021).

There are many such analysis studies evaluating studies in different fields of FC approach. Although the importance of each of these studies’ contribution to the field cannot be denied, the lack of a study in which the FC approach is considered as a whole attracts the attention. In other words, there is no study that will provide a greater perspective to the field and describe all the studies related to FC from past to present in depth. On the other hand, content analysis, systematic review and bibliometric studies have difficulties for large datasets because of both the number of articles and the indicators from which they take their basis, and because they are relatively manual analyzes (Gurcan et al., 2021; Yang et al., 2016). To eliminate these difficulties and for in-depth analysis, topic modeling studies based on data/text mining are frequently used (Gurcan et al., 2021; Hu et al., 2014; Kang et al., 2019). In other words, topic modeling studies are regarded as ideal studies to reveal the themes and trends of studies in any field in a comprehensive framework (Gurcan et al., 2021). In this context, with the present study, it is aimed to reveal the trends and tendencies in the articles based on the FC approach from the past to the present. This study, conducted with 2959 articles, is the most comprehensive study in FC as well as the first innovative study to focus on topic modeling-based semantic analysis. In this context, this study is necessary and important as it will look at the field from a broad perspective and describe all the works in detail. In this context, the following research questions were tried to be answered:

RQ1: What is the distribution of articles published in the field of FC by years?

RQ2: What are the prominent journals, authors and subject areas in the articles published in the field of FC?

RQ3: Which are the leading affiliation and countries in the articles published in the field of FC?

RQ4: What are the prominent topics in the articles published in the field of FC?

RQ5: How do the prominent topics in the articles published in the field of FC change over time?

Method

In this study, the descriptive characteristics of the articles in the field of FC were revealed and then the topic modeling analysis was performed. Topic modeling analysis has been frequently preferred in recent years to retrieve hidden semantic patterns in large text sets that are difficult to handle manually, such as natural language processing, information retrieval, and literature review (De Mauro et al., 2018; Gurcan et al., 2021). Latent Dirichlet Allocation (LDA) method was used in the topic modeling analysis. LDA is a generative model based on Bayes’ inference that provides an effective approach to systematically analyze large collections of textual documents; therefore, it is widely used for topic modeling analysis (Blei et al., 2003; Li et al., 2019). Accordingly, LDA, a probabilistic model for content analysis based on topic modeling, was used for the semantic analysis of the empirical corpus of FC literature. LDA is based on the assumption that each document in a dataset covers more than one topic. LDA calculates the topic distribution per document, the word distribution per topic, and the topic and word assignments per document using an iterative process based on the Dirichlet distribution (Blei et al., 2003).

Search strategy and study selection

Scopus database was used to access articles in the field of FC. Scopus is covering more than 7,000 publishers and 240 disciplines worldwide, including Elsevier, Emerald, IEEE, Sage, Springer, Taylor & Francis, Wiley Blackwell, and others (Gurcan et al., 2021; Scopus, 2022a; Yung & Khoo-Lattimore, 2019). On the other and, Scopus is the largest abstract and citation database of peer-reviewed literature such as scientific journals, books, and conference proceedings (Scopus, 2022b). From this point of view, Scopus was deemed sufficient by the researchers. In order to include the maximum number of articles on Scopus, the title, abstract and keywords were searched. The following query was created by examining the studies in the literature by the researchers, two field experts:

TITLE-ABS-KEY ( “flipp* class*” OR “flipp* learn*” OR “flipp* teach*” OR “flipp* educat*”) AND ( LIMIT-TO ( DOCTYPE, “ar” ) OR LIMIT-TO ( DOCTYPE, “re” ) ) AND ( EXCLUDE ( PUBYEAR, 2022 ) ) AND ( LIMIT-TO ( SRCTYPE, “j” ) ) AND ( LIMIT-TO ( PUBSTAGE, “final” ) ).

With this query, all journal articles (research and review) published from the beginning to the end of 2021 was retrieved. The query was run on March 17, 2022. The search yielded a total of 2959 articles (2807 research articles and 152 review articles). The title, abstract and author keywords of these articles were added to the data set. This information, recorded in csv format, was used as an input for the topic modeling analysis.

