Based upon the relationship between carbonate content and stratal velocity and density, we attempted to apply an artificial neural network to the inversion of carbonate content summarized from high-resolution seismic data limited by controlled well measurements. The method was applied to the slope area of the northern South China Sea near ODP Sites 1146 and 1148, with satisfactory results. The key to this method is the collection of several properties from seismic profiles across or near the wells. Then the progressive regression method was used to determine the six seismic properties most closely related to carbonate content variations, which are defined as average frequency, integrated absolute amplitude, dominating frequency, reflection time, derivative instantaneous amplitude, and instantaneous frequency. Finally, the stratal carbonate content is reversed. The reversal results thus obtained, with the errors of carbonate content mostly ranging within + or -5% relative to that measured from sediment samples, show a relatively accurate picture of carbonate distribution along the slope profile.