Source Code for Module TEES.ExampleBuilders.FeatureBuilders.NodalidaFeatureBuilder

 1  from FeatureBuilder import FeatureBuilder 
 2   
 3 -class NodalidaFeatureBuilder(FeatureBuilder): 
 4 -    def __init__(self, featureSet): 
 5          FeatureBuilder.__init__(self, featureSet) 
 6       
 7 -    def buildShortestPaths(self, graph, tokenPath, position=0, newPath=None): 
 8          if newPath == None: 
 9              assert(position == 0) 
10              newPath = [tokenPath[0]] 
11          else: 
12              newPath = newPath + [tokenPath[position]] 
13           
14          if position == len(tokenPath) - 1: 
15              return [newPath] 
16          forwardEdges = [] 
17          if graph.has_edge(tokenPath[position],tokenPath[position+1]): 
18              forwardEdges.extend(graph.get_edge(tokenPath[position],tokenPath[position+1])) 
19          reverseEdges = [] 
20          if graph.has_edge(tokenPath[position+1],tokenPath[position]): 
21              reverseEdges.extend(graph.get_edge(tokenPath[position+1],tokenPath[position])) 
22          newPaths = [] 
23          for reverseEdge in reverseEdges: 
24              newPaths.extend( self.buildShortestPaths(graph, tokenPath, position+1, newPath + [(reverseEdge,"reverse")]) ) 
25          for forwardEdge in forwardEdges: 
26              newPaths.extend( self.buildShortestPaths(graph, tokenPath, position+1, newPath + [(forwardEdge,"forward")]) ) 
27          return newPaths 
28       
29 -    def buildTokenGramFeatures(self, tokenPath, sentenceGraph): 
30          txtGrams = [""] 
31          annTypeGrams = [""] 
32          posGrams = [""] 
33          for token in tokenPath: 
34              featureList = self.getTokenFeatures(token, sentenceGraph) 
35              for feature in featureList: 
36                  if feature.find("txt_") != -1: 
37                      newGrams = [] 
38                      for gram in txtGrams: 
39                          newGrams.append(gram + feature) 
40                      txtGrams = newGrams 
41                  elif feature.find("POS_") != -1: 
42                      newGrams = [] 
43                      for gram in posGrams: 
44                          newGrams.append(gram + feature) 
45                      posGrams = newGrams 
46                  elif feature.find("annType_") != -1: 
47                      newGrams = [] 
48                      for gram in annTypeGrams: 
49                          newGrams.append(gram + feature) 
50                      annTypeGrams = newGrams 
51          for gram in txtGrams + annTypeGrams + posGrams: 
52              if gram != "": 
53                  self.setFeature(gram, 1) 
54       
55 -    def buildEdgeGramFeatures(self, edgePath): 
56          string = "" 
57          print edgePath 
58          for edge in edgePath: 
59              string += edge[0].attrib["type"] + "-" + edge[1] 
60          self.setFeature(string, 1) 
61       
62 -    def buildNGrams(self, paths, sentenceGraph, n=3): 
63          for path in paths: 
64              assert(len(path)%2==1) 
65              tokenPhase = True 
66              for i in range(len(path)): 
67                  # Token n-grams 
68                  if tokenPhase: 
69                      tokenGram = [] 
70                      for j in range(i, max(-1,i-n*2), -2): 
71                          tokenGram = [path[j]] + tokenGram 
72                          self.buildTokenGramFeatures(tokenGram, sentenceGraph) 
73                  # Dependency n-grams 
74                  else: 
75                      edgeGram = [] 
76                      for j in range(i, max(0,i-n*2), -2): 
77                          edgeGram = [path[j]] + edgeGram 
78                          self.buildEdgeGramFeatures(tokenGram) 
79                  tokenPhase = not tokenPhase 
80