Data analysis, pre-processing and fitting topic modeling

Within the scope of the study, firstly, the descriptive characteristics of the articles in the field of FC are given. In accordance with this, the number of publications by years, prominent journals, authors and subject areas were determined. Besides, prominent affiliations and countries were also identified. The numerical data obtained for answers of the first three research questions (RQ1, RQ2 and RQ3) were presented graphs and tables. Then, LDA-based topic modeling analysis was conducted to answer the fourth and fifth research questions (RQ4 and RQ5). Topic modeling analysis is a text mining application and requires some data preprocessing steps. The purpose of these processes is to obtain an processable clean dataset of raw data (Aggarwal & Zhai, 2013; Gurcan & Cagiltay, 2019). The pre-processes are divided into three stages as tokenization, removal of meaningless words and lemmatatazion. In the first stage, the texts were converted into simple words, special characters and punctuation marks in the words were cleaned. Then, stop words (meaningless words, etc.) were discarded. The words in the query have also been added to the stop word list. With this step, it is aimed to prevent domain-specific words from being repeated over and over in topics. In the last stage, words are reduced to stem with lemmatatizion. After these stages, the final corpus was created and prepared for analysis. The combination of the title, abstract and keywords was subjected to topic modeling analysis.

Text mining and topic modeling tools in Python were used for both data mining preprocessing stages and for fitting and applying LDA to FC corpus. In order to fit the LDA model to the empirical corpus, firstly, the values ​​of the previous parameters that provide the optimization of the model were selected. The preliminary parameters of α, which determines the distribution of topics in documents, and β, which determines the distribution of words in topics, were taken α = 0.1 and β = 0.01 as the recommended values ​​for topic modeling of short texts (Blei & Lafferty, 2007). An iterative and intuitive process is suggested for model fitting (Gurcan et al., 2021; Mimno et al., 2011). In order to determine the ideal number of topics (K) empirically in LDA-based topic modeling analysis, a model was created for all K values ​​between 10 and 40. Coherence value was calculated for each K value. The coherence value was taken into account in determining the appropriate number of topics in the LDA analysis. The optimum level of the coherence value is accepted as 0.7 (Blei et al., 2003). A coherence value is calculated in each model and the number of topics producing the closest result to 0.7 is considered ideal. As a result of the analysis, a model with 16 topics was decided. The graph of the number of topics-cohorence value is given in Fig. 1.

Fig. 1

Number of topics-coherence score graph

Topic modeling analysis generates topics (such as topic 0, topic 1, etc.) and terms that make up topics. Terms representing topics are produced in order according to their representation rate. Topics are labeled and named by considering these terms (Hindle et al., 2013; Gurcan et al., 2021; Özyurt & Ayaz, 2022). In this direction, the researcher named the topics by considering the terms that represent them. Then, this nomenclature was reviewed together by two educational technologists and the final versions of the topics names were decided together. In addition, the percentage of each topic per document, the distribution of words in each topic, and the distribution of topics within all articles were calculated. At the end of this whole process, first 15 terms with the highest frequency representing that topic were determined for each of the 16 topics.

Findings

The results of the study are presented under two main sections to answer the first three research questions (RQ1, RQ2, and RQ3) and the 4th and 5th research questions (RQ4 and RQ5). Accordingly, firstly, the corpus was analyzed to present the bibliographic features of the FC articles in the corpus descriptively, and the findings were given under the title of “findings of descriptive analysis”. Afterwards topic modeling analysis was applied to discover the implicit topics in these articles and the findings were presented under the title of “findings of topic modeling analysis”.

Findings of descriptive analysis

In order to answer the first three research questions (RQ1, RQ2, and RQ3), the descriptive characteristics of the articles in the field of FC were retrieved. In this context, first of all, the number of articles (n = 2959) in the field of FC from past to present according to years (RQ1) and the slope graph that emerged accordingly are given in Fig. 2.

Fig. 2

Number of publication according to years and slope graph

As seen in Fig. 2, the number of publications increases linearly in the field where 5 articles were published in 2012. Especially the increase in 2015 draws attention and creates a break. Findings on journals, authors and subject areas (RQ2) that stand out in the articles in the field of FC are given in Fig. 3; Tables 1 and 2, respectively.

