Source Code for Module dataprocessor.input.rankreader

  1  #!/usr/bin/python 
  2  # -*- coding: utf-8 -*- 
  3  ''' 
  4  Created on Jun 27, 2011 
  5   
  6  @author: Lukas Poustka, Eleftherios Avramidis 
  7  ''' 
  8   
  9   
 10  from xml.dom.minidom import parse 
 11  from sentence.parallelsentence import ParallelSentence 
 12  from sentence.sentence import SimpleSentence 
 13  from xml.sax.saxutils import unescape 
 14  from dataprocessor.input.genericreader import GenericReader 
 15    
 16   
 17 -class RankReader(GenericReader): 
 18      """ 
 19      Reader able to parse the ranking results from taraxu 1st evaluation round, as exported by cfedermann 
 20      """ 
 21   
 22 -    def __init__(self, input_xml_filename, load = True): 
 23          """ 
 24          Constructor. Creates an XML object that handles ranking file data 
 25          @param input_xml_filename: the name of XML file 
 26          @type input_xml_filename: string 
 27          @param load: by turning this option to false, the instance will be  
 28                       initialized without loading everything into memory 
 29          @type load: boolean  
 30          """ 
 31          self.input_filename = input_xml_filename 
 32          self.loaded = load 
 33          if load: 
 34              self.load() 
 35       
 36 -    def load(self): 
 37          """ 
 38          Loads the data of the file into memory. It is useful if the Classes has 
 39          been asked not to load the filename upon initialization 
 40          """ 
 41          self.xmlObject = parse(self.input_filename) 
 42   
 43 -    def unload(self): 
 44          self.xmlObject.unlink() 
 45           
 46       
 47   
 48 -    def get_parallelsentences(self): 
 49          """ 
 50          This function parses a ranking xml file and returns a list of parallel 
 51          sentence objects. 
 52          @return ps_list: list of tuples in format (ranking-item_id, ParallelSentence) 
 53          @type ps_list: list of tuples 
 54          """ 
 55          r_items = self.xmlObject.getElementsByTagName('ranking-item') 
 56          ps_list = [] 
 57          for r_item in r_items: 
 58              stc_id = r_item.getAttribute('sentence_id') 
 59              src = '' 
 60              tgt_list = [] 
 61               
 62              
 63              for rank_child in r_item.childNodes: 
 64                  if rank_child.nodeName == 'source': 
 65                      src = SimpleSentence(unescape(rank_child.childNodes[0].nodeValue)) 
 66                  elif rank_child.nodeName != '#text': 
 67                      tgt = SimpleSentence(unescape(rank_child.childNodes[0].nodeValue)) 
 68                      for attribute_name in rank_child.attributes.keys(): 
 69                          attribute_value = rank_child.getAttribute(attribute_name) 
 70                          tgt.add_attribute(attribute_name, attribute_value) 
 71                      tgt.add_attribute('system', rank_child.getAttribute('name')) 
 72  #                   tgt.add_attribute('rank', rank_child.getAttribute('rank')) 
 73                      tgt_list.append(tgt) 
 74               
 75              ps = ParallelSentence(src, tgt_list) 
 76              #TODO: this was old, may have to change the attribute key. Commented because overlapping with other features 
 77  #            if not ps.get_attributes().has_key("id"):  
 78  #                ps.add_attributes({'id': stc_id}) 
 79              ps.add_attributes({'sentence_id': stc_id}) 
 80              ps_list.append(ps) 
 81          return ps_list 
 82       
 83       
 84       
 85 -class R2RankReader(GenericReader): 
 86      """ 
 87      Reader able to parse the ranking results from taraxu 2nd evaluation round, as exported by cfedermann 
 88      """ 
 89   
 90 -    def __init__(self, input_xml_filename, load = True): 
 91          """ 
 92          Constructor. Creates an XML object that handles ranking file data 
 93          @param input_xml_filename: the name of XML file 
 94          @type input_xml_filename: string 
 95          @param load: by turning this option to false, the instance will be  
 96                       initialized without loading everything into memory 
 97          @type load: boolean  
 98          """ 
 99          self.input_filename = input_xml_filename 
100          self.loaded = load 
101          if load: 
102              self.load() 
103       
104 -    def load(self): 
105          """ 
106          Loads the data of the file into memory. It is useful if the Classes has 
107          been asked not to load the filename upon initialization 
108          """ 
109          self.xmlObject = parse(self.input_filename) 
110   
111 -    def unload(self): 
112          self.xmlObject.unlink() 
113           
114       
115   
116 -    def get_parallelsentences(self): 
117          """ 
118          This function parses a ranking xml file and returns a list of parallel 
119          sentence objects. 
120          @return ps_list: list of tuples in format (ranking-item_id, ParallelSentence) 
121          @type ps_list: list of tuples 
122          """ 
123          r_items = self.xmlObject.getElementsByTagName('ranking-item') 
124          ps_list = [] 
125          for r_item in r_items: 
126              stc_id = r_item.getAttribute('id') 
127              src = SimpleSentence('') 
128              tgt_list = [] 
129               
130              
131              for rank_child in r_item.childNodes: 
132                  if rank_child.nodeName == 'source': 
133                      src = SimpleSentence(unescape(rank_child.childNodes[0].nodeValue)) 
134                  elif rank_child.nodeName == 'translation': 
135                      tgt = SimpleSentence('') 
136                      for attribute_name in rank_child.attributes.keys(): 
137                          attribute_value = rank_child.getAttribute(attribute_name) 
138                          tgt.add_attribute(attribute_name, attribute_value) 
139  #                     tgt.add_attribute('system', rank_child.getAttribute('name')) 
140  #                   tgt.add_attribute('rank', rank_child.getAttribute('rank')) 
141                      tgt_list.append(tgt) 
142               
143              attributes = {'sentence_id': stc_id, 
144                            'id': stc_id,   
145                                 'langsrc': 'de', 
146                                 'langtgt': 'en' 
147                                 } 
148               
149               
150              ps = ParallelSentence(src, tgt_list, None, attributes) 
151              #TODO: this was old, may have to change the attribute key. Commented because overlapping with other features 
152  #            if not ps.get_attributes().has_key("id"):  
153  #                ps.add_attributes({'id': stc_id}) 
154               
155              ps_list.append(ps) 
156          return ps_list 
157