From dc078a058595ddb6b6b35d86d0cefc5d0e631b04 Mon Sep 17 00:00:00 2001
From: Patrick Simianer <p@simianer.de>
Date: Mon, 16 Nov 2015 17:16:28 +0100
Subject: new cdec2json.py

---
 example/cdec/cdec2json.py | 94 ++++++++++++++++++++++++-----------------------
 1 file changed, 48 insertions(+), 46 deletions(-)

(limited to 'example/cdec')

diff --git a/example/cdec/cdec2json.py b/example/cdec/cdec2json.py
index 66bc8b4..fda5604 100755
--- a/example/cdec/cdec2json.py
+++ b/example/cdec/cdec2json.py
@@ -2,52 +2,64 @@
 
 import cdec
 import sys, argparse
-import json, gzip
+import json
+import codecs
 
 def hg2json(hg, weights):
   """
-  hackishly output a JSON representation of a cdec hypegraph
-
-  TODO: cdec now uses boost's serialization instead of JSON
-        (github repo @b88176dc4fd53480e77d601ff63bf5300cf8fc7f
-         still uses the old format)
+  output a JSON representation of a cdec hypegraph
+  (see http://aclweb.org/aclwiki/index.php?title=Hypergraph_Format )
   """
   res = ''
   res += "{\n"
-  # rules
-  res += '"rules":[\n'
-  rules = []
-  for i in hg.edges:
-    s = json.dumps(str(i.trule))
-    try:
-      rules.index(s)
-    except:
-      rules.append(s)
-  res += ",\n".join(rules)
-  res += "\n],\n"
-  # nodes
+  res += '"weights":{'+"\n"
+  a = []
+  for i in weights:
+    a.append( '%s:%f'%(json.dumps(i[0]), i[1]) )
+  res += ", ".join(a)+"\n"
+  res += "},\n"
   res += '"nodes":'+"\n"
   res += "[\n"
   a = []
-  a.append('{ "id":0, "symbol":"root", "span":[-1,-1] }')
+  a.append( '{ "id":-1, "cat":"root", "span":[-1,-1] }' )
   for i in hg.nodes:
-    a.append('{ "id":%d, "symbol":"%s", "span":[%d,%d] }'%(i.id+1, i.cat, i.span[0], i.span[1]))
+    a.append( '{ "id":%d, "cat":"%s", "span":[%d, %d] }'%(i.id, i.cat, i.span[0], i.span[1]) )
   res += ",\n".join(a)+"\n"
   res += "],\n"
-  # edges
   res += '"edges":'+"\n"
   res += "[\n"
   a = []
   for i in hg.edges:
     s = "{"
-    s += '"head":%d'%(i.head_node.id+1)
-    s += ', "rule":%s'%(rules.index(json.dumps(str(i.trule))))
-    # tails
-    if len(list(i.tail_nodes)) > 0:
-      s += ', "tails":[ %s ],'%(",".join([str(n.id+1) for n in i.tail_nodes]))
+    s += '"head":%d'%(i.head_node.id)
+    xs = ' "f":{'
+    b = []
+    for j in i.feature_values:
+      b.append( '"%s":%s'%(j[0], j[1]) )
+    xs += ", ".join(b)
+    xs += "},"
+    c = []
+    for j in i.tail_nodes:
+      c.append(str(j.id))
+    if len(c) > 0:
+      s += ', "tails":[ %s ],'%(",".join(c))
     else:
-      s += ', "tails":[ 0 ],'
-    s += ' "score":%s }'%(i.prob)
+      s += ', "tails":[ -1 ],'
+    s += xs
+    f =  []
+    for x in i.trule.f:
+      if type(x) == type(u'x'):
+        f.append(codecs.encode(x, 'utf-8'))
+      else:
+        f.append(str(x))
+    e = []
+    for x in i.trule.e:
+      if type(x) == type(u'x'):
+        e.append(codecs.encode(x, 'utf-8'))
+      else:
+        e.append(str(x))
+    s += " \"rule\":\"%s ||| %s ||| %s\""%(str(i.trule.lhs), json.dumps(" ".join(f))[1:-1], json.dumps(" ".join(e))[1:-1])
+    s += ' }'
     a.append(s)
   res += ",\n".join(a)+"\n"
   res += "]\n"
@@ -55,38 +67,28 @@ def hg2json(hg, weights):
   return res
 
 def main():
-  parser = argparse.ArgumentParser(description='get a nice json representation of cdec hypergraphs')
+  parser = argparse.ArgumentParser(description='get a proper json representation of cdec hypergraphs')
   parser.add_argument('-c', '--config', required=True, help='decoder configuration')
   parser.add_argument('-w', '--weights', required=True, help='feature weights')
-  parser.add_argument('-g', '--grammar', required=False, help='grammar')
   args = parser.parse_args()
   with open(args.config) as config:
     config = config.read()
   decoder = cdec.Decoder(config)
   decoder.read_weights(args.weights)
   ins = sys.stdin.readline().strip()
-  if args.grammar:
-    if args.grammar.split('.')[-1] == 'gz':
-      with gzip.open(args.grammar) as grammar:
-        grammar = grammar.read()
-    else:
-      with open(args.grammar) as grammar:
-        grammar = grammar.read()
-    hg = decoder.translate(ins, grammar=grammar)
-  else:
-    hg = decoder.translate(ins)
+  hg = decoder.translate(ins)
 
-  sys.stderr.write("input:\n '%s'\n"%(ins))
-  sys.stderr.write("viterbi translation:\n '%s'\n"%(hg.viterbi()))
+  sys.stderr.write( "input:\n '%s'\n"%(ins) )
+  sys.stderr.write( "viterbi translation:\n '%s'\n"%(hg.viterbi()) )
   num_nodes = 0
   for i in hg.nodes: num_nodes+=1
-  sys.stderr.write("# nodes = %s\n"%(num_nodes))
+  sys.stderr.write( "# nodes = %s\n"%(num_nodes) )
   num_edges = 0
   for i in hg.edges: num_edges+=1
-  sys.stderr.write("# edges = %s\n"%(num_edges))
-  sys.stderr.write("viterbi score = %s\n"%(round(hg.viterbi_features().dot(decoder.weights), 2)))
+  sys.stderr.write( "# edges = %s\n"%(num_edges) )
+  sys.stderr.write( "viterbi score = %s\n"%(round(hg.viterbi_features().dot(decoder.weights), 2)) )
 
-  print hg2json(hg, decoder.weights).encode('utf-8')
+  print hg2json(hg, decoder.weights)
 
 if __name__=="__main__":
   main()
-- 
cgit v1.2.3