Abstract
Studies on path dependence (PD) and lock-in have indicated several limitations, such as chaotic literature information, interdisciplinarity complexity, and evolution vagueness. In this study, literature on path dependence and lock-in are obtained from the database of Web of Science Core Collection from 2001 to 2017. The knowledge mining model of the evolution of path dependence and lock-in is constructed based on knowledge mining process and content. The evolution of this area is also explored. Results show that the evolution of path dependence and lock-in can be categorized as follows: (1) English, American, and German scholars have led the research on path dependence and lock-in. (2) The research topics have undergone four stages. (3) The research levels have been classified as policy system (macro-level), regional economy (meso-level), and organizational path dependence (micro-level). (4) The research design can be represented as theory basis, case discussion, lock-in or reformation, and path creation. (5) The research method can be transformed from case study into lock-in mathematical modeling. A panoramic knowledge mapping of path dependence and lock-in is displayed by the research system to enhance understanding on research trajectories and future research directions.
Introduction
The research on path dependence (PD) and lock-in in humanities and social sciences begins with David’s (1988) analysis of the technology lock-in of the QWERTY keyboard, which combines state and process [1]. The PD in the process shows a non-ubiquitous random dynamic process and it is influenced by historical events. The result of PD shows that the lock-in state is locked in the past structure or path [2]. Thus, PD forms a lock-in after a certain development process, which is a type of lock-in. Lock-in results from PD, and PD results in lock-in. Furthermore, PD and lock-in are unequal to the history of “accident”, and are not determined by initial conditions [3–5].
The current literature on PD and lock-in are numerous and complex. However, few studies focus on the research and analysis of PD and lock-in using knowledge mining (KM). So this study constructed the KM model (KMM) of PD and lock-in evolution and obtained the literature on PD and lock-in from the database of Web of Science (WoS) Core Collection as an example to realize KM and knowledge discovery (KD). The WoS is a worldwide authoritative citation index database, and the relevant literature collected from it satisfies the KM and explores its evolution. In addition, CiteSpace is a common-used tool for mapping knowledge domains, and its input data adopts the WoS data format as a standard to realize knowledge visualization by mapping knowledge domains. The research helps scholars grasp the research trajectories, select future research directions, and enlighten other researchers.
Research status
Many organizations and management journals have applied the proposed PD and lock-in theory, and numerous scholars have adopted it to help explaining topics such as institutional changes [6, 7], economic evolution [8], geographical area [9], organizational changes [10], and technological innovation [11, 12]. In 2013, Kay initiated a heated debate about relevant concepts, formation factors, results, and even research paradigms in PD and lock-in, which are what scholars have mainly focused on [3]. After nearly 30 years of development, the literature that focused on PD and lock-in has also been enriched. 20.6 percent of papers published in Research Policy used the PD theory [3]. However, several limitations exist in the complex literature, such as literature information chaos, interdisciplinarity complexity, and vagueness.
To research method, the simulation, which involves a number of activities related to the experimental determination of the effects that occur in the system, process or model which imitates them, is beneficial to restore true facts, such as the process of PD and lock-in [13]. However, Vergne J.P. and Durand R. highlighted the lack of interaction between the theory and empirical analysis of PD and lock-in and demonstrated the methodological significance of three empirical research methods: computer simulation, experimental research, and counterfactual investigation [2]. Vergne J.P. (2013) stated that the empirical research on PD is limited [4].
In general, the notion of simulation, involves a number of activities related to the experimental determination of the effects that occur in the system, process or model which imitates them.
However, the scope of these studies is limited. The conclusions based on qualitative analysis are extensive and lacking in measurement basis. Moreover, scholars have focused on the realization of knowledge visualization and discovery, and the use of KM method and technology in recent years. However, the literature using KM method to explore the evolution of PD and lock-in has not yet been discovered.
Therefore, the study applied the method of KM and obtained the literature on PD and lock-in from WoS Core Collection. Visual analysis is then performed to support the above statement. Furthermore, the stages of research theme changes and development trends for the recent 30 years are divided. Research topics at each stage are explored. Research hotspots and frontiers in the past 5 years are analyzed to grasp the trend in the filed.
The remainder of this paper is structured as follows. Section 2 describes the research status of PD and lock-in. Section 3 introduces the KM model of PD and lock-in evolution, and the CiteSpace implementation of KM model. The results discussion and conclusion are respectively presented in Sections 4 and 5.
