function [X,L,trainindices,testindices] = makedataset 

files = {};
for i=1:8
 for j=1:8
   files{1+length(files)} = sprintf('q%dnc%d',i,j);
 end
end	

X = {};

for i=1:4
i
  tei{i} = [16*i-15:16*i];
  te4{i} = files(tei{i});
  tr4{i} = files(setdiff([1:64],tei{i}));
  fid = fopen(sprintf('trainfiles%d.txt',i),'wt');
  for j=1:length(tr4{i})
    fprintf(fid,'%s\n',tr4{i}{j});
  end
  fclose(fid);
  fid = fopen(sprintf('testfiles%d.txt',i),'wt');
  for j=1:length(te4{i})
    fprintf(fid,'%s\n',te4{i}{j});
  end
  fclose(fid);

  cmd = sprintf('perl ngrams.pl -s textfeats%d.txt -f 2 -e 1 -o -w -n -x 3 -c trainfiles%d.txt',i,i);
cmd
  system (cmd);

  cmd = sprintf('perl ngrams.pl -r textfeats%d.txt -t tf%d -c testfiles%d.txt',i,i,i);
cmd
  system (cmd);

  cmd = sprintf('perl ngrams.pl -r textfeats%d.txt -t tf%d -c trainfiles%d.txt',i,i,i);
cmd
  system (cmd);

  cmd = sprintf('wc -l textfeats%d.txt > tmp.dat',i);
cmd
  system (cmd);

  z = loadcell('tmp.dat');
  z = strread(z{1},'%s','delimiter',' ');

  D = str2num(z{1}) + 1;

  L{i} = [];
  a = 0;

  indi = [];
  indj = [];
  vals = [];
 
  istest = [];
  trainindices{i} = [];
  testindices{i} = [];

  for j=1:length(files)
    [y,x] = svmlread(sprintf('%s_text_tf%d.dat',files{j},i));
    L{i} = vertcat(L{i},y);
    b = a+length(y);
    [ii,jj,v]=find(x);
    indi = vertcat(indi,a+ii);
    indj = vertcat(indj,jj);
    vals = vertcat(vals,v);

    istest(a+1:b) = length(tei{i}) - length(setdiff(tei{i},j));
    a = b;
  end

  N = length(L{i});

  testindices{i} = find(istest);
  trainindices{i} = setdiff([1:N],testindices{i});

  X{i} = sparse(indi,indj,vals,N,D); 

end

L=L{1};