Source Code for Module dataprocessor.input.taraxureader

 1  #!/usr/bin/python 
 2  # -*- coding: utf-8 -*- 
 3   
 4   
 5  """ 
 6  Created on 15 Οκτ 2010 
 7   
 8  @author: Eleftherios Avramidis 
 9  """ 
10   
11   
12  from xml.dom.minidom import parse 
13  from sentence.parallelsentence import ParallelSentence 
14  from sentence.sentence import SimpleSentence 
15  from xml.sax.saxutils import unescape 
16  from xmlreader import XmlReader 
17   
18 -class TaraXUReader(XmlReader): 
19      """ 
20      classdocs 
21      """ 
22   
23 -    def __init__(self, inputFilename): 
24          """ 
25          Constructor. Creates an XML object that handles the XML 
26          """ 
27          self.TAG_DOC = "doc" 
28          self.TAG_SENT = "sentence" 
29          self.TAG_SRC = "source" 
30          self.TAG_TGT = "target" 
31          self.TAG_REF = "reference" 
32          self.TAG_ANNOTATIONS = "annotations" 
33          self.TAG_ANNOTATION = "annotation" 
34          self.xmlObject = parse(inputFilename) 
35       
36       
37   
38 -    def get_parallelsentences(self, start = None, end = None): 
39          """ 
40          @return: a list of ParallelSentence objects 
41          """ 
42          judgedCorpus = self.xmlObject.getElementsByTagName(self.TAG_DOC) 
43          langsrc = judgedCorpus[0].attributes["source_language"].value 
44          langtgt = judgedCorpus[0].attributes["target_language"].value 
45          if not start and not end: 
46              sentenceList = judgedCorpus[0].getElementsByTagName('sentence') 
47          else: 
48              sentenceList = judgedCorpus[0].getElementsByTagName('sentence')[start:end] 
49          newssentences = []  
50          for xmlEntry in sentenceList: 
51              srcXML = xmlEntry.getElementsByTagName('source') 
52              tgtXML = xmlEntry.getElementsByTagName('target') 
53              refXML = xmlEntry.getElementsByTagName('reference') 
54               
55              src = SimpleSentence (unescape(srcXML[0].childNodes[0].nodeValue.strip()) , self.__read_attributes__(srcXML[0]) ) 
56               
57              #Create a list of SimpleSentence objects out of the object 
58              tgt = map( lambda x: SimpleSentence(unescape(x.childNodes[0].nodeValue.strip()), self.__read_attributes__(x) )  , tgtXML )  
59               
60              ref = SimpleSentence() 
61              try:     
62                  ref = SimpleSentence (unescape(refXML[0].childNodes[0].nodeValue.strip()) ,  self.__read_attributes__(refXML[0])) 
63              except LookupError: 
64                  pass 
65               
66              #Extract the XML features and attach them to the ParallelSentenceObject 
67              attributes = self.__read_attributes__(xmlEntry) 
68              attributes["langsrc"] =  langsrc 
69              attributes["langtgt"] =  langtgt 
70                           
71              #create a new Parallesentence with the given content 
72              curJudgedSentence = ParallelSentence(src, tgt, ref, attributes) 
73               
74               
75              newssentences.append(curJudgedSentence) 
76          return newssentences 
77