Research method
KMM of evolution construction
According to the analysis of evolution dimensions, and the KM framework, the KMM of evolution was constructed.
Evolution dimensions analysis
To clarify the evolution of PD and lock-in, three dimensions should be started from which are research distribution, research topics, and research hotspots and frontiers. Research distribution explains the time and space distribution among exciting literature and examines the temporal trends and spatial distribution of the field. Research topics include the research stages and research themes evolvement. Research stages present the diachronic question. Research themes evolvement includes topic analysis and detailed descriptions of each stage. Research hotspots and frontiers focus on related literature over the recent five years and identify future research trends.
KM framework
KM acts as the key factor of knowledge discovery, particularly emphasizing the processes and steps. KM is the process of transforming data into knowledge from the perspective of concept, and information visualization, which can be used to reveal the relationship between abstract information and hidden features, is indispensable between them [14]. KM is divided into three layers: data, information and knowledge layer. The KM framework is then summarized according to the present study (Fig. 1).
KM framework.
From the process perspective, KM is followed by database selection, data preparation, data filter, data mining, knowledge representation, knowledge evaluation, and knowledge base. Among them, the first three processes are the data layer, and data mining corresponds to information layer. The last three belong to the knowledge layer. From the content perspective, the data structure in the data layer can be categorized into the structured, semi-structured, and unstructured data. Structured data mostly refer to specifications formation and structural descriptions, such as relational data. Unstructured data are shown through text, video, and picture. However, semi-structured data are usually visible in HTML files. The information layer includes data mining target, content, and technology. The knowledge layer contains knowledge discovery and application.
A KM model of PD and lock-in evolution is constructed through combining the evolution dimensions and KM framework (Fig. 2).
KMM of PD and lock-in evolution.
In the data layer, the network sci-tech literature has been selected due to its massive research papers, rich proprietary research and great academic value. In addition, they also include time, countries, institutions, title, abstract, keywords, citation, and other contents such as research topic and future research direction.
In the information layer, the research topic determines the data mining content. The cooperation analyses of countries and institutions are matched not only through statistics on the number of contributions by countries and institutions, but also among their relationship networks. The research stages division and themes evolvement considering the large amount of data in diachronic research. Thus, cluster analysis is adopted in this study. The knowledge base can be reflected by co-citation, which indicates themes evolvement in the field. Co-occurrence analysis examines the research hotspots and frontiers shown as keywords.
The knowledge layer involves the evolution of PD and lock-in including the research levels, research design, and research method. The research findings help researchers grasp the topics in this field and enlighten future researchers.
CiteSpace is a common-used software for scientific knowledge mapping [15]. To achieve national and institutional cooperation, co-citation cluster and co-occurrence analyses, CiteSpace is selected to assist the KM analysis of PD and lock-in evolution. A CiteSpace implementation model of KM (Fig. 3) is developed by combining KM content and CiteSpace conditions. The implementation of data collection and filter, mapping, and special node selection is separately explained as follows.
CiteSpace implementation of the KMM of PD and lock-in evolution.
In July 15, 2017, some literature on “lock-in” and “path dependence”, or “path-dependent”, or “path dependency” were obtained from the WoS Core Collection database. To ensure the reliability of the results, literature with incomplete abstract or keywords information and themes that do not match with this study were removed. Finally, 328 valid data has been selected by this paper. Among them, 319 are journal papers and 9 are review papers. A total number of 172 data are from the past 5 years accounting for 55% of total. Citation information index is then obtained, and the CiteSpace5.1.R1 is applied to remove duplicate data.
Mapping drawing
CiteSpace processes literature information using large number and out of order as input, and exports the knowledge mapping of co-citation and keywords co-occurrence analyses. Three visualization methods, cluster, time-line, and time-zone, are applied to examine the research themes evolvement, hotspots, and frontiers in PD and lock-in. In the co-citation analysis, the phenomenon of two citing references is mentioned by the same literature and reveals the research topics evolution through the analyses of clusters, key nodes, and mapping color. Keywords co-occurrence analysis involves two key words appeared in the same literature and reflects research hotspots and frontiers by examining the frequency and betweenness centrality of keywords.
Special nodes selection
Special nodes selection is vital because it occupies a key position in the knowledge network and plays a specific role in the evolution of knowledge structure. These special nodes are usually transformed into contribution frequency, betweenness centrality, and Sigma value.
