I agree. I think your only option is to collapse the data by the first three or four strings, and from that derive a table of "make" values. After extracting that, the remainder becomes the "model". There are just too many odd possibilities to automate the detection. However,one optimization might be to determine if there are only two words, and if there are assume that they are make and model. The makes extracted from that can then be used against the remainder, and any matches can be handled in an automated fashion. This should leave a very small residue for manual evaluation. I've dealt with the same problem with municipality and province, and there's really no count-based solution. Tom
... View more