Preliminary PCA of some ME Semitic lgs, made by @PhDniX & me using Militarev's 2015 Swadesh100 data ( jolr.ru/files/(164)jlr… ) to get a table of lexical cognacy %ages, and finding its eigenvectors in R. Note how Lebanese clusters with Meccan & Maltese, then Qur'anic... @nntaleb 
..while Biblical A, Palestinian A, Syriac, & Mandaic clustering together as Aramaic, with Hebrew slightly further off; Mehri is the clear outgroup here. It's unclear how robust this is if you add more lgs or more vocab, but this result matches traditional conclusions very nicely.
(You can view the tables used at docs.google.com/spreadsheets/d… ; no doubt there will be arguable points. Ugaritic & Phoenician were excluded because of too many gaps.)
