ABSTRACT. International academic collaborations gain from face-to-face interactions. We take advantage of a quasi-experiment that the construction of sino-US nonstop flights (NF) to test the influence on academic co-publication. Based on the perspective of knowledge spillover and face-to-face contact theory, results show that researchers enjoy a productivity boost both in quality and quantity, using data set of highly-cited papers (2009-2018) in the Web of Science and DID method. Furthermore, it expounds that participating in international academic conferences and reporting papers is a main channel to promote high-level transnational scientific co-publication. Besides, large amount of robustness checks such as time lag effect and the treatment intensity effect show that policy effect of NF in promoting scientific cooperation remains robust. Last, we discuss the relationship between international transportation and scientific and technological collaboration.

ABSTRACT. National scientific research cooperation plays an important role in the development of national scientific research capabilities. Quantitative analysis of scientific cooperation between countries can provide references for guiding the country to select scientific research partners. Based on the 356,405 papers on the AMiner platform from 1953 to 2010, we analyze the collaboration patterns and its evolution of papers between 51 countries. From the perspectives of national scientific cooperation network, considering paper output and academic influence, we compare cooperation network centrality, collaboration intensity, cooperation relationship selection and other indicators to analyze the relationship of scientific cooperation. Several interesting patterns are observed: (1) the scientific cooperation among countries is very frequent, but most of the cooperative relations between them are weak; (2) there are some special super ties between countries; (3) super ties occur at an average rate of 1 in 7 collaborators; (4) the average cooperation strength of high-yielding countries is generally higher than that of low-yielding countries; (5) when high-yielding countries become strong relations with low-yielding countries, it can greatly drive the growth of scientific research output of high-yielding countries. Then, the regression method is used to measure the impact of super ties on the publication rate and citations of countries, controlling pertinent factors such as prestige and cooperation scale. The regression results indicate that super ties increase the productivity and citations of national papers, confirming that super-tie is an important driving factor in the development of science.

ABSTRACT. Artificial intelligence (AI) has brought significant revolutions to human society. This paper reveals the evolution of collaborative scholarship among researchers of AI from 1995 to 2019. It explores collaborative pattern, co-authorship networks in three perspectives (individual, institutional and international) and changing status of countries within collaboration. In total, 49,320 publications were collected and divided into 5 groups from which co-authorship networks were abstracted. Results show that 1) collaborations expanded individually, institutionally and internationally. During AI’s rapid recovery era (from 1993 to 2010), most of the collaborations were conducted within one institution in one country. During AI’s boom era (from 2011 to 2019), large and even mega teams (some teams have more than 20 authors) became prominent, collaborations between two institutions were the most popular collaboration pattern, and multi-national collaboration became a reality. 2) The breadth of authorial collaboration expanded, the collaboration coalesced, and the nationalities of collaborating authors diversified during the study period. United States and United Kingdom prove to be the most important countries in collaboration. China and Switzerland are two countries with rapidly growing significance of collaboration, especially China.

ABSTRACT. This article takes the field of stem cells as an empirical case to explore the evolutionary model of the scientific community and the law of the author's influence. (1) Taking 5 years as a time slice, a sliding window method is used to construct scientific cooperation networks, and more than ten complex network indicators are used to measure the overall structural evolution of the networks. (2) We measure the influence of authors in the cooperation networks with indicators such as degree centrality, closeness centrality, and betweenness centrality, then analyze the growth patterns of authoritative experts, domestic experts, and academic new stars. (3) The Leiden algorithm is used to extract the community structure in the cooperation network, and the community structure model is described according to the author's influence difference. We use indicators such as continuity and variability to measure community similarity, and then quantitatively analyze several evolutionary modes of community: expansion, contraction, division, merger, and personnel flow. (4) The influence change of scientific researchers will drive the evolution of the scientific team, and the changes in the scientific team will cause the influence of scientific researchers to increase or decrease. This article explores the concomitant relationship between the influence change of researchers and the evolution of scientific teams and hopes to provide constructive suggestions for the construction and development of scientific teams.

