HarpLDA+: Optimizing latent dirichlet allocation for parallel efficiency | IEEE Conference Publication | IEEE Xplore