Fig. 3

Top ten prominent journal and number of publication

Table 1

Top ten authors who have the highest number of publication and their number of publication

Authors	n
Hwang, G.J.	20
Hew, K.F.	18
Lo, C.K.	16
Jeong, J.S.	15
González-Gómez, D.	14
López-Belmonte, J.	14
Belmonte, J.L.	10
Zainuddin, Z.	10
Gopalan, C.	9
McLaughlin, J.E.	9

Table 2

Top ten prominent subject area and number of publication

Subject area	n
Social Sciences	2128
Computer Science	555
Medicine	407
Engineering	400
Arts and Humanities	228
Mathematics	157
Nursing	157
Psychology	152
Business, Management and Accounting	137
Health Professions	101

According to Fig. 3, the journals with the highest number of publication are journals in different fields of educational technologies and sciences. The top three journals with the highest number of publications are Primus (n = 45), Sustainability (n = 40) and Education Sciences (n = 39), respectively. In addition, journals on health sciences and chemistry education draw attention. The first prominent author in this field is Hwang, G.J. (National Taiwan University of Science and Technology), with 20 articles, as seen in Table 1. He is followed by Hew, K.F., with 18 articles, and Lo, C.K. (from The University of Hong Kong) with 16 articles. In terms of the subject areas of the publications it is seen that the first field is Social Sciences (n = 2128), which also includes educational sciences. When the top ten prominent fields are examined, it is possible to say that there are publications on FC in different disciplines (Computer science, medicine, engineering, business etc.). The findings on prominent affiliation and country (RQ3) in the articles in the field of FC are given in Tables 3 and 4, respectively.

Table 3

Top ten prominent affiliation and number of publication

Affiliation	n
Universidad de Granada	38
The University of Hong Kong	35
National Taiwan University of Science and Technology	27
The University of North Carolina	25
Universidad de Extremadura	23
Harvard Medical School	22
The Education University of Hong Kong	19
Monash University	19
Universidad Politécnica de Madrid	19
National Taiwan Normal University	18

Table 4

Top ten countries with the highest number of publication, their publication numbers and rates

Country	n	f
United States	870	29.40%
China	251	8.48%
Spain	233	7.87%
Taiwan	177	5.98%
Australia	142	4.80%
United Kingdom	124	4.19%
South Korea	119	4.02%
Turkey	105	3.55%
Malaysia	93	3.14%
India	80	2.70%
Total	2194	74.15%

As seen in Table 3, in terms of the affiliation origins of the publications the first three places are shared by “Universidad de Granada”, “The University of Hong Kong” and “National Taiwan University of Science and Technology”. In terms of the country origins of the publications, as in Table 4, “United States”, “China” and “Spain” share the first three ranks. The ratio of the publications originating from the top 10 countries to the total publications was 74.15%.

Findings of topic modeling analysis

Findings on prominent topics (RQ4) and trends (RQ5) in articles in the field of FC are given in this section. As a result of the LDA-based topic modeling analysis, 16 topics were discovered. The proportion of each topic was calculated based on the number of articles assigned to each topic. The topics are given in Table 5 in order of volume. In addition, the first 15 terms that make up each topic are listed in order according to the density of creating that topic.

Table 5

Discovered topics, terms that make up the topics, the number of articles assigned to the topics, volume rates and accelerations