ABSTRACT. In the process of the development of technology market in the late developing countries, large-scale science and technology infrastructure will bring new possibilities and continuous influence to science and technology innovation in the industry. Taking Beidou satellite navigation system as an example, this paper studies the changes of innovation activities of enterprises, universities and other scientific research subjects in different policy environments based on Beidou System in 2004-2018. We have collected dozens of Beidou industrial support policies with different influence areas and implementation objects, and divided them into two types of policy tools through content analysis: support output policies and encourage cooperation policies. Any innovation subject is located in a different location and receives different policy support. In the analysis, we use the data in patent database to select 2430 patent applicants from different countries, and analyze the characteristics of patent citation network. The preliminary empirical results show that: (1) the promotion effect of large-scale science and technology infrastructure on industrial technology depends on the diversity of participants, and the more types of participants, the more significant the promotion effect; (2) the policy of encouraging cooperation has a significant impact on the centrality of technology network, and the national policy has a greater impact than the regional industrial policy; (3) the policy of encouraging cooperation It is more effective than direct support policy, which may be because technology cooperation network is still the main driving force in the continuous progress of technology.

ABSTRACT. Research hotspots reveal academic frontiers and display the most innovative scientific results. To promote the development of a discipline and avoid repetitive researches, it is essential for researchers to acquire valuable and up-to-date researches timely. Special issue(SI) acts as "vanguard(s) of knowledge that creates a path into new topics", and the editorial summary of it is a highly condense material that may indicate the development frontiers of disciplines. This research is intended to find a new way for research hotspots identification through the editorial summary of SI. The bibliographic data of the editorial material of SI is searched and collected from Web of Science. In order to reveal the research hotspots, the research fields of SI is demonstrated using science overlap mapping and cortext river graph. Their topicality evolution is explored by topic modeling and term clumping. The research motivation of SI is also analyzed with the help of text extraction and named-entity recognition. There is an increasing tendency in the publication of SI, especially in the past decades, indicating that more journals notice and admit the importance of SI. Many SI articles have been published in the fields of computer science and engineering, followed by material science and environmental science and technology. This paper demonstrates a holistic view of the development trends of SI publication activity in all disciplines. By exploring the essential roles of SI, it offers new insights into the measurement of research hotspots and frontiers.

ABSTRACT. Latent Dirichlet Allocation (LDA) has become the cornerstone for probabilistic topic modeling of text collections. In the context of science analysis and scientific policy design and monitoring, these tools have been widely used over the last few years to model topic evolution, detect emerging topics, or to analyze lead-lag between data sets. Many datasets related to Science, Technology and Innovation (STI) (scientific articles, patent applications, funding proposals, etc) have been analyzed with these tools and some examples can be found in the last editions of the Global Tech Mining Conference. Apart from offering a thematic overview of a document collection, topic models can also be used to obtain an intermediate representation that can later be used to train automatic classifiers (with respect to specific taxonomies) or to build semantic graphs. However, topic models depend on a set of algorithm hyper-parameters, including in many cases the number of topics, and their selection may significantly affect the quality of the models. Some common procedures to select them rely on coherence definitions and subjective evaluation. In this work, we propose to exploit document graphs based on available metadata for hyperparameter selection, and compare this strategy with topic coherence and topic model stability approaches. Our results on several STI-related data sets show that these strategies provide relevant indicators to build high-quality topic models.

ABSTRACT. Transformative research topics (TRTs) can subvert the original concept of disciplines and technologies, redefine their connotation, and ultimately lead to the revolution of discipline paradigm and technological trajectory. The evolution of TRT usually requires a long accumulation time, therefore its early discovery is more conducive to strategic planning in technology R&D and industrial layout, but the early identification is difficult due to the lack of accumulated information. Therefore, how to identify its warning signals at an early stage, and evaluate its transformative potential in the future, is challenging but critical. This study aim to identify and predict TRTs based on catastrophe theory for general purpose, explore the mechanism of discontinuous mutation in the continuously evolution system and the relationship between continuous changing factors. Through the measurement of knowledge potential during the topic evolution, we can finally achieve early detection and monitoring to TRTs. This paper identifies TRTs with eleven indicators from three dimensions: topic growth rate, potential societal and economic influence, reduction trend of uncertainty and ambiguity. An empirical research in stem cells proves there exist differences in mutation potential in different emerging topics. The TRTs detection model based on catastrophe theory can reflect the transformative patterns of the system, is capable to identify and predict prediction TRTs. The method we propose can identify the transformative potential of emerging topics, contribute to the scientific and technological planning and management.