References or keywords with high frequencies, also can be called a statical analysis of data, reveal the development or evolution: research distribution, current and previous research hotspots in many fields from social science to engineering technology, such as the development of the migration problems [16], and knowledge discovery issues [17]. The tree rings of nodes represent knowledge mapping frequency. When the radius is large, the frequency is high. The high betweenness centrality of nodes is the key nodes and turning points of knowledge evolution in this field. Thus, the node plays a key role and it is marked by a purple circle in the knowledge mapping. High burst refers to the rapid increase of keywords or references in a short time, which becomes the research focus in this time period and is tagged in red circle in the knowledge mapping. CiteSpace adopts Kleinberg’s algorithm (2003) to compute the occurrence of citations through burst parameters [17]. The concerned areas can be summarized by viewing the cited references according to burst degree. These areas, to some extent, represent the research topics and hot issues in corresponding stages and even research themes evolvement. Sigma algorithm reflects centrality and burst.
Results and discussion
Research distribution of PD and lock-in
To clarify the research distribution of PD and lock-in, the time and space distribution were explored.
Time distribution
As illustrated in Fig. 4, the number of PD and lock-in studies has increased annually, and they can be divided into three stages according to time: start-up stage from 2001 to 2013, gradual growth stage from 2004 to 2010, and blowout growth stage from 2011 to 2017. The size of studies has significantly increased in 2011, compared with that in 2010. In 2012, the number of studies peaked at 40. The gradual annual growth of PD and lock-in internationally indicates that they became a recent research focuses.
Time distribution of PD and lock-in from 2001 to 2017.
Table 1 indicates that the USA is a leading core country with a number of 78 studies accounting for 24% of total. The second is England with 54, accounting for 17%, followed by Germany with 41, accounting for 13%. However, only 10 published papers are in China. China’s contribution is not outstanding, but it is in steady growth. Moreover, USA, England, and Germany are also the top three in terms of betweenness centrality, indicating that they have more cooperation with other countries. From the perspective of literature institutions, most papers in this field originated from universities and the representative research institutions such as Erasmus University, Aalto University, Free University Berlin, University Michigan, and University Melbourne. However, the University Helsinki, University Namur, and American Political Science Association are the top three in terms of betweenness centrality. They play important roles in network structure with increased cooperation from other institutions.
Country and institution distribution of PD and lock-in
Country and institution distribution of PD and lock-in
CiteSpace 3.9.R1 is used to carry out the co-citation analysis of the 328 literature, and the time span applied is from 2001 to 2017. The node types are the cited references. The top N per slice equals 50, and the minimum spanning tree to prune the network is selected. The co-citation network with 311 nodes and 415 lines were obtained in 10.699 seconds. After clustering, the module value Q equals 0.7115, and S equals 0.7903. Both values are greater than 0.5, indicating that the network structure of the 16 clusters is clear and the clustering results are distinct. However, the citations from clusters 11 to 15 are both 5 (less than 10). Thus, only 11 contents, from cluster 0 to 10, are presented in this study (Fig. 5). The cluster label is based on the tf*idf algorithm from the extracted word title citation annotation. Figure 5 shows intuitively that the period between 1985 and 2010 is the development peak of PD and lock-in.
Time-line mapping of 328 cited references on PD and lock-in.
Figure 5 exhibits five important literature with high frequency and high betweenness centrality. David P.A. (1985, 85/0.11. 85 is the frequency value, whereas 0.11 represents centrality, and the following number remain the same) introduced the concept of PD in social sciences by analyzing the technology lock-in of QWERTY keyboard [1]. Arthur W.B. (1989, 57/0.37) first proposed the concept of path lock-in based on the study of David and explored the lock-in by historical events [18]. Arthur W.B. (1994, 44/0.23) described the PD of economic growth and stressed increasing returns, self-reinforcing mechanism, and small event lock-in of PD by analyzing competition, industry location pattern, city system, and strategic pricing [19]. Pierson P. (2000, 75/0.57) studied PD in the political development process based on the mechanism of increasing returns [20]. From the type and operation pattern of historical sequence events and development process and result of the event chain, Mahoney J. (2000, 64/0.25) dissected the formation and recognition of PD in historical institutionalism [21].
