The next generation of spectroscopic surveys, both Galactic and extra-galactic (e.g., SDSS, 2dF), present the challenge of classifying spectra in an efficient and objective manner. The standard approach to this problem has been to visually classify spectra based on a number of spectral features (e.g., the equivalent widths of emission lines). The size of new spectral surveys (> 106 galaxies) and the desire to compare the luminosity and environments of galaxies with their spectral properties make these techniques infeasible. We describe here an automated classification scheme that is being developed for the SDSS.