Topic name	Topic terms	n	Rate	AccT
Performance and perception	performance, assessment, perception, active, exam, higher, pharmacy, undergraduate, examination, material, grade, final, activity, university, significant	839	28.35%	20.67
Nursing education	nursing, effect, achievement, efficacy, self, satisfaction, strategy, motivation, significant, regulated, directed, design, regulation, skill, undergraduate	417	14.09%	12.04
Effectiveness and motivation	learner, teacher, effectiveness, effect, motivation| perception, questionnaire, significant, instructional, training, impact, digital, environment, online, performance	405	13.69%	11.10
Thinking skills	thinking, skill, critical, learner, ability, technology, design, activity, performance, teacher, content, reading, college, development, mobile	277	9.36%	6.90
Programming learning	programming, performance, computer, experience, load, cognitive, program, design, trauma, lab, assessment, blended, pas, effort, handoff	226	7.64%	7.00
Active learning strategies	technology, active, information, strategy, learning, implementation, innovation, training, student, work, development, teacher, practice, pedagogical, virtual	190	6.42%	6.23
English foreign language	english, language, writing, efl, learner, foreign, instruction, skill, speaking, technology, college, teacher, university, activity, effect	119	4.02%	3.43
Simulation based learning	simulation, development, mobile, neurology, skill, material, feedback, ultrasound, analytics, social, radiation, knowledge, learner, small, activity	100	3.38%	2.71
Mathematics education	mathematics, technology, calculus, blended, informa, trading, limited, design, student, university, instructional, video, published, publication, effect	94	3.18%	1.65
Medical education	medical, curriculum, health, clinical, knowledge, patient, skill, care, student, medicine, resident, training, program, practice, confidence	85	2.87%	2.03
Information literacy	online, information, literacy, mooc, technology, library, service, practice, instruction, open, professional, reality, reform, project, training	57	1.93%	0.73
Engineering education	engagement, activity, engineering, online, university, active, learning, blended, content, higher, experience, strategy, faculty, evaluation, implementation	43	1.45%	0.57
Achievement and attitude	university, achievement, attitude, effect, higher, perception, gamification, technology, significant, motivation, teacher, skill, positive, design, questionnaire	38	1.28%	1.04
Experiences	experience, activity, design, knowledge, environment, peer, technology, practice, challenge, instruction, material, university, team, support, active	37	1.25%	0.80
Professional skills	skill, profession, evidence, competency, faculty, module, technology, instruction, university, program, assessment, smart, team, practice, industry	22	0.74%	0.18
Chemistry education	video, online, chemistry, content, chemical, material, watching, instructor, undergraduate, introductory, collaborative, society, experience, design, concept	10	0.34%	-0.09

AccT: Total Acceleration

Table 5 shows that the most studied (the number of publications with the highest number of publications is called the volume throughout the text) and the three fastest accelerating topics in the FC field are “Performance and perception” (n = 839; f = 28.35%; Acc = 20.67), “Nursing education” (n = 417; f = 14.09%; Acc = 12.04), and “Effectiveness and motivation” (n = 405; f = 13.69% and Acc = 11.10). While calculating the acceleration (Acc), the slope formula was used and the slope value of the data in a range was taken into account as acceleration. The fact that this acceleration value is negative or positive is accepted as an indicator of a decrease or increase in the number of publications on that topic. The time period was divided into certain periods in order to determine the distribution of the discovered topics by years, and the trend of each topic within itself and according to other topics. Since the data set is ten years old (2012–2021), five two-year periods were created. In Table 6, the number of publications of the topic in the periods and the slope values ​​of each topic in the periods are given. With the help of the data in Table 6, the volume (number of articles) and slope graphs for each topic in periods were created. In Fig. 4, the number of articles of the first three topic- period graph is given in order according to the slope.

Table 6

Distribution of the number of articles on the topics according to two-year periods and their acceleration values

	Periods					Total	AccP
	2012–2013	2014–2015	2016–2017	2018–2019	2020–2021
	n
Performance and perception	8	80	175	236	340	839	82
Nursing education	2	25	68	129	193	417	48.6
Effectiveness and motivation	5	29	71	120	180	405	44.1
Thinking skills	6	19	58	88	106	277	26.9
Programming learning	2	14	37	74	99	226	25.4
Active learning strategies	2	8	31	57	92	190	22.9
English foreign language	2	7	23	35	52	119	12.8
Simulation based learning	0	10	17	31	42	100	10.5
Mathematics education	1	16	19	28	30	94	7
Medical education	1	8	21	22	33	85	7.8
Information literacy	2	13	12	15	15	57	2.8
Engineering education	1	10	8	10	14	43	2.6
Achievement and attitude	0	2	5	14	17	38	4.6
Experiences	1	2	9	12	13	37	3.4
Professional skills	1	4	6	4	7	22	1.2
Chemistry education	0	3	1	2	4	10	0.7
Total	34	250	561	877	1237	2959

AccP = Acceleration in Periods

Fig. 4

Volume and slope graphs of the top three topics with the highest volume

As seen in Fig. 4, the first three topics whose volumes increase the fastest over time are respectively “Performance and perception” (AccP = 82), “Nursing education” (AccP = 48.6), and “Effectiveness and motivation” (AccP = 44.1). This shows that these three topics are both the most voluminous topics and the topics that increase the number of publications withing themselves the fastest in time. Volume-period graphs of all topics are given in Appendix 1 in order of acceleration.

