Scientometrics最新文献

Funding acknowledgments are important objects of study in the context of science funding. This study uses a mixed-methods approach to analyze the funding acknowledgments found in 2.3 million scientific publications published between 2008 and 2021 by authors affiliated with research institutions in the Middle East and North Africa (MENA). The aim is to identify the major funders, assess their contribution to national scientific publications, and gain insights into the funding mechanism in relation to collaboration and publication. Publication data from the Web of Science is examined to provide key insights about funding activities. Saudi Arabia and Qatar lead the region, as about half of their publications include acknowledgments to funding sources. Most MENA countries exhibit strong linkages with foreign agencies, mainly due to a high level of international collaboration. The distinction between domestic and international publications reveals some differences in terms of funding structures. For instance, Turkey and Iran are dominated by one or two major funders whereas a few other countries like Saudi Arabia showcase multiple funders. Iran and Kuwait are examples of countries where research is mainly funded by domestic funders. The government and academic sectors mainly fund scientific research in MENA whereas the industry sector plays little or no role in terms of research funding. Lastly, the qualitative analyses provide more context into the complex funding mechanism. The findings of this study contribute to a better understanding of the funding structure in MENA countries and provide insights to funders and research managers to evaluate the funding landscape.

Online academic profiles are used by scholars to reflect a desired image to their online audience. In Google Scholar, scholars can select a subset of co-authors for presentation in a central location on their profile using a social feature called the “co-authroship panel”. In this work, we examine whether scientometrics and reciprocality can explain the observed selections. To this end, we scrape and thoroughly analyze a novel set of 120,000 Google Scholar profiles, ranging across four dieffectsciplines and various academic institutions. Our results seem to suggest that scholars tend to favor co-authors with higher scientometrics over others for inclusion in their co-authorship panels. Interestingly, as one’s own scientometrics are higher, the tendency to include co-authors with high scientometrics is diminishing. Furthermore, we find that reciprocality is central in explaining scholars’ selections.

Billions of scientific papers lead to the need to identify essential parts from the massive text. Scientific research is an activity from putting forward problems to using methods. To learn the main idea from scientific papers, we focus on extracting problem and method sentences. Annotating sentences within scientific papers is labor-intensive, resulting in small-scale datasets that limit the amount of information models can learn. This limited information leads models to rely heavily on specific forms, which in turn reduces their generalization capabilities. This paper addresses the problems caused by small-scale datasets from three perspectives: increasing dataset scale, reducing dependence on specific forms, and enriching the information within sentences. To implement the first two ideas, we introduce the concept of formulaic expression (FE) desensitization and propose FE desensitization-based data augmenters to generate synthetic data and reduce models’ reliance on FEs. For the third idea, we propose a context-enhanced transformer that utilizes context to measure the importance of words in target sentences and to reduce noise in the context. Furthermore, this paper conducts experiments using large language model (LLM) based in-context learning (ICL) methods. Quantitative and qualitative experiments demonstrate that our proposed models achieve a higher macro F₁ score compared to the baseline models on two scientific paper datasets, with improvements of 3.71% and 2.67%, respectively. The LLM based ICL methods are found to be not suitable for the task of problem and method extraction.

Beyond citations, the impact of scientific publications is often measured by usage metrics, such as downloads, save counts and sharing counts. However, the motivations behind the utilization of these publications and their influencing factors have not yet been well studied. Therefore, it remains questionable whether and to what extent usage metrics can reflect the impact of publications. Based on expectancy-value theory, the aim of the present study was to examine the differences in behavioral characteristics and driving factors between article downloading, sharing, and saving, especially document characteristics. For the present study, survey data from 480 respondents across Chinese universities were collected and investigated in terms of the frequency and purpose of three literature usage behaviors, namely, downloading, sharing, and saving. Additionally, 11 document characteristics were used to construct three variables in the research models: intrinsic interest value, attainment value, and utility value. Their effects on three usage behaviors were examined based on path analysis via SmartPLS. The results showed that the overall frequency of article downloading and saving was greater than that of article sharing. The primary purposes of downloading and saving were closely related to scientific research, such as for review and citing. The sharing of articles on social media was mainly for agreeing with their opinions. Both intrinsic interest value and utility value exhibited a significant positive influence on article-downloading, whereas attainment value and intrinsic interest value showed a significant relationship with sharing and saving, respectively. In conclusion, different literature usage behaviors can be triggered and driven by the distinct values of research articles. The results obtained in this study could help to clarify the determinants of different usage behaviors; additionally, they might promote the reasonable application of usage metrics or altmetrics in scientific evaluation.

Author networks in science often rely on citation analyses. In such cases, as in others, network interpretation usually depends on supplementary data, notably about authors’ research domains when disciplinary interpretations are sought. More general social networks also face similar interpretation challenges as to the semantic content specificities of their members. In this research-in-progress, we propose to infer author networks not from citation analyses but from topic similarity analyses based on a topic-model of published documents. Such author networks reveal, as we call them, “hidden communities of interest” (HCoIs) whose semantic content can easily be interpreted by means of their associated topics in the model. We use an astrobiology corpus of full-text articles (N = 3,698) to illustrate the approach. Having conducted an LDA topic-model on all publications, we identify the underlying communities of authors by measuring author correlations in terms of topic distributions. Adding publication dates makes it possible to examine HCoI evolution over time. This approach to social networks supplements traditional methods in contexts where textual data are available.

