We present an approach to estimate information carried by experimentally observed neural spike trains elicited by known stimuli. This approach makes use of an embedding of the observed spike trains into a set of vector spaces, and entropy estimates based on the nearest-neighbor Euclidean distances within these vector spaces [L. F. Kozachenko and N. N. Leonenko, Probl. Peredachi Inf. 23, 9 (1987)]. Using numerical examples, we show that this approach can be dramatically more efficient than standard bin-based approaches such as the "direct" method [S. P. Strong, R. Koberle, R. R. de Ruyter van Steveninck, and W. Bialek, Phys. Rev. Lett. 80, 197 (1998)] for amounts of data typically available from laboratory experiments.