Source Code for Module support.preprocessing.jcml.jcml2arff

  1  ''' 
  2  Created on 06 Mar 2012 
  3   
  4  @author: Eleftherios Avramidis 
  5  ''' 
  6  from dataprocessor.input.jcmlreader import JcmlReader 
  7   
  8 -class ArffWriter: 
  9 -    def __init__(self, output_filename, class_type, class_values=[], numeric_attribute_names=[], string_attribute_names=[], nominal_attributes={}): 
 10          self.f = open(output_filename, 'w') 
 11          self.class_type = class_type 
 12          self.class_values = class_values 
 13          self.numeric_attribute_names = numeric_attribute_names 
 14          self.string_attribute_names = string_attribute_names 
 15          self.nominal_attributes = nominal_attributes 
 16   
 17 -    def write_header(self, relation_name, ): 
 18          self.f.write("@RELATION {0}\n\n".format(relation_name)) 
 19           
 20          for attribute_name in self.numeric_attribute_names: 
 21              self.f.write("@ATTRIBUTE {0}\tNUMERIC\n".format(attribute_name)) 
 22          for attribute_name in self.string_attribute_names: 
 23              self.f.write("@ATTRIBUTE {0}\tSTRING\n".format(attribute_name)) 
 24          for attribute_name, values in self.nominal_attributes.iteritems(): 
 25              self.f.write("@ATTRIBUTE %s\t{%s}\n" % (attribute_name, ",".join(values))) 
 26   
 27          if self.class_type == "numeric": 
 28              self.f.write("@ATTRIBUTE class\tNUMERIC\n") 
 29          elif self.class_type == "nominal": 
 30              self.f.write("@ATTRIBUTE class\t{%s}\n\n" % (",".join(self.class_values))) 
 31          self.f.write("@DATA\n") 
 32       
 33 -    def write_instance(self, attributes ={}, class_value = ""): 
 34           
 35          for attribute_name in self.numeric_attribute_names: 
 36              try: 
 37                  self.f.write("{0},".format(attributes[attribute_name].replace("inf", "999999").replace("nan", "0"))) 
 38              except KeyError: 
 39                  self.f.write("0,") 
 40          for attribute_name in self.string_attribute_names: 
 41              try: 
 42                  self.f.write("{0},".format(attributes[attribute_name].replace("nan", "0"))) 
 43              except KeyError: 
 44                  self.f.write("0,") 
 45          for attribute_name, values in self.nominal_attributes.iteritems(): 
 46              try: 
 47                  self.f.write("{0},".format(attributes[attribute_name].replace("nan", "0"))) 
 48              except KeyError:     
 49                  self.f.write("{0},".format(list(values)[0])) 
 50          self.f.write("{0}\n".format(class_value)) 
 51   
 52 -    def close(self): 
 53          if self.f: 
 54              self.f.close() 
 55 -    def __del__(self): 
 56          self.close() 
 57   
 58 -class Jcml2Arff: 
 59 -    def process(self, jcml_filename, arff_filename,  
 60                  hidden_attribute_names, discrete_attribute_names, string_attribute_names,  
 61                  relation_name, class_name, class_type, class_values): 
 62          dataset = JcmlReader(jcml_filename).get_dataset() 
 63          attribute_names = set(dataset.get_all_attribute_names()) - set(hidden_attribute_names) 
 64          nominal_attributes = dataset.get_discrete_attribute_values(discrete_attribute_names) 
 65          numeric_attribute_names = attribute_names - set(discrete_attribute_names) 
 66          arff = ArffWriter(arff_filename, class_type, class_values, numeric_attribute_names, string_attribute_names, nominal_attributes) 
 67          arff.write_header(relation_name) 
 68           
 69          for ps in dataset.get_parallelsentences(): 
 70   
 71              atts = {} 
 72              atts.update(ps.attributes) 
 73              atts.update(ps.get_nested_attributes()) 
 74              class_value = atts[class_name] 
 75              del(atts[class_name]) 
 76              for attname in hidden_attribute_names: 
 77                  try: 
 78                      del(atts[attname]) 
 79                       
 80                  except: 
 81                      pass 
 82              arff.write_instance(atts, class_value) 
 83          arff.close() 
 84   
 85   
 86  if __name__ == '__main__': 
 87      hidden_attributes = ["tgt-1_berkeley-tree", "src_berkeley-tree","ref_berkeley-tree",                          
 88                           "testset", "judgment-id", "langsrc", "langtgt", "ps1_judgement_id",  
 89                                 "id", "tgt-1_score" ,  "tgt-1_system" ,     
 90                                   ] 
 91      discrete_attributes = [ "src_reuse_status",   
 92                         "src_terminologyAdmitted_status",  
 93                         "src_total_status",  
 94                         "src_spelling_status",   
 95                         "src_style_status",   
 96                         "src_grammar_status",   
 97                         "src_terminology_status", 
 98                         "src_resultStats_projectStatus",  
 99                         ] 
100       
101      import sys 
102      input_sgml = sys.argv[1] 
103      output_arff = sys.argv[2] 
104      try: 
105          class_name = sys.argv[3] 
106      except: 
107          class_name =  "tgt-1_score" 
108      try: 
109          class_type = sys.argv[4] 
110      except: 
111          class_type = "numeric" 
112       
113      Jcml2Arff().process(input_sgml,  
114                          output_arff,  
115                          hidden_attributes, discrete_attributes, [], "qe", class_name, class_type, []) 
116