other ideas

Ideas for Activities

ID	Ideas	Name	Date
001	Use Universal Sentence Encoder+ML Reference: Daniel Cer, Yinfei Yang, Sheng-yi Kong, Nan Hua, Nicole Limtiaco, Rhomni St. John, Noah Constant, Mario Guajardo-Céspedes, Steve Yuan, Chris Tar, Yun-Hsuan Sung, Brian Strope, Ray Kurzweil. Universal Sentence Encoder. arXiv:1803.11175, 2018	Sam Tomioka
002	Build word embedding from CDISC documentation	Sam Tomioka
003	Graph approach? Structure Learning, Graph Convolutional Network	Sam Tomioka
004	Build conventions for output $Y$	Sam Tomioka
005	Maybe do multi step approaches rather than end-to-end where the first is to classify the domains, then classify drop or keep, then classify them into SDTM variables so we could reduce the amount of training data?	Sam Tomioka
006	Collect metadata of raw variables used for SDTM	Sam Tomioka