Yi Bu

Understanding scientific collaboration: Homophily, transitivity, and preferential attachment

Scientific collaboration is essential in solving problems and breeding innovation. Coauthor network analysis has been utilized to study scholars collaborations for a long time, but these studies have not simultaneously taken different collaboration features into consideration. In this paper, we present a systematic approach to analyze the differences in possibilities that two authors will cooperate as seen from the effects of homophily, transitivity, and preferential attachment. Exponential random graph models (ERGMs) are applied in this research. We find that different types of publications one author has written play diverse roles in his/her collaborations. An authors tendency to form new collaborations with her/his coauthors collaborators is strong, where the more coauthors one author had before, the more new collaborators he/she will attract. We demonstrate that considering the authors attributes and homophily effects as well as the transitivity and preferential attachment effects of the coauthorship network in which they are embedded helps us gain a comprehensive understanding of scientific collaboration.

Understanding persistent scientific collaboration

Common sense suggests that persistence is key to success. In academia, successful researchers have been found more likely to be persistent in publishing, but little attention has been given to how persistence in maintaining collaborative relationships affects career success. This paper proposes a new bibliometric understanding of persistence that considers the prominent role of collaboration in contemporary science. Using this perspective, we analyze the relationship between persistent collaboration and publication quality along several dimensions: degree of transdisciplinarity, difference in coauthors scientific age and their scientific impact, and research‐team size. Contrary to traditional wisdom, our results show that persistent scientific collaboration does not always result in high‐quality papers. We find that the most persistent transdisciplinary collaboration tends to output high‐impact publications, and that those coauthors with diverse scientific impact or scientific ages benefit from persistent collaboration more than homogeneous compositions. We also find that researchers persistently working in large groups tend to publish lower‐impact papers. These results contradict the colloquial understanding of collaboration in academia and paint a more nuanced picture of how persistent scientific collaboration relates to success, a picture that can provide valuable insights to researchers, funding agencies, policy makers, and mentor–mentee program directors. Moreover, the methodology in this study showcases a feasible approach to measure persistent collaboration.

Understanding scientific collaboration from the perspective of collaborators and their network structures

Scientific collaboration is one of the key factors to trigger innovations. Coauthorship networks have been taken as representations of scholars’ collaboration for a long time. This study investigates how the authors’ attributes and the coauthorship network structures simultaneously influence the scientific collaboration among them. Exponential random graph models (ERGMs) are adopted in this research. We find that an author has a propensity to coauthor with the other scholar if they have different levels of productivity. We also find that the effect of network’s transitivity strongly influence authors’ collaboration. We demonstrate that taking the effects from both authors’ attributes and the network structures into consideration helps gain a comprehensive understanding of scientific collaboration.

Understanding success through the diversity of collaborators and the milestone of career

Scientific collaboration is vital to many fields, and it is common to see scholars seek out experienced researchers or experts in a domain with whom they can share knowledge, experience, and resources. To explore the diversity of research collaborations, this article performs a temporal analysis on the scientific careers of researchers in the field of computer science. Specifically, we analyze collaborators using 2 indicators: the research topic diversity, measured by the Author‐Conference‐Topic model and cosine, and the impact diversity, measured by the normalized standard deviation of h‐indices. We find that the collaborators of high‐impact researchers tend to study diverse research topics and have diverse h‐indices. Moreover, by setting PhD graduation as an important milestone in researchers careers, we examine several indicators related to scientific collaboration and their effects on a career. The results show that collaborating with authoritative authors plays an important role prior to a researchers PhD graduation, but working with non‐authoritative authors carries more weight after PhD graduation.

Functional structure identification of scientific documents in computer science

The increasing number of open-access full-text scientific documents promotes the transformation from metadata- to content-based studies, which is more detailed and semantic. Along with the benefits of ample data, the confused internal structure introduces great difficulties to data organization and analysis. Each unit in scientific documents has its own function in expressing authors’ research ideas, such as introducing motivations, describing methods, stating related work, and drawing conclusions; these could be used to identify functional structure of scientific documents. This paper firstly proposes a clustering method to generate domain-specific structures based on high-frequency section headers in scientific documents of a domain. To automatically identify the structure of scientific documents, we categorize scientific documents into three types: (1) strong-structure documents; (2) weak-structure documents; and (3) no-structure documents. We further divide the identification into three levels—section header-based identification, section content-based identification, and paragraph-based identification—corresponding to the three types of documents. Our experiments on documents in the field of computer science show that: (1) section header-based identification is the most direct and simplest method, but its accuracy is limited by unknown words in section headers; (2) section content-based identification is more stable and obtains good performance; and (3) paragraph-based identification is promising in identifying functions of no-structure documents. Additionally, we apply our methods to two tasks: academic search and keyword extraction. Both tasks demonstrate the effectiveness of functional structure.

