Source Code for Module test.pairwise_test

  1  ''' 
  2  Created on 14 Jul 2012 
  3   
  4  @author: Eleftherios Avramidis 
  5  ''' 
  6   
  7  from dataprocessor.input.jcmlreader import JcmlReader 
  8  from dataprocessor.sax.saxps2jcml import Parallelsentence2Jcml 
  9  from sentence.pairwisedataset import AnalyticPairwiseDataset, CompactPairwiseDataset, FilteredPairwiseDataset 
 10  import os 
 11  import unittest 
 12  from numpy.ma.testutils import assert_equal 
 13  from sentence.rankhandler import RankHandler 
 14  from sentence.dataset import DataSet 
 15  from dataprocessor.output.xmlwriter import XmlWriter 
 16   
 17   
 18   
 19   
 20 -class PairwiseTesting(unittest.TestCase): 
 21      ''' 
 22      classdocs 
 23      ''' 
 24 -    def setUp(self): 
 25          self.filename = "pairwiseparallelsentence_test.jcml" 
 26          self.mydataset = JcmlReader(self.filename).get_dataset() 
 27       
 28   
 29   
 30 -    def test_pairwise_reconstruct_twice(self): 
 31          """ 
 32          Loads a dataset, converts that to pairwise once and reconstructs it. Then it loads that again and reconstructs it once more 
 33          This was helpful to detect a problem of wrong  
 34          """ 
 35           
 36           
 37          #first perform typical cleanup of the test set 
 38          analytic_testset = AnalyticPairwiseDataset(self.mydataset) #this  
 39          filtered_dataset = FilteredPairwiseDataset(analytic_testset, 1.00) 
 40          filtered_dataset.remove_ties() 
 41   
 42          output_filename = "filtered_1.jcml" 
 43          Parallelsentence2Jcml(filtered_dataset.get_parallelsentences(), shuffle_translations=False, sort_attribute="system").write_to_file(output_filename) 
 44   
 45          reconstructed_dataset = filtered_dataset.get_multiclass_set() 
 46  #        reconstructed_dataset.remove_ties() 
 47           
 48          output_filename = "reconstructed_1.jcml" 
 49          Parallelsentence2Jcml(reconstructed_dataset.get_parallelsentences(), shuffle_translations=False, sort_attribute="system").write_to_file(output_filename) 
 50           
 51          #retrieve clean test set from the file and repeat the handling 
 52          simple_testset = JcmlReader(output_filename).get_dataset() 
 53          analytic_testset_2 = AnalyticPairwiseDataset(simple_testset) #this  
 54          compact_testset_2 = CompactPairwiseDataset(analytic_testset_2) 
 55   
 56          output_filename = "filtered_2.jcml" 
 57          Parallelsentence2Jcml(compact_testset_2.get_parallelsentences(), shuffle_translations=False, sort_attribute="system").write_to_file(output_filename) 
 58           
 59  #        self.assertEqual(len(filtered_dataset.get_parallelsentences()), len(compact_testset_2.get_parallelsentences()))  
 60           
 61          reconstructed_dataset_2 = compact_testset_2.get_multiclass_set() 
 62          output_filename = "reconstructed_2.jcml" 
 63          Parallelsentence2Jcml(reconstructed_dataset_2.get_parallelsentences(), shuffle_translations=False, sort_attribute="system").write_to_file(output_filename) 
 64           
 65          reconstructed_1 = reconstructed_dataset.get_parallelsentences() 
 66          reconstructed_2 = reconstructed_dataset_2.get_parallelsentences() 
 67          self.assertEqual(len(reconstructed_1), len(reconstructed_2), "The number of sentences when reconstructing the same set twice has changed") 
 68           
 69  #        for sentence_id in compact_testset_2.get_pairwise_parallelsentence_sets().iterkeys(): 
 70  #            pset1 = compact_testset_2.get_pairwise_parallelsentence_set(sentence_id) 
 71  #            print pset1.get_system_names() 
 72  #            pset2 = compact_testset_2.get_pairwise_parallelsentence_set(sentence_id) 
 73  #            print pset2.get_system_names() 
 74           
 75          for p1, p2 in zip(reconstructed_1, reconstructed_2): 
