Chenwei Zhang

International scientific collaborations bring together diverse academic and cultural resources, but they also create coordination challenges that make leadership especially important. This study examines leaders’ structural positions within internationally composed scientific teams by distinguishing majority-country leaders, whose institutional country is the numerically dominant country in the team, from minority-country leaders. Using large-scale bibliometric data from computer science, we compare team characteristics and citation outcomes across these two leadership contexts. We find that majority-country-led teams are more common and slightly larger, whereas minority-country-led teams tend to exhibit greater leader-member gaps in academic age and productivity and somewhat higher country diversity. In baseline citation models, majority-country leadership is associated with slightly higher 2-year and 5-year citation counts. However, this association is substantially attenuated after accounting for the team’s dominant country: the short-term association becomes statistically insignificant, while only a small long-term association remains. Further analyses show that the remaining citation association is concentrated mainly among U.S./China-dominated teams rather than being a universal pattern across all dominant-country contexts. Overall, the findings suggest that leaders’ majority-minority country positions are associated with team composition and citation outcomes, but the citation premium is modest, context-dependent, and intertwined with the global visibility of dominant scientific systems.

Scientific collaboration has become increasingly dominant in the production of scientific knowledge, with research teams growing larger over time. The resulting division of scientific labor, encompassing both labor specialization and task division, has gradually attracted scholarly attention. Existing studies, however, usually focus on only one aspect of this division and have largely overlooked a systematic, quantitative measure of the overall division of labor—an issue of growing importance in the era of big science. To address this gap, this study proposes a concentration-based measure of the degree of division of labor in scientific collaborations that integrates both scientific labor specialization and task division, using contribution statement data. Analyzing more than 148,000 author contribution statements from Public Library of Science journals, we obtain several key findings: first, the DDoSL increases with team size and stabilizes at a moderate level once the team size exceeds eight members; second, against a backdrop of an overall increase in DDoSL over time, smaller and larger teams exhibit opposite temporal trends, suggesting different talent requirements; third, DDoSL levelled off considerably after the CRediT system was introduced. These findings contribute to the literature by providing new insights into the division of labor in scientific research and by offering evidence to inform science policy-making.

In an era of rapidly expanding publications, attracting readers' attention has become increasingly difficult, highlighting the importance of clear and well-structured writing. Although prior studies have examined linguistic features, much less is known about how the logical progression of ideas influences the scholarly impact. To address this gap, we develop a large language model (LLM)-based agent, prompted by established writing guidelines, to assess how effectively each paragraph in a paper's introduction performs its intended logical functions. Using a data set of 125,825 research articles in six disciplines, we analyze how the logical progression and adherence to the writing guidelines relate to the citation outcomes. Our findings show that introductions with clearer functional roles, sufficient background, and guideline-aligned sequencing receive more citations in Computer Science, Mathematics, and Physics. In contrast, disciplinary writing conventions and reputation-based citing behaviors likely explain the divergent trends observed in Medicine and Economics.

Aspect-based sentiment analysis (ABSA) aims to extract all aspect terms from review texts and classify their sentiment polarity. Although supervised methods exhibit excellent performance on the ABSA task, review texts in new domains typically lack labels, and it is time-consuming and costly to annotate for them. Unsupervised domain adaptation can transfer the knowledge learned from the source domain to the target domain to alleviate the problem of insufficient fine-grained labeled data. In this paper, we propose a novel label-review-opinion generative model (LRO-Gen) for cross-domain ABSA task. Specifically, we generate pseudo-labels for target domain review texts. Then, the model uses syntactic similarity rules alongside the pseudo-labels to generate new target-specific labeled review texts, which narrows the gap between domains. Finally, the generative extraction model is used in place of the traditional extraction model to improve domain generalization. In particular, we strengthen the correlation between sentiment elements by transforming generating a pair of sentiment elements into a pseudo-opinion sentence. Experimental results on four datasets demonstrate that our model is more effective than previous cross-domain ABSA models.

Aspect Sentiment Triplet Extraction (ASTE) is a fine-grained sentiment analysis task that involves identifying aspect and opinion terms and conducting sentiment analysis for each aspect-opinion pair. We categorize existing methods into tagging-based, span-based, and generation methods. However, despite notable achievements, these methods still face certain challenges: span-based methods struggle to distinguish similar spans, and generation methods are prone to slower decoding speeds. To address these challenges, we propose a novel ASTE framework, called DiffuSyn, which leverages diffusion models and syntactic dependency parsing to improve the performance of sentiment analysis. Specifically, we first add Gaussian noise to the indices at the beginning and end of aspect words and opinion words in sentences based on a non-autoregressive diffusion model. We obtain accurate boundaries of aspect and opinion terms through boundary denoising, and thus identify precise spans. Secondly, we introduce a syntactic dependency parser to capture the syntactic dependencies within review sentences, thereby providing effective syntactic information for sentiment analysis. Finally, we complete the matching of aspects and opinions by leveraging the extracted boundaries and syntactic semantic information, facilitating the prediction of sentiment relationships. We conduct experiments on four public datasets (ASTE-Data-V2), and the results indicate the effectiveness of our approach in the ASTE task. Furthermore, our method achieves state-of-the-art performance in both Aspect Term Extraction (ATE) and Opinion Term Extraction (OTE) tasks.

Online one-on-one tutoring serves as a supplementary approach to traditional classroom instruction. It has been shown to enhance personalized learning and academic performance. However, the dynamics of dialogic interactions within this educational setting are not fully understood. Thus, we present a computational analysis of dialogic interactions in online one-on-one mathematics tutoring. Specifically, we devised a coding scheme tailored to online tutoring sessions and leveraged advanced artificial intelligence techniques to construct an automated model for annotating dialog acts. We then investigated the basic characteristics and interaction patterns in a dataset encompassing online one-on-one tutoring dialogs within K-12 mathematics education and obtained insightful findings. First, tutors were found to often apply both didactic and other effective teaching strategies. Second, off-task chatting accounted for a significant proportion of tutoring sessions. Third, high school students exhibited greater engagement and cognitive abilities than primary and middle school students through their more active participation and superior reasoning skills. Primary school students, despite their less active participation, responded positively when engaged by tutors. The findings highlight the importance of optimizing strategies applied by tutors and students to create a more dynamic and effective learning environment and provide valuable insights into the nature of online one-on-one tutoring.