ABSTRACT. Existing Chinese technical terminology extraction methods focus on the high-frequency characteristics, while technical domain correlation characteristic and the unithood indicator of terminology are paid less attention. Aiming at these problems, this paper introduces the background corpus into C-value method, proposes the concept of word frequency distribution rate of technical terminology, and constructs the DC-value method to effectively solve the problem of technical domain correlation characteristics. At the same time, the information entropy is introduced to compute the left and right information entropy of strings, measure the uncertainty of strings’ left and right borders, and meet the unithood indicator of technical terminology. In addition, the structural characteristics and lexical rules are obtained through the analysis of a great number of Chinese patent corpuses, to filter the candidate terms and ensure the accuracies of terminology. Experiments are done to extract technical terminology in the domain of information and communication, with the Chinese patent literature in the domain of information and communication as domain corpus and the ones in the domain of electric vehicle as background corpus. The results show that the presented algorithm can effectively extract the technical terminology in Chinese patent literatures, and have a better precision than the log-likelihood ratio method and the mutual information method.

ABSTRACT. Understanding the evolution of research topics is essential to discover new trends in science and to grasp the background of a field. However, due to information overload, aging acceleration and increasingly fierce competition in science and technology, the identification process is full of difficulty and low accuracy. This paper proposes a new method of topic evolution discovery based on dynamic network and dynamic community, and analyzes the way of topic evolution. First of all, based on the piecewise linear representation (PLR), this paper finds the turning point of research trend in the whole research field to divide the time period. At the same time, based on word2vec word vector model, this paper constructs a dynamic keyword relationship network. Secondly, this paper identifies communities based on random walk community discovery algorithm and uses Z-score method to find the research topic. Finally, by measuring the similarity between topics in different time periods, we can identify the evolution relationship of topics and analyze the evolution path of topics. The effectiveness of this method is verified by empirical research in the field of information science.

ABSTRACT. This paper develops an approach to map knowledge innovation ecosystems. As an empirical test case, the paper identifies and charts the presence of regionalized knowledge ecosystems in the Artificial Intelligence (AI) sector in Canada and the UK. Methodological and policy insights are drawn from this case. The paper will develop an analytical framework to examine knowledge innovation ecosystems in emerging technologies and use bibliometric, patent and social network analysis methods to investigate and compare AI knowledge innovation ecosystem in the UK and Canada. We will use a newly-developed and up-to-date AI bibliometric search approach. This approach, which incorporates related AI technologies such as machine learning, neural nets, and deep learning, is used for AI publication and (adjusted by relevant classifications) patent searches. The paper will map who participates in AI research and innovation, identifying regional clusters and key research and innovation participants. It will then probe the extent to which these clusters form knowledge innovation ecologies, examining the role of internal and external knowledge linkages and networks. Knowledge relationships between key actors, including R&D institutions, companies, and other organizations will be probed, including roles of proximate and non-proximate organizations. The dynamics of change over time will be analyzed.

ABSTRACT. The role of science and patenting is often viewed as the focus of SMEs in post catch-up conditions, as they seek to pivot from performing low value-added activities to new technological ventures in searching for new niches. SMEs – particularly those in city areas – are incentivized to commit to research that is linked to scientific knowledge and patenting activities. This study is interested to explore whether these SMEs upgrade, patent and commit to long term upgrading. Their performances are benchmarked to SMEs of Silicon Valley (the highly cited Marshallian industrial cluster). We configured an extraction process for bulk patenting data and architected a sorting procedure to derive a list of relevant indexes from patents assigned to the bottom 40 (and 60) percent of the total assignees of a region. We discovered that SMEs in Taipei, Seoul, Singapore, Tel Aviv, Hong Kong and Dublin had indeed upgraded to adopt emerging (science-based) technologies and appropriate them – both to supply for existing market needs, and to build their competitive edge for future endeavours. While many cities were found to be relatively ahead in producing technologies that are classified as long cycle and science-based, Taipei saw a rising number of SMEs committed to science amid the majority which invested in non-science technologies. Taipei nonetheless stands out as the city which bears many characteristics of what a Marshallian cluster is like. This study sheds new light on the pursuit of SMEs in post catch-up conditions.