With the help of Table 6, the percentages of the number of publications on each topic in each period were calculated. The slope value was determined with these percentages. With the help of these data, the ratio of the number of publications on each topic in the periods has been revealed. In this way, the acceleration of each topic in itself was obtained from the percentages. While this process is being done, the operations on Table 6 are made on a row basis. For example, the number of publications on the topic “Performance and perception” in the periods is 8, 80, 175, 236 and 349, respectively. Accordingly, the percentages of the publications on this topic in the periods are obtained as 0.95%, 9.54%, 20.86%, 28.13%, 40.52%, respectively. In other words, while 0.95% of the publications on this topic were made in the first period, 40.52% were made in the last period. The acceleration of the percentages of this topic in periods (AccPP = Percentage acceleration in periods) was calculated as 9.77. In Fig. 5, the percentices of the first three topics in order of acceleration and slope graphs are given.

Fig. 5

Volume ratio graphs of the top three topics with the highest percentage increase in the number of publications in periods

As can be seen in Fig. 5, the first three subjects whose number of publications increased proportionally over time were respectively “Achievement and attitute” (AccPP = 12.11), “Active learning strategies” (AccPP = 12.05) and “Nursing education” (AccPP = 11.65). When Fig. 5; Table 6 are interpreted together, it can be seen that the top three topics with the most volume are “Performance and perception”, “Nursing education” and “Effectiveness and motivation”, while the top three topics with the fastest increase in percentage over time are “Achievement and attitute”, “Active learning strategies” and “Nursing education”. In other words, the top three topics with the highest slope, the ratio of the publications on that topic within itself is increasing as time goes by. In Fig. 6, the percentige slope (AccPP) values ​​of all topics in the periods are given in order.

Fig. 6

Slope values of all topics’ publication rates in periods

As can be seen in Fig. 6, the publication rates of all topics in periods are gradually increasing. However, the top three topics that accelerated the most within themselves - although the slope values ​​are close to each other - are “Achievement and attitute”, “Active learning strategies” and “Nursing education”, respectively, and they are followed by “Programming learning”. The last three topics with the slowest rate of increase are “Engineering education”, “Professional skills” and “Information literacy”, respectively.

In addition to the percentages of the topics over time within themselves, the percentages compared to other topics in each period were also obtained from the data in Table 6. Considering the number of publications of the relevant topic in each period, the percentage rate was calculated compared to other topics. A slope value (AccPTP = Percent acceleration relative to other topics in periods) was determined again with these percentages. With the support of these data, the percentage weight of each topic was calculated compared to other topics in the relevant period. While this process is being done, column-based operations are performed on Table 6. For example, in the first period, a total of 34 publications were made, 8 of which (23.53%) were on “Performance and perception”. The number of publications of this topic in other periods and the number of publications in that period are in pairs (80;250), (175;561), (236;877) and (340;1237), respectively. Therefore, the percentage weight of the “Performance and perception” topic in each period compared to the other topics was calculated 23.53%; 32.00%; 31.19%; 26.91%; 27.49% respectively. In summary, while the percentage weight of this topic was 23.53 (23.53% of all publications on this topic) in the first period, it was 27.49 in the last period. This calculation was made for all topics and the percentage rate of each topic in the relevant period was calculated compared to other topics. Based on these ratios, an acceleration value (AccPTP) was calculated for each topic. In Fig. 7a and b, the ratio and slope graphs of the top three topics, which increase and decrease their weight over time, are given respectively.

Fig. 7

The top three topics that increased (a) and decreased (b) the percentage rate compared to other topics in periods

As can be seen in Fig. 7, the top three topics that increase their weight in percentage among other topics over time are respectively “Nursing education” (AccPTP = 2.41), “Programming learning” (AccPTP = 0.71) and “Active learning strategies” (AccPTP = 0.64). However, the top three topics, whose percentage weight decreased over time compared to other topics, were “Engineering education” (AccPTP = -0.65), “Information literacy” (AccPTP = -1.28) and “Thinking skills” (AccPTP= -1.57). Considering the slope values ​​of all topics, the evaluation of the weights in the periods is given in Fig. 8.