Industrial clusters, geographical concentrations of interconnected companies, aim to achieve technological innovation by acquiring common technology, which is the technology shared by all companies in an industrial cluster. Obtaining patents from universities is a primary way to gain common technology. However, existing patent recommendation methods have primarily focused on meeting the technological needs of individual companies, thus falling short in addressing the common technological requirements of all companies within an industrial cluster. To address the problem, we propose a deep learning (DL) method that recommends patents to industrial clusters based on common technological needs mining (DL_CTNM). The proposed method mines the common needs from patents owned by the companies and domain knowledge about potential technologies common to industries. Specifically, we mine the technological needs of the companies from their patents using long short-term memory networks and obtain their patent-based common needs by designing a candidate patent-aware attention mechanism. Then, we extract implicit technology directions from the domain knowledge using a capsule network and obtain domain knowledge-based common needs by designing an industrial cluster-aware attention mechanism. We evaluate the proposed method through offline and online experiments, comparing it to various benchmark methods. The experimental results demonstrate that our method outperforms these benchmarks in terms of recall and normalized discounted cumulative gain.

The purpose of this paper is to consider the joint action of structural, cognitive and relational social capital, and to explore the formation mechanism of the innovation network of new energy vehicles (NEV). The research data come from China's NEV cooperative invention patent applications from 2001 to 2019. This paper uses the exponential random graph model (ERGM) to study the impact of different dimensions of social capital on the NEV industry-university-research (I-U-R) innovation network. The results show that from the perspective of structural capital, the closed network structure has a positive impact on the formation of NEV I-U-R innovation network. From the perspective of cognitive capital, the homogeneity of knowledge base has a positive effect on the formation of the NEV I-U-R innovation network, and the innovation subjects with the same knowledge base breadth and the same knowledge base depth are more inclined to form a cooperative relationship. For relational capital, institutional environment similarity and organizational structure similarity are important factors affecting the formation of NEV I-U-R innovation network to a similar extent. The findings of this study provide scientific references for promoting the sustainable development of I-U-R innovation network in NEV industry.

Technology fusion refers to the phenomenon in which distinct technology domains overlap. Despite its importance in technology innovation and evolution, few studies have explored the general pattern of the cross-domain search process leading to technology fusion. This paper proposes that the stretching between distinct technology domains could be viewed as searching in a two-dimensional knowledge partition landscape and then empirically validates the model based on a large patent dataset derived from the U.S. Patent and Trade Office (USPTO). The findings show that the general pattern of the search processes leading to technology fusion could be viewed as searching across a broad technology scope to identify limited valuable linking points within existing technology domains, and the search processes are mainly “divergent”; that is, innovative agents gradually extend the search scope to pursue new hybrid technologies. However, the cross-domain search would be more targeted if the two technology domains were closer to each other. In addition, compared to searching across a broader technology scope, digging in certain technology areas is more important for the generation of new high-impact hybrid technologies. This study provides a novel perspective for understanding the new knowledge creation process and technology fusion.

Academic collaboration is fundamental to the advancement of scientific research. However, with the growing number of publications and researchers, it becomes increasingly challenging to identify suitable collaborators. Academic collaborator recommendation is a promising solution to this problem. Traditional recommendation methods based on collaborative filtering suffer serious data sparsity. In recent years, network topology-based methods have shown good recommendation performance while alleviating the data sparsity issue to some extent by exploiting the relationships between nodes and their attributes. Nevertheless, these methods are typically based on homogeneous collaboration networks that consist only of scholar nodes and collaboration relationships, leading to suboptimal performance. In reality, collaboration involves many different types of nodes and relations that accumulate multiplex information. To address this issue, we construct a heterogeneous academic information network comprising four types of nodes: scholars, papers, organizations, and publication venues. An academic collaborator recommendation model is designed to capture multi-type attribute features and network topology features of nodes through metapaths based on the network. Specifically, the attribute features of nodes are embedded by a node type-aware embedding method. The topology features are then extracted through the node type-aware aggregation and metapath instance aggregation procedure. After that, we utilize a metapath aggregation method to gather different types of metapaths, each representing a factor that affects collaboration. Thus, the topology information and attribute information are preserved, while encompassing multi-type factors of collaboration. Finally, we compute the vector similarity to determine collaborators. Through rigorous experimentation on a large-scale interdisciplinary academic dataset, we found that the proposed model exhibits outstanding performance in practical applications. Unlike traditional approaches confined to homogeneous collaboration networks, our model delves deeper by mining and leveraging diverse node attributes and multiple collaboration influencing factors. This approach significantly enhances the accuracy and effectiveness of collaborator recommendations. Ultimately, we aspire to contribute to a more efficient and accessible platform that simplifies the search for suitable collaborators.

As a phased achievement of technological innovation, patent analysis holds extraordinary research significance. By constructing patent citation networks, scholars have proposed various centrality algorithms (such as citation count, PageRank, LeaderRank, etc.) for evaluating the quality and influence of patents. However, these centrality algorithms suffer from age bias, which means these algorithms are more inclined to obtain higher rankings for older patents, thus losing fairness to younger patents. Additionally, the selection of algorithm performance evaluation indicators is crucial. If the indicators are not chosen appropriately, the results may be affected. Therefore, based on the background of Chinese green patents, this paper develops an unbiased evaluation ranking algorithm to identify significant Chinese green patents earlier. The results demonstrate that the combination of the rescaled method and the AttriRank algorithm can effectively obtain the importance of patents, and provide a systematic and reasonable evaluation method for measuring patent value.