High frequency literature, which can be classified by citation burst (Table 2), does not explicitly explain research themes evolvement. Figure 5 and Table 2 display the four stages of PD and lock-in study: between 1985 and 2004, between 2005 and 2007, between 2008 and 2012, and between 2013 and 2017.
High citation burst references
To further explore the themes evolvement in PD and lock-in, the co-citation cluster mapping is drawn at each stage. The clustering labels of the three algorithms are used to discover the themes evolvement in PD and lock-in based on title and keyword extraction.
Between 1985 and 2004: Policy institution– customer lock-in– technological change
The data period is from 2001 given the limitation of the database. However, the citations reflect the knowledge base of the field, and the previous literature (before 2011) can be analyzed by above citations. Thus the summary of the themes is not affected.
Figure 6 shows that Clusters 1 and 4 are included in 2001. Among them, Cluster 1 has 29 cited references, the label of which is PD and contains several tag words, such as system and democratic consolidation. Cluster 1 mainly discusses PD at the institution level. Cluster 4 has 15 cited references and is marked by firms in which the tag words are customer lock-in, network externality, substitution theory, chaos theory, and QWERTY keyboard. This clustering also explains PD and lock-in at the firm level.
Cluster mapping of cited references between 2001 and 2004.
Clusters 2 and 6 belong to the 2002 category. Cluster 2 has 35 cited references, the label of which is institutional transfer and the tag words are labor relation, system transfer, East European economies, economy, and mainly concerns the PD of institutional transfer at the level of institution. Cluster 6 is labeled by innovation from 14 cited references, and it contains several tag words, such as complex technology, technological change, internationalization, system, network, and time and is all about the analysis of PD at the level of technological innovation.
Clusters 0 and 3 are included in 2003. Cluster 0 is labeled by bio technology from 38 cited references, and tagged with institutional isomorphism. This cluster still explores institutional PD. Cluster 3 has 18 cited references labeled by path creation, and it contains several tag words, such as environment, substitution theory, network externality, and lock-in. Furthermore, it includes PD and path creation in the view of environment.
Clusters 5 and 7 belong to the 2004 category. Cluster 5 is composed of 15 cited references and labeled by France. The tag words are labor market policy, politics, France, Germany, and social protection. Institutional PD is emphasized in the political development process. Cluster 7 has 10 cited references and is labeled by urban structure including urban areas, traffic, lock-in, land use, and decision making. This cluster focuses on PD of city location and traffic in the view of economic geography.
Therefore, they all boil down to PD at the policy and institution level, which includes institutional change, political development policy, economic geography, urban location and traffic, and environmental regulation. Important references and Clusters 0, 1, 2, 3, 5, and 7 are all included. Therefore, the study of this stage focuses on policy and system, stressing the institutional change and convergence. The second is the customer lock-in at the company, technical change, and innovation levels. Thus, the development track can be summed up as follows: policy system – customer lock-in – technological change. Moreover, as the forming factors of PD and lock-in, network externality is more analyzed at this stage.
Figure 7 shows that Cluster 3 belongs to the 2005 category. Cluster 3 has 16 cited references and is labeled by cluster learning. Furthermore, this cluster contains several tag words, such as regional economies, regional economic development, Germany model, financial market, knowledge transfer, cross-border integration, strategic alliances, switching cost, lock-in, and switching cost. It is found that the regional economy becomes a new research area that focuses on new research topics, such as the development and evolution of regional economic and cluster development. Strategic alliances, knowledge transfer, and switching cost are the keywords at this stage.
Cluster mapping of cited reference between 2005 and 2007.
Cluster 1 is included in the 2006 category, and the label is prospective voluntary agreement extracted from 16 cited references. The tag words are environmental policies, policies and methods, avoid technical system lock-in, technological lock-in, technical defects, PD, sustainable development, income increase, framework, and innovation. Therefore, the PD study in the view of environment is stills the research topic at this stage, but it is no longer confined to the environmental policy level at the last stage. The analysis of the lock-in of environmental technology and technological system from the perspective of sustainable development and strategic framework formulation are emphasized to avoid lock-in.
Cluster 2 corresponds to the 2007 category with 16 cited references. The label is transport planning, and the tag words are Pakistan and PD. Furthermore, it stresses the research on PD and lock-in in government transportation planning.
