Browsing by Author "Jiang, Ming"

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Evaluating BERT-based scientific relation classifiers for scholarly knowledge graph construction on digital library collections
    (Springer, 2021-11-02) Jiang, Ming; D'Souza, Jennifer; Auer, Sören; Downie, J. Stephen; Human-Centered Computing, School of Informatics and Computing
    The rapid growth of research publications has placed great demands on digital libraries (DL) for advanced information management technologies. To cater to these demands, techniques relying on knowledge-graph structures are being advocated. In such graph-based pipelines, inferring semantic relations between related scientific concepts is a crucial step. Recently, BERT-based pre-trained models have been popularly explored for automatic relation classification. Despite significant progress, most of them were evaluated in different scenarios, which limits their comparability. Furthermore, existing methods are primarily evaluated on clean texts, which ignores the digitization context of early scholarly publications in terms of machine scanning and optical character recognition (OCR). In such cases, the texts may contain OCR noise, in turn creating uncertainty about existing classifiers’ performances. To address these limitations, we started by creating OCR-noisy texts based on three clean corpora. Given these parallel corpora, we conducted a thorough empirical evaluation of eight BERT-based classification models by focusing on three factors: (1) BERT variants; (2) classification strategies; and, (3) OCR noise impacts. Experiments on clean data show that the domain-specific pre-trained BERT is the best variant to identify scientific relations. The strategy of predicting a single relation each time outperforms the one simultaneously identifying multiple relations in general. The optimal classifier’s performance can decline by around 10% to 20% in F-score on the noisy corpora. Insights discussed in this study can help DL stakeholders select techniques for building optimal knowledge-graph-based systems.
  • Thumbnail Image
    Item
    Methylation of dual-specificity phosphatase 4 controls cell differentiation
    (Cell Press, 2021) Su, Hairui; Jiang, Ming; Senevirathne, Chamara; Aluri, Srinivas; Zhang, Tuo; Guo, Han; Xavier-Ferrucio, Juliana; Jin, Shuiling; Tran, Ngoc-Tung; Liu, Szu-Mam; Sun, Chiao-Wang; Zhu, Yongxia; Zhao, Qing; Chen, Yuling; Cable, LouAnn; Shen, Yudao; Liu, Jing; Qu, Cheng-Kui; Han, Xiaosi; Klug, Christopher A.; Bhatia, Ravi; Chen, Yabing; Nimer, Stephen D.; Zheng, Y. George; Iancu-Rubin, Camelia; Jin, Jian; Deng, Haiteng; Krause, Diane S.; Xiang, Jenny; Verma, Amit; Luo, Minkui; Zhao, Xinyang; Pediatrics, School of Medicine
    Mitogen-activated protein kinases (MAPKs) are inactivated by dual-specificity phosphatases (DUSPs), the activities of which are tightly regulated during cell differentiation. Using knockdown screening and single-cell transcriptional analysis, we demonstrate that DUSP4 is the phosphatase that specifically inactivates p38 kinase to promote megakaryocyte (Mk) differentiation. Mechanistically, PRMT1-mediated methylation of DUSP4 triggers its ubiquitinylation by an E3 ligase HUWE1. Interestingly, the mechanistic axis of the DUSP4 degradation and p38 activation is also associated with a transcriptional signature of immune activation in Mk cells. In the context of thrombocytopenia observed in myelodysplastic syndrome (MDS), we demonstrate that high levels of p38 MAPK and PRMT1 are associated with low platelet counts and adverse prognosis, while pharmacological inhibition of p38 MAPK or PRMT1 stimulates megakaryopoiesis. These findings provide mechanistic insights into the role of the PRMT1-DUSP4-p38 axis on Mk differentiation and present a strategy for treatment of thrombocytopenia associated with MDS.
  • Thumbnail Image
    Item
    National wetland mapping in China: a new product resulting from object-based and hierarchical classification of Landsat 8 OLI images
    (Elsevier, 2020-06) Mao, Dehua; Wang, Zongming; Du, Baojia; Li, Lin; Tian, Yanlin; Jia, Mingming; Zeng, Yuan; Song, Kaishan; Jiang, Ming; Wang, Yeqiao; Earth Sciences, School of Science
    Spatially and thematically explicit information of wetlands is important to understanding ecosystem functions and services, as well as for establishment of management policy and implementation. However, accurate wetland mapping is limited due to lacking an operational classification system and an effective classification approach at a large scale. This study was aimed to map wetlands in China by developing a hybrid object-based and hierarchical classification approach (HOHC) and a new wetland classification system for remote sensing. Application of the hybrid approach and the wetland classification system to Landsat 8 Operational Land Imager data resulted in a wetland map of China with an overall classification accuracy of 95.1%. This national scale wetland map, so named CAS_Wetlands, reveals that China’s wetland area is estimated to be 451,084 ± 2014 km2, of which 70.5% is accounted by inland wetlands. Of the 14 sub-categories, inland marsh has the largest area (152,429 ± 373 km2), while coastal swamp has the smallest coverage (259 ± 15 km2). Geospatial variations in wetland areas at multiple scales indicate that China’s wetlands mostly present in Tibet, Qinghai, Inner Mongolia, Heilongjiang, and Xinjiang Provinces. This new map provides a new baseline data to establish multi-temporal and continuous datasets for China’s wetlands and biodiversity conservation.