Just saw this in skimming email... On Aug 4, 2011, at 4:58 AM, Clément Notin wrote:
> Hello, > > I'm tweaking and rewriting parts of the ARFF converter to meet my needs (I > want to create NamedVectors with the the first column as name). > I asked it to output the dictionnary it used but the file is always empty ! > (BTW, I'm using 0.6-SNAPSHOT) Patch would be great, if you think it is general purpose enough. > > In the code, it is the label binding which are written to the dictionnary. > > Am I right if I think that : > > - the labels are the column names (@attribute) in the ARFF file ? Yes. > - the dictionnary should be the mapping between strings and their long id Yes. > ? > > > Correct me if I'm wrong but when my clustering job (yes I want to do > clustering ;) ) is finished it is useful to know what strings are behind > numerical values. Yes, that is usually helpful. > Because the vector values are ordered so it's easy to know > which column is which column... > > Thanks for your anwsers. > > Regards. > > -- > *Clément **Notin* -------------------------------------------- Grant Ingersoll http://www.lucidimagination.com