In addition, the lines between the nodes of Clusters 0 and 4 are complex and span for 3 years. They are bound to be the main research topics at this stage. Cluster 0 is composed of 17 cited references and labeled by case study method. The tag words are examples, qualitative research, complex causal relationship, PD, regional economic evolution, regional lock-in, regional path innovation, increasing returns, network externality, and national sovereignty. Cluster 4 is labeled by Sweden from 12 cited references, including industrial emission control, New Zealand case study, public investment, and the United States. It is an important research direction that involves adopting case analysis method to explain the PD and lock-in of regional economic evolution, national sovereignty, and public investment in different countries.
Figure 8 shows that the color of the internal lines of Clusters 1, 2, and 3 are unitary, whereas those of others are complicated. Therefore, the clusters are summarized by labels and theme words. They are labeled from P1 to P7 with the time sequence (Fig. 8).
Cluster mapping of cited reference between 2008 and 2012.
P1 is labeled by policy system and contains Clusters 0, 1, and 4. Among them, Japan is the label of Cluster 0, and the theme words are unbalanced processes, institutional changes, public assistance policies, and urban development. The cluster is labeled by stagnation, and the theme words are as follows: Australian social policy, war, and path. Cluster 4 is national park tourism partnership, which includes PD, policy changes, and European welfare states. The PD and lock-in explanation of policy and institution changes can be obtained from the literature on national development. Therefore, the policy system is used as a summary.
P2 is the customer lock-in and corresponds to Cluster 3. Cluster 3 is labeled by unsustainable consumption with several tag words, such as sustainable consumption, unsustainable consumption, luxury goods, individual demand, locking, and consumer lock-in.
P3 corresponds to Cluster 5, which is labeled by urban areas, including dependent processes, overcome lock-in, automotive technology, and radical ecological innovation.
Cluster 6 belongs to P4 and is labeled by case study. The theme words are case study of Belgian agriculture lock-in, resistance wheat varieties, and case.
P5 is the debate between lock-in and reform, and it is composed of Clusters 2 and 8. Cluster 2 reinforces lock-in and includes several themes, such as adjust the infrastructure system, lock in, change, and promote reforms. This cluster mainly focuses on discussing whether to strengthen lock-in or push reform in the case of PD and lock-in. The debate is the label of Cluster 8, which includes response and historical events to clarify the relationship between the formulation of PD and lock-in and historical events, and is to discuss how to cross lock-in, and to achieve the reform and innovation. Therefore, Clusters 2 and 8 can be summarized as strengthening lock-in or reforming and stressing the importance of reform and innovation after considering their negative effects.
Cluster 7 belongs to P6 and is labeled by organizational PD. Tag words include strategic technological intervention and several themes, such as organizational inertia, organizational structure, organizational transformation, management, lock-in, unlock planning, strategic alliance, cooperation, strategic niche management, technology innovation system, and path creation of new technology. This cluster explores PD and lock-in of organizational structure, management, and technology due to the organizational inertia and then discusses how to cross lock-in by strategic alliance, cooperation, and strategic niche management.
Modeling is the label of P7, and it corresponds to Cluster 9 with only 5 cited references. It includes result dependent, result independent, distinguish PD, and measure. The mathematical programming model is constructed to simulate the prototyping characteristics of PD and lock-in. However, it is not sufficient to become the mainstream at this stage, but it provides ideas for future research. The following research hot topics also confirm it.
The evolution and changes of policy system (P1) and traffic development in urban areas (P3) have been widespread concerned at the first two stages. Customer lock-in (P2) is the continuation of consumer lock-in between 1985 and 2004. Case study is the continuation of the case study method between 2005 and 2008. Therefore, these contents are insufficient as research topics at this stage. Thus, lock-in or reform (P5), organizational PD (P6), and modeling (P7) became the research topics and trajectory at this stage. In addition, the terms of unlocking, crossing, and avoidance of PD and lock-in have been repeatedly mentioned and discussed.
According to the co-occurrence relationship path in Fig. 9, three distinct knowledge groups, the path lock-in, firm reforms, and political institutions, are identified. They are the research hotspots of PD and lock-in in the last five years. The inner threads of each knowledge group are connected by keywords, and the internal context is clear.
Keyword co-occurrence mapping in the area of PD and lock-in between 2013 and 2017.
