Re: [help-GIFT] Unpacking features in more user friendly format

[Wolfgang Mueller]

Dear Epa,

It depends a lot on what sizes you have. This perl is 32-bit centric.

If you want to be sure to read the same features as GIFT does, look at
gift-generate-inverted-files and see what it uses to read the .fts
files. I am currently not in a position to do this by myself. If
problems persist, please kick me again end of next week or beginning of
the week after.

Best,
Wolfgang

On Fri, 09 Mar 2007, Epa wrote:

> Hey,
> 
> I read GIFT literature and I have an idea of what GIFT does and what I can 
> expect from GIFT. Using Perl script, I am trying to unpack the .fts files so 
> I 
> can see featureID and their frequencies. I used the script from Dr. Muller 
> with small modification so I can sort features in ascending order but I am 
> wary of accepting the result.
> 
> Here is my my script:
> 
>        open (FILE,shift)|| die "can't open the file $file";
>        %features= ();
>        read (FILE,$lString,4) || die ("Cannot read the file $file");
>          while (read FILE,$lString,8|| die ("Cannot read the file $file")){
>          my ($FeatureID,$Frequency)=unpack("If",$lString);
>                if ($Frequency>0){
>                $features{"$FeatureID"}=$Frequency;
>                                 }
>          }
>   
>         foreach $key (sort{$a<=>$b} keys %features){
>               print "$key:$features{$key}";}
> 
> My output seems to have some big numbers both in FeatureID and Frequencies 
> and 
> I am wondering if this is accurate or something went wrong during features 
> extraction.
> Here is an example of some of the features form one .fts file:
> 
> 0:3.48554469996741e+28
> 1006665857:1.22526735123633e-40
> 4181688320:2.15799963506022e-43
> 4215242752:1.82168800362226e-43
> 
> I would appreciate some help or advice on how to convert the .fts into a more 
> user friendly because I am questioning these bif numbers in my output.
> 
> Thank you very much
> 
> Epa Uwimana
>  
> 
> 
>  
> 
> 
> 
> 
> 
> _______________________________________________
> help-GIFT mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/help-gift
--
Dr. Wolfgang Mueller
LS Medieninformatik
Universitaet Bamberg