 76              systemrank1 = set([(tgt.get_attribute("system"), tgt.get_attribute("rank")) for tgt in p1.get_translations()])  
 77              systemrank2 = set([(tgt.get_attribute("system"), tgt.get_attribute("rank")) for tgt in p2.get_translations()]) 
 78              self.assertEqual(systemrank1, systemrank2) 
 79               
 80   
 81 -    def test_pairwise_save_reconstruct(self): 
 82          """ 
 83          Loads a dataset, converts that to pairwise once and reconstructs it. Then it loads that again and reconstructs it once more 
 84          This was helpful to detect a problem of wrong  
 85          """ 
 86           
 87           
 88          #first perform typical cleanup of the test set 
 89          analytic_testset = AnalyticPairwiseDataset(self.mydataset) #this  
 90          filtered_dataset = FilteredPairwiseDataset(analytic_testset, 1.00) 
 91          filtered_dataset.remove_ties() 
 92   
 93          output_filename = "filtered_1.jcml" 
 94          Parallelsentence2Jcml(filtered_dataset.get_parallelsentences(), shuffle_translations=False, sort_attribute="system").write_to_file(output_filename) 
 95   
 96          reconstructed_dataset = filtered_dataset.get_multiclass_set() 
 97  #        reconstructed_dataset.remove_ties() 
 98           
 99          output_filename = "reconstructed_1.jcml" 
100          Parallelsentence2Jcml(reconstructed_dataset.get_parallelsentences(), shuffle_translations=False, sort_attribute="system").write_to_file(output_filename) 
101           
102          #retrieve pairwise test set from the file and repeat the handling 
103          simple_testset = JcmlReader("filtered_1.jcml").get_dataset() 
104          analytic_testset_2 = AnalyticPairwiseDataset(simple_testset) #this  
105          compact_testset_2 = CompactPairwiseDataset(analytic_testset_2) 
106   
107          output_filename = "filtered_2.jcml" 
108          Parallelsentence2Jcml(compact_testset_2.get_parallelsentences(), shuffle_translations=False, sort_attribute="system").write_to_file(output_filename) 
109           
110  #        self.assertEqual(len(filtered_dataset.get_parallelsentences()), len(compact_testset_2.get_parallelsentences()))  
111           
112          reconstructed_dataset_2 = compact_testset_2.get_multiclass_set() 
113          output_filename = "reconstructed_2.jcml" 
114          Parallelsentence2Jcml(reconstructed_dataset_2.get_parallelsentences(), shuffle_translations=False, sort_attribute="system").write_to_file(output_filename) 
115           
116          reconstructed_1 = reconstructed_dataset.get_parallelsentences() 
117          reconstructed_2 = reconstructed_dataset_2.get_parallelsentences() 
118          self.assertEqual(len(reconstructed_1), len(reconstructed_2), "The number of sentences when reconstructing the same set twice has changed") 
119           
120  #        for sentence_id in compact_testset_2.get_pairwise_parallelsentence_sets().iterkeys(): 
121  #            pset1 = compact_testset_2.get_pairwise_parallelsentence_set(sentence_id) 
122  #            print pset1.get_system_names() 
123  #            pset2 = compact_testset_2.get_pairwise_parallelsentence_set(sentence_id) 
124  #            print pset2.get_system_names() 
125           
126          for p1, p2 in zip(reconstructed_1, reconstructed_2): 
127              systemrank1 = set([(tgt.get_attribute("system"), tgt.get_attribute("rank")) for tgt in p1.get_translations()])  
128              systemrank2 = set([(tgt.get_attribute("system"), tgt.get_attribute("rank")) for tgt in p2.get_translations()]) 
129              self.assertEqual(systemrank1, systemrank2) 
130   
131   
132  #        rank_vector_1 
133           
134           
135  #        equal_sentences(reconstructed_dataset_2, reconstructed_dataset) 
136           
137          #first see if number of sentences at the compact sets are equal 
138