Measuring the stability of scientific collaboration

Stability has long been regarded as an important characteristic of many natural and social processes. In regards to scientific collaborations, we define stability to reflect the consistent investment of a certain amount of effort into a relationship. In this paper, we provide an explicit definition of a new indicator of stability, based on the year-to-year publication output of collaborations. We conduct a large-scale analysis of stability among collaborations between authors publishing in the field of computer science. Collaborations with medium–high degree of stability tend to occur most frequently, and on average, have the highest average scientific impact. We explore other “circumstances”, reflecting the composition of collaborators, that may interact with the relationship between stability and impact, and show that (1) Transdisciplinary collaborations with low stability leads to high impact publications; (2) Stable collaboration with the collaborative author pairs showing greater difference in scientific age or career impact can produce high impact publications; and (3) Highly-cited collaborators whose publications have a large number of co-authors do not keep stable collaborations. We also demonstrate how our indicator for stability can be used alongside other similar indicators, such as persistence, to better understand the nature of scientific collaboration, and outline a new taxonomy of collaborations.

Innovation or imitation: The diffusion of citations

Citations in scientific literature are important both for tracking the historical development of scientific ideas and for forecasting research trends. However, the diffusion mechanisms underlying the citation process remain poorly understood, despite the frequent and longstanding use of citation counts for assessment purposes within the scientific community. Here, we extend the study of citation dynamics to a more general diffusion process to understand how citation growth associates with different diffusion patterns. Using a classic diffusion model, we quantify and illustrate specific diffusion mechanisms which have been proven to exert a significant impact on the growth and decay of citation counts. Experiments reveal a positive relation between the “low p and low q” pattern and high scientific impact. A sharp citation peak produced by rapid change of citation counts, however, has a negative effect on future impact. In addition, we have suggested a simple indicator, saturation level, to roughly estimate an individual articles current stage in the life cycle and its potential to attract future attention. The proposed approach can also be extended to higher levels of aggregation (e.g., individual scientists, journals, institutions), providing further insights into the practice of scientific evaluation.

Understanding the formation of interdisciplinary research from the perspective of keyword evolution: a case study on joint attention

Understanding the formation of interdisciplinary research (IDF) is critically important for the promotion of interdisciplinary development. In this paper, we adopt extracted keywords to investigate the features of interdisciplinarity development, as well as the distinct roles that different participating domains play in various periods, and detect potential barriers among domains. We take joint attention (JA) as the study domain, since it has undergone a development process from a topic of one domain to interdisciplinary research (IDR). Our empirical study has yielded interesting findings. First, we detect the phenomenon of knowledge diffusion as it evolved through three domains of JA. It enabled us to observe the shift of roles the domains played during the process of IDF, as well as the existence of potential barriers among these domains. Second, according to the diffusion and development process of JA among domains, three phases that an IDR field in general goes through were identified: a latent phase, an embryo phase, and a mature phase. Third, domains may play different roles in distinct periods, with the formation of IDR. Four roles are identified: knowledge origin, knowledge receiver, knowledge respondent, and interdisciplinary participant. This paper showcases how to detect the evolution of IDR by analyzing keyword evolution. By giving the profiles of IDR fields and descriptions of keyword evolution, it would be valuable for policy makers and regulators to promote IDR development.

A quantitative exploration on reasons for citing articles from the perspective of cited authors

Citation is regarded as one of the “norms of science” (Merton in Am Sociol Rev 22(6):635–659, 1957) and is deeply researched by the field of scientometrics. The motivations authors have for citing one another are considered significant and have been the subject of extensive qualitative research such as content analysis, questionnaires, and interviews of citing authors. However, the existing qualitative studies have covered a limited number of samples. To expand the dataset, this paper proposes a quantitative method applied to detecting citation reasons from the angle of citation networks and the attributes of cited authors, including their publication count (the number of single-authored publications, collaborative and first-authored publications as well as collaborative but non-first-authored publications, and number of whole publications), citation count, research topic interests, and gender. By applying the Exponential Random Graph Models (ERGMs), the current study revealed that authors in the field of information retrieval tend to cite those with more single-authored, collaborative and first-authored, and collaborative but not first-authored publications. Besides, in this field, the number of publications, similar topical domains, and same gender are proven to be significantly favorable in selecting references in our experiment.

Using the appearance of citations in full text on author co-citation analysis

As a frequently used method of depicting scientific intellectual structures, author co-citation analysis (ACA) has been applied to many domains. However, only count-based information is involved as the input of ACA, which is not sufficiently informative for knowledge representations. This article catches several metadata in full text of citing papers but not aims at content-level information, which increases the amount of information input to ACA without increasing computational complexity a lot. We propose a new method by involving information including the number of mentioned times in a citing paper and the number of context words in a citing sentence. We combine these pieces of information into the traditional ACA and compare the results between ACA and the proposed approach by using factor analysis, network analysis, and MDS-measurement. The result of our empirical study indicates that compared with the traditional ACA, the proposed method shows a better clustering performance in visualizations and reveals more details in displaying intellectual structures.