Fig. 8

Change in weights of all topics over time compared to other topics

As can be seen in Fig. 8, among all the studied topics while the percentage of seven topics increased over time, the weight of nine topics gradually decreased. Nursing education, which is the topic with the most increased weight among other topics, is also in the second rank among the most voluminous topics. The most voluminous topic, “Performance and perception”, was the second to last topic with increasing weight compared to other topics.

Finally, the prominence time of the prominent topics among the other topics were examined on the periods. In this way, the periods in which the first seven topics with positive momentum value became prominent were described. For this reason, the percentages in each period were listed among these topics and the first three were taken. “Performance and perception”, “Effectiveness and motivation” and “Thinking skills” topics constitute the first three ranks in the first period. The topic “Thinking skills” is not shown in the schedule as it has a decreasing weight. Starting from the first period, this process was iteratively examined in all periods. Accordingly, the chart shown in Fig. 9 was created.

Fig. 9

Timeline of prominence of prominent topics compared to other topics

As can be seen in Fig. 9, the prominence times of seven topics, whose weight is gradually increasing compared to other topics, differ from each other. As a matter of fact, the first two topics that have continued since 2012 till our day were determined as “Performance and perception” and “Effectiveness and motivation”. Since 2014, “Nursing education”, “Programming learning” and “Simulation based learning” topics have started to become prominent. While the topic of “Active learning strategies” gained momentum since 2016, the topic of “Achievement and attitute” has become prominent since 2018.

Discussion and conclusions

In this section, the results obtained from the study are discussed in accordance with the research questions and a future perspective is presented in the light of these. The chapter is organized under two headings. The first title addresses the first three research questions; the second title was created for the 4th and 5th research questions.

Reflections from the descriptive analysis

In order to answer the first three research questions (RQ1, RQ2 and RQ3), the total number of articles by years, the prominent journals, authors and subject areas in the articles and the publication trend in the field of FC in the context of publication by institution and country were examined. The results of the study showed that (in the context of RQ1), the number of articles in 2021 reached 634 in the field where there were 5 articles in 2012. In this context, Parslow (2012) successfully predicted the threshold of the quantum leap of flipped madness in the second decade of the millennium in his study. In addition, the “flipped classroom” in the academic field was first discussed in a Horizon Report in 2012 and was introduced as a promising model that could open the door to new approaches in education (Johnson et al., 2012). In this context, the results of the study confirm these. In terms of the year of initiation of the studies in the field of FC and the distribution of publications by years it is possible to say that the FC approach will attract more and more attention in the following years by an increasing number of practitioners, researchers and scientists. This prediction is based on the rapid increase in technology-based education, which is seen as a savior especially in difficult times when face-to-face physical meetings and education cannot be held or applied. The best example of this is the current COVID-19 pandemic, which is promoting and promoting emergency distance learning globally (Kushairi & Ahmi, 2021). It is known that there is a general increase in FC-related publications during the Covid-19 period (Elkhatat & Al-Muhtaseb, 2021; Haqiyah et al., 2021; Pozo-Sánchez et al., 2021).

In terms of the prominent author, journal and subject areas in the articles published in the field of FC (in the context of RQ2) it is seen that the author with the most publications in this field is Hwang, G.J. (n = 20), journal with the highest number of publication is Primus (n = 45), and the most prominent subject area is Social Sciences (n = 2128). Hwang, G.J. (n = 20) is followed by Hew, K.F. with 18 publications and Lo, C.K. with 16 publications. When the journals in which the articles are published are examined, it is seen that journals with the highest number of publications are journals in different fields of educational technologies and sciences. In addition, it was seen that the journals in the field of health sciences and chemistry education were among the top ten journals. Although the number of articles published in the journals is different, it can be said that most of the journals in the studies in the literature (Julia et al., 2020; Kushairi & Ahmi, 2021) are in the top ten journals in our study. In addition, the fact that medical and chemical education journals have become prominent in this study, as in the literature (Kushairi & Ahmi, 2021; Yang et al., 2017) can be related to the state of being interdisciplinary, which will become more meaningful when the subject area distribution is examined. Indeed, when the top ten subject areas were examined, it was seen that the first area was “Social Sciences” (n = 2128), which also includes educational sciences. The fact that the field of educational sciences is under social sciences has highlighted this field (Kushairi & Ahmi, 2021). It was observed that this area was followed by the “Computer Science” (n = 555) and “Medicine” (n = 407) areas, respectively. Apart from these, there are different disciplines such as “Engineering”, “Nursing”, “Psychology” in the top ten. Although the ranking in the studies in the literature changes, the distribution of the subject area obtained in the study is very close to the study of Kushairi & Ahmi (2021). Indeed, the first four subject areas are the same, and the subject areas in 3rd and 4th rows have changed places. the “Engineering” and “Medicine” fields, which shared the 3rd and 4th places in the other study, changed palces with each other in our study. The subject areas of the publications in this field are generally similar, which supports the emphasis that the field has the opportunity to be applied on an interdisciplinary platform (Julia et al., 2020; Kushairi & Ahmi, 2021; Yang et al., 2017).