ABSTRACT. A new round of scientific and technology revolution, and the corresponding industrial upgrading meets the new chance and challenges. Such new technologies as block chain, big data and deep learning and so forth, are constantly emerging, which have a profound impact on economic and social development. To forecasting the technology opportunities and development trends has become an important issue which affects the competitive strategies of countries, companies and research institutions etc.

This paper proposes a hybrid approach based on topic modeling, k-mean clustering and patent analysis to forecast the technology opportunities, and the blockchain is selected as a case study.

In the case study, firstly, we retrieve and download the patent data from the USPTO database. In text preprocessing, we focus on the Abstract and Title fields. Then, we divide the cleaned data into several sub-datasets according to the time sequences. Secondly, we apply Latent Dirichlet Allocation (LDA) model to each sub-dataset, which is a kind of text mining method to extract the technical topics appearing in patent data. Thirdly, we propose an clustering algorithm based on several weighting indicators to analyze the technology distributions. Finally, based on the technical analyses of blockchain, we analyze and forecast the development trends of the blockchain patents in each path(cluster). This paper attempt to contribute to forecast technology opportunities for a specific new/emerging technology, as well as to further understand their emergence and development.

ABSTRACT. New societal trends are currently unfolding, such as digitalisation, sharing economy and changing consumer awareness. These trends might highly influence future energy demand and depending on their realisation might enhance or counterbalance projected energy efficiency gains. This study is an attempt to quantitatively identify and verify new societal and technology trends that are likely to shape future energy demand in the European countries, using tech mining tools (Vantage Point).

The methodological approach is based on qualitative (literature review, expert consultations) and quantitative (bibliometric analysis, technology mining) analysis. It includes the following stages:

Step 1. Qualitative analysis of trends using literature review and expert consultations. Step 2. Quantitative analysis of WoS publications (bibliometric analysis and technology mining with Vantage Point and VOSviewer). Step 3. Combination of qualitative and quantitative results and their analysis.

As a result of this study, the relative assessment of new societal and technology trends influencing the development of energy sector in the European countries was conducted using quantitative data processing and expert procedures and the most relevant trends were analysed (f.e. health importance, sharing economy and prosumer models, human-machine interfaces, awareness of personal footprint, circular economy and decarbonisation, etc.). The results obtained can be taken as a guide by researchers, business representatives or energy policy makers.

ABSTRACT. An effective selection of partners is a core factor affecting the collaborative innovation performance, and many guidelines or references have been developed for this purpose. Some studies propose quantitative indicators using the bibliographic information or considering the subject correlation of papers or patents, and some others use the network analysis method. Our previous study developed a systematic selecting framework based on the highly correlated connotation of science and technology innovation through topic analysis of both papers and patents, to achieve the expected synergy and avoid vicious competition. Here, we extend it by focusing on those organizations that have both papers and patents (Paper-patent-Org) and integrating link prediction of cooperative network, to provide more references for partner selection decision. We proposed an extended framework including 7 steps, the first three steps were from our previous study, and the next four steps included “Constructing collaboration network of papers’ or patents’ organizations respectively”, “Identifying Paper-patent-Org and constructing the fusion network of paper and patent organizations by coalescing the data properly”, “Using the link prediction method on the fusion network”, and “comprehensively considering the results of analysis of previous steps to recommend collaborative innovation partners”. Empirical analysis demonstrated the feasibility and advantage of this fusion method; however, the verification and application of the method need to be carried out on larger datasets of a greater number of years.

