#!/usr/bin/env ruby require 'zipf' require 'trollop' conf = Trollop::options do opt :foreign, "foreign file", :type => :string, :required => true opt :english, "english file", :type => :string, :required => true opt :size, "one size", :type => :int, :required => true opt :repeat, "number of repetitions", :type => :int, :default => 1 opt :prefix, "prefix for output files", :type => :string end fn = conf[:foreign] fn_ext = fn.split('.').last f = ReadFile.readlines fn en = conf[:english] en_ext = en.split('.').last e = ReadFile.readlines en size = conf[:size] nlines_f = `wc -l #{fn}`.split()[0].to_i nlines_e = `wc -l #{en}`.split()[0].to_i if nlines_f != nlines_e STDERR.write "Unbalanced files (#{nlines_f} vs. #{nlines_e}), exiting!\n" exit 1 end prefix = conf[:prefix] a = (0..nlines_e-1).to_a i = 0 conf[:repeat].times { b = a.sample(size) ax = a.reject{|j| b.include? j} `mkdir split_#{i}` new_f = WriteFile.new "split_#{i}/#{prefix}.train.#{i}.#{fn_ext}" new_e = WriteFile.new "split_#{i}/#{prefix}.train.#{i}.#{en_ext}" ax.each { |j| new_f.write f[j] new_e.write e[j] } new_f.close; new_e.close new_f = WriteFile.new "split_#{i}/#{prefix}.test.#{i}.#{fn_ext}" new_e = WriteFile.new "split_#{i}/#{prefix}.test.#{i}.#{en_ext}" b.each { |j| new_f.write f[j] new_e.write e[j] } new_f.close; new_e.close i += 1 }