作者: Xiang Ren , Daniel A. McFarland , Hancheng Cao , Mengjie Cheng , Zhepeng Cen
DOI:
关键词: Data science 、 Translational research 、 Knowledge transfer 、 Text corpus 、 Computer science 、 Unit of analysis 、 Government
摘要: What kind of basic research ideas are more likely to get applied in practice? There is a long line investigating patterns knowledge transfer, but it generally focuses on documents as the unit analysis and follow their transfer into practice for specific scientific domain. Here we study translational at level concepts all fields. We do this through text mining predictive modeling using three corpora: 38.6 million paper abstracts, 4 patent documents, 0.28 clinical trials. extract (i.e., phrases) from corpora instantiations "research ideas", create concept-level features motivated by literature, then trajectories over 450,000 new (emerged 1995-2014) identify factors that lead only small proportion these be used inventions drug Results our suggest several mechanisms distinguish which concept will adopted practice, not. also demonstrate derived can explain predict with high accuracy. Our work provides greater understanding researchers, practitioners, government agencies interested encouraging research.