In terms of the affiliation and country origins of the articles published in the field of FC, (in the context of RQ3) the top three institutions with the most publications were Universidad de Granada (n = 38; origin:Spain); The University of Hong Kong n = 35; origin: Hong Kong) and National Taiwan University of Science and Technology (n = 27; origin: Taiwan). When the origins of the articles were examined, it was seen that the most publications were from the United States (n = 870; f = 29.40%). It is followed by China (n = 251; f = 8.48%) and Spain (n = 233; f = 7.87%), respectively. Almost one-third of the articles originated in the United States. Although the results are in parallel with the prominent institutions and countries in the studies in the literature (the first two countries are in the same rank) ) (Çakır et al., 2020; Kushairi & Ahmi, 2021; Yang et al., 2017), in our study it has been observed that publications originating from countries such as “Spain”, “United Kingdom” have increased. Indeed, unlike the current studies, the institution with the highest number of publication (Universidad de Granada) originated in Spain. While Yang et al., (2017) and Çakır et al., (2020) state that the institutions and countries in the top ten are mostly of western origin, the results of our study are parallel with the results of Kushairi & Ahmi (2021) and showed that FC applications tends to increase in developing Asian countries especially such as Taiwanese, Turkey, Malesia and India. This situation can be interpreted as an indicator of the importance that developing countries attach to educational technologies (Kushairi & Ahmi, 2021).

Reflections from the topic modeling analysis

The results of the topic modeling analysis (in the context of RQ4) showed that the articles in the field of FC were grouped under 16 main topics. Among these topics, the three most voluminous – having the highest number of publication/the most studied – are respectively “Performance and perception” (n = 839; f = 28.35%), Nursing education (n = 417; f = 14.09%) and Effectiveness and motivation (n = 409; f = 13.69%). The number of publications on these three topics constitutes more than half of the total number of publications (56.13%). “Programming learning”, “Active learning strategies” and “English foreign language” follow these topics in terms of volume. The least voluminous topics are respectively “Experiences”, “Professional skills” and “Chemistry education”. FC has evolved into useful learning at his own pace for all types of students or mastery learning paradigm known since the age of constructivist theory (Kushairi & Ahmi, 2021). From this perspective, the FC approach suggests the student-centered strategy which has been promoted by leading scholars for along time. In this regard, it is not surprising that the effects of FC-based learning environments on students’ learning performance, perceptions, motivations and the effectiveness of these environments are among the most studied topics. It is possible to mention many studies on these topics in the literature (Blair et al., 2016; Thai et al., 2017; Touchton, 2015). In addition to performance and perception, topics such as “Achievement and attitude”, “Experiences”, “Professional skills” have emerged as FC-based learning environments create an active learning environment and therefore the effects of these environments on learning experiences are examined from different aspects withing these topics. In this regard, it can be accepted as an indication that the FC approach is frequently applied and evaluated from different dimensions.