ABSTRACT. Identifying potential emerging technologies and business models are crucial for the industry to gain first-mover advantages in the market competition and enhance their competitiveness, especially for startups. In the meanwhile, startups also play a vital role in boosting the emergence of technologies and business models. Compared with analyzing publication and patent databases, startup profiles can provide the first-hand technologies and business models in practice, which bridges the gap from research and development (R&D) to the market. This paper proposes the framework of emerging technology and business model identification based on keyword extraction from startup profiles, which validates the reciprocity between technology mining and entrepreneurship monitoring based on 240 startups of German energy sector. Furthermore, it proves the possibility that unstructured texts from webpages, reports and LinkedIn can supplement technology mining and the evolution identification of technologies as well as business models. Besides, this paper presents the technology and business model evolution during German energy transition.

ABSTRACT. This paper presents a systematic literature review on organizational capabilities in the project business and assesses publications identified from the Web of Science Core Collection from a thematic perspective using a temporal aspect. Briefly, organizational capabilities in the business context refer to a firm’s ability to do something. This study approached the literature systematically through bibliographic coupling clustering and term analysis. Based on a carefully selected set of 93 papers published between 1993 and 2019, bibliometric analysis confirmed a steady increase in publishing activity in the past ten years. Bibliographic coupling visualization with VOSviewer software resulted in eight article clusters. The thematic analysis was conducted based on the titles and abstracts of the articles in each cluster. The term analysis features of ATLAS.ti software were applied in theme identification. Based on the cluster sizes, the three themes attracting the most attention were 1) Capabilities development in innovation and knowledge management, 2) Dynamic capabilities and project management, and 3) Organizational capabilities and IT systems. Temporal analysis demonstrated the publishing activity per cluster and visualized the prevalence of the themes in a timeline, indicating how the number of publications has changed over time in certain themes. The paper contributes to thematic outlining of the literature by presenting a methodological approach to combine bibliometrics and term analysis. Organizational capability research benefits from a systematically identified and visualized thematic structure. The paper suggests avenues for future research, including how to explore the effects on results of including full article texts in term analysis.

ABSTRACT. This paper presents a tech mining method adapted to the technological intelligence requests detected in the author’s collaboration with firms. Certain industries (power electronics, e.g.) produce complex systems formed by particular configurations of materials and/or devices (MD), and the performance of the systemoften depends on the successful incorporation of such MDs in the solution offered to the customer. We developed a method adapted to such needs for the automated detection of the MDs in patent data, together with a system for positioning the MDs in the Technology Life Cycle (TLC) curve. The method is complemented with a subsampling strategy aimed at providing the Science & Technology (S&T) decision maker with the patent subsamples that show evidences of concentrating their inventive capacity on the MDs that qualify as “emergent” in the TLC model. Our method adds value to technology watch professionals in sectors where incremental innovations of the product depend on achieving a succesful recombination of MDs that could easily go under the radar, particularly in their emergent stages.

ABSTRACT. Virtual reality (VR) is a highly integrated technology based on computer, which combines sensor technology, high-definition simulation technology, 3D modeling technology and multi-disciplinary knowledge. There are three stages of patent applications for virtual reality. The purpose of this study is to address the development trend of virtual reality technology by ARMA model with the quoted frequency of virtual reality patent applications per year. This paper mines patent data of virtual reality technology form DII. In order to ensure integrity and accuracy, the data was retrieved with the manual code and the subject from 1963 to 2018 on August 13, 2019. The retrieval results were 18,588 patent families. After data cleaning and manual denoising, the missing or duplicate patents of data value were deleted. And the data of 18,321 patent families was adopted for the research. According to the total number of virtual reality technology patents cited annually, ARMA model is used for time series analysis in this paper. On the basis of patent citation analysis, this paper recognizes the importance of quoted frequency and applies the data to ARMA modeling. It is verified that the valid model is ARMA (2,1) which can conduct forecast effectively.