In the knowledge group of path lock-in, the keywords with high betweenness centrality are lock-in, historical institutionalism, model, increasing returns, and cities. Thus, the group revolves around the phenomena of PD based on historical system and urban development, focuses on some questions such as the path of urban development and lock-in of historical system, and establishes the lock-in model based on the lock-in factors such as increasing returns. The concern is that the factors and modeling of path lock-in are important research topics over the last five years. Furthermore, the line color sparked by lock-in in the yellow and orange represents the maximum, and corresponds to 2016 and 2017. Path lock-in then becomes a hot topic and research frontier at present and even will be one in the future.
In the knowledge group of firm changes, the keywords with high betweenness centrality are firms, economic geography, dynamics, technology, industry, and organizations. This group is sparked by evolution, which includes economic geography changes, organizational changes, technological changes, industrial changes, and firm changes. Among them, technology research and development power, technology capacity, and technology diffusion belong to technological changes. In addition, research at the firm level is particularly prominent and becomes the essential connection point of mapping. The connection nodes are the network and transformation, that is, the research hotspots and key turning point of PD and lock-in over the last five years.
In the knowledge group of political institutions, the keywords with high betweenness centrality are growth, institutions, governance, policy, history, performance, Europe, path, creation, adaptation, China, and European Union. The research is around the institutions mainly including China, Asia, Europe, the European Union, and governance. Keywords such as labor, employment, historical sociology, and institutionalism are included in politics. The research on adaptation contains several keywords such as complexity, choice, performance, and path creation. In this group, the keywords with high frequency are path creation, governance, and performance. The group focuses on the PD of policies and institutions in the political development process, stresses governance and performance, and promotes institutional management, path creation, and innovation.
In conclusion, the research is on the core concept of PD. It explains some key issues such as innovation, evolution, path lock-in, and path creation in PD and lock-in from the last five years. These studies involve research on institution, technology, organization, policy, economic geography, politic, industry, and firm. The areas include China, Europe, and the European Union. From previous studies, historical institutions, economic geography, traffic development, and policy changes are the hotspots at each stage in the area of PD and lock-in. In addition, the changes at the level of firms, lock-in, path creation, and innovation, lock-in factors and models also act as hotspots with more room for future research.
A model of KM for the evolution of PD and path lock-in is proposed in this study. The study collected 328 references from the WoS database, performed data mining, drawn the knowledge mapping, and concluded the evolution in this area. The following conclusions are drawn:
English, American, and German scholars lead the research on PD and lock-in, and scholars are mostly from colleges and universities. The contribution of China does not stand out but grows steadily. Research topics have gone through four stages: the first stage of policy institution, customer lock-in, and technological change between 1985 and 2004, the second of case study method, regional economic, and environment technology lock-in between 2005 and 2007, the third stage of lock-in or reform, organizational PD, and modeling between 2008 and 2012, and the last one being the knowledge groups of path lock-in, firm changes, and political institutions. The research levels are from macro to micro-level with the trajectory of policy institution, regional economy, and organizational PD. Among them, national policies and political institutions are within the macro-level, the evolution of regional economy belongs to the meso-level, and the organizational transformation is in the micro-level. However, the research on micro-level is relatively few. The research design can be represented in theoretical basis, case discussion, lock-in or reform, and path creation. PD and lock-in theories are knowledge-based, whereas case deduction is the application. The topics of lock-in or reform and path creation extend the theory and are classified into the development research. The research method can be transformed from case study into lock-in mathematical modeling. The case study on PD and lock-in includes qualitative investigation. However, the mathematical model of lock-in contains a quantitative study. The empirical analysis based model would be particularly essential.
The study clarifies the evolution of PD and lock-in from five aspects, which are, research distribution, research topic changes, research level, research design, and research method. Furthermore, it explores the research hotspots and frontiers of the last five years and confirms the lack of micro-level research such as firms and the lack of research method such as empirical analysis. This study establishes a strong basis for researchers to grasp the development trend of the field, and it also inspires ideas for future research. In addition, due to the limitations of CiteSpace software, the data only comes from the SCI Core Collection database. If all the literature can be analyzed, a more comprehensive grasp of the search progress in the field can be achieved.
Footnotes
Acknowledgments
The authors acknowledge the National Key Technology Research and Development Program of the Ministry of Science and Technology of China (Grant: 2015BAG08B04), Soft Science Research Program of Science and Technology Department of Hubei Province, China (Grant: 2017ADD009).