The TMSCORE Procedure

The TMSCORE procedure scores textual data in SAS Viya. In text mining, scoring is the process of applying parsing and singular value decomposition (SVD) projections to new textual data. The TMSCORE procedure performs this scoring of new documents, and its primary outputs are the Outparent data table (which holds the parsing results of the term-by-document matrix) and the Outdocpro data table (which holds the reduced-dimensional representation of the score collection). PROC TMSCORE uses some of the output data tables of the TEXTMINE procedure as input data to ensure consistency between scoring and training. During scoring, the new textual data must be parsed using the same settings that the training data were parsed with, indexed using only the subset of terms that were used during training, and projected onto the reduced-dimensional subspace of the singular value decomposition that was derived from the training data. To facilitate this process, you specify the CONFIG=, TERMS=, and SVDU= options in PROC TEXTMINE to create three data tables (Outconfig, Outterms, and Svdu, respectively), and then you specify those three data tables as inputs to PROC TMSCORE. For more information about these data tables, see the CONFIG=, TERMS=, and SVDU= options, respectively, in the section PROC TMSCORE Statement.