ABSTRACT. The research of technological speciation catches the interest of several authors (Adner und Levinthal 2002; Moehrle und Caferoglu 2019). As a counterpart of scientific-driven technologies, technological speciation sheds light on the phenomenon of demand-pull technologies. Inspired by evolutionary biology, technological speciation argues that new technologies might be a result of an adaption process, in which a speciation technology branches-off from a mainstream technology, in order to fulfill unsatisfied customer needs reflected in application niches. Moehrle and Caferoglu (2019) present a first semantic patent analysis approach, which identifies technological speciation candidates by measuring attributes such as novelty, growth, persistence and community. Although this approach seems to be promising, it has only been developed and tested for a B2C market. Since there is a lack of further studies testing this method, our research aims to find out if and how the methodical approach is applicable for a B2B market. In this study, we test the methodical approach for the technology of electric drives. Doing so, we identify several application fields such as automotive, flight vehicles, medical technology, military or gaming. For each of those application fields, we identify several technological speciation candidates, for instance, e-vehicle, e-bike, hoverboard, drone, surgical instrument. Our approach delivers some theoretical as well as managerial implications as it helps to understand the evolution of electric drive technology and shows the usability of the methodical approach for a B2B technology.

ABSTRACT. The explosive growth of research literature pertaining to COVID-19 presents analytical challenges. This paper spotlights approaches to address those.

ABSTRACT. The COVID-19 pandemic presented a challenge to the global research community as they rushed to find solutions to the devastating crisis. Drawing expectations from resilience theory, this paper explores how the trajectory of coronavirus research was affected by the COVID-19 pandemic. Using terms featured in articles in early COVID-19 research, and pathways of knowledge characterized through term extraction, evolutionary pathways and statistical analysis, the results reveal that the pandemic disrupted existing lines of coronavirus research to a large degree. While broad communities of coronavirus research are similar pre- and during COVID-19, topics themselves change significantly and there is less cohesion amongst early COVID-19 research compared to that within research before the pandemic. We find that some lines of research revert to research pursued almost a decade earlier, whilst others pursue brand new trajectories, and that Chinese researchers in particular appear to be driving the more novel research. These findings suggest that the global research community maintained a similar broad research focus from before the pandemic, but shifted the detail of the research in the COVID-19 period, exploring diverse pathways, and that specific actors are more flexible than others in doing so. The findings raise further questions about whether the shifts are advantageous for global scientific progress, and whether the research community will return to the original equilibrium or reorganize into a different knowledge configuration.

ABSTRACT. The Covid-19 pandemic has spurred a global race for therapies and vaccines; but drug development is notoriously uncertain and prolonged. With cases increasing exponentially, we need a way to decrease the time it takes for actors to search the network of biomedical knowledge for optimal solutions. Unfortunately, we have a limited understanding of search processes in drug development, perhaps because the traditional methods used to measure knowledge—cited references—are unreliable when applied to clinical trials. Here we demonstrate a new application of the NLM Medical Text Indexer which allows us to link clinical trials, biomedical research and the actors’ internal knowledge through the MeSH vocabulary into a common network. We use event history models that take advantage of the pandemic’s properties as a natural experiment and the small world structure of the MeSH network to identify how search availability and problem distance affect the time it takes for each MeSH term to be linked to a Covid-19 clinical trial. We then use the phase outcomes of those trials to understand how different search strategies affect the probability of success. This paper makes contributions to scientometric measurement, organizational learning on networks and the ongoing policy response to the Covid-19 pandemic.

ABSTRACT. Since the beginning of 2020, many tens of thousands of academic papers about COVID-19 have been published and hundreds of new papers continue to be published every day. This incredible rate of scientific productivity leads to information overload, making it difficult for researchers, clinicians, and public health officials to keep up with the latest findings. Automated text mining techniques for finding and summarizing papers are promising strategies for addressing information overload. In this talk, we survey the landscape of existing resources and systems that have been developed for text mining over COVID-19 papers. In particular, we highlight the COVID-19 Open Research Dataset (CORD-19), a growing resource of scientific papers on COVID-19 and related historical coronavirus research. CORD-19 is designed to facilitate the development of text mining systems over its rich collection of metadata and structured full text papers. Since its release, CORD-19 has supported the development of dozens of publicly-available systems aimed at helping biomedical experts and policy makers search for effective treatments and management policies for COVID-19. Finally, we discuss how CORD-19 has enabled the rapid organization of several shared tasks, which have pooled community resources to collect valuable data to support development and evaluation of these systems.