When the topics obtained from the topic modeling analysis are examined, it is seen that topics such as field teaching also draws attention. As a matter of fact, “Nursing education” (the second highest voluminous topic), “Programming learning”, “English foreign language”, “Mathematics education”, “Medical education”, “Engineering education” and “Chemistry education” are the topics emerged in this direction. The increasing complexity of the nursing and healthcare environments in recent years (Barbour & Schuessler, 2019; Njie-Carr et al., 2017) has brought along significant changes and transformations in the medical field, especially in nursing education. With this change, innovative learning strategies have been implemented and FC has been widely used for this. Many studies in the literature have emphasized that the FC approach effectively supports learning in the field of nursing and health and increases the quality of learning (Özbay & Çınar, 2021; Youhasan et al., 2021).

The fact that the FC approach, which forces students to learn basic medical information and create skills and abilities beyond remembering information, is regarded as a promising modality in medical education (Barbour & Schuessler, 2019) may have highlighted the studies in this field. “Programming learning” is among the most voluminous area-based topics. Difficulties in programming learning (Ruiz de Miras et al., 2022) have triggered student-centered learning approaches and many studies have been implemented in this direction (Ruiz de Miras et al., 2022; Taşpolat et al., 2021). The belief that the positive results of the FC approach help to overcome the difficulties in programming learning may have led to intense studies in this area. In addition, the positive aspects of the FC approach and its frequent use in the field of education have triggered the frequent implementation of studies in different fields.

Finally, in order to determine the trends in FC studies (in the context of RQ5), the acceleration of the topics over time and according to other topics was examined. The first three topics, whose intensity of study increased the most over time, are “Achievement and attitute”, “Active learning strategies” and “Nursing education”, respectively. The last three topics with the slowest rate of increase are “Engineering education”, “Professional skills” and “Information literacy”, respectively. Nursing education is one of the three most voluminous topics in terms of volume. In other words, “Nursing education” is common among the top three most voluminous topics and the top three topics that increase the intensity of study the most over time. On the other hand, in terms of the trends of the topics over time, the first three topics that increased their weight in percentage among other topics over time are “Nursing education”, “Programming learning” and “Active learning strategies”, respectively. The topics of “Nursing education” and “Active learning strategies” are also among the topics whose intensity of study increases over time. The fact that nursing education is an important criterion for the quality of the health system (Kim et al., 2019) has encouraged researchers in this field to implement innovative studies (Youhasan et al., 2021). The importance given to “nursing education” may have triggered this topic to be one of the most accelerated topics. Similarly, the difficulties and troubles experienced in programming learning (Özyurt & Özyurt, 2018; Ruiz de Miras et al., 2022) may have highlighted the implementations in this field. Enabling student-centered learning environments in FC studies flourishes by the student’s taking responsibility for learning. FC constructivism is an important approach that encourages active learning (Shapran & Shapran, 2019; Stöhr & Adawi, 2018). In this context, as aresult of the increase in FC studies in recent years the topic of “Active learning strategies” may have become prominent. The first three topics, whose percentage weight decreased over time compared to other topics, were obtained as “Engineering education”, “Information literacy” and “Thinking skills”.

There are a total of 7 topics, which increased in intensity over time (showing positive momentum) compared to other topics. According to the timeline divided into two-year periods, it has been observed that the topics of “Performance and perception” and “Effectiveness and motivation” have been studied since 2012. This situation can be interpreted as the evaluation of the effects of the first studies in the field of FC on learning outcomes. Since 2014, “Nursing education”, “Programming learning” and “Simulation based learning” topics have started to become prominent. While the topic of “Active learning strategies” gained momentum since 2016, the topic of “Achievement and attitute” has become prominent since 2018.

Limitations and future works

With this study, it is aimed to examine the research interests and trends of the articles in the field of FC from the past to the present (until the end of 2021). The most innovative aspect of the study is that it is the most comprehensive study based on topic modeling-based semantic analysis. However, it is also possible to mention some limitations of the study. First, the dataset is limited to peer-reviewed articles in the Scopus database. Databases other than scopus and all document types can be included in the corpus by scanning more comprehensive databases. In addition, the widely used - accepted in the literature - LDA algorithm was used for the topic modeling analysis. In future studies, comparative studies can be carried out with different algorithms. Another important issue is the necessity of repeating such studies throughout FC or its sub-topics (especially the most voluminous or the fastest accelerating issues) in certain time periods. In this way, it can be seen how the situation in the current study continues in the future and how the increasing/decreasing trends in the topics have changed. In this way, trend changes in the field will be followed in a healthy way.