From 9c7b9953567206c10a7eb0dc8a93f6b35c6122b8 Mon Sep 17 00:00:00 2001 From: Patrick Simianer Date: Wed, 14 Jan 2015 19:09:16 +0100 Subject: cdec2json.py --- example/cdec2json.py | 100 +++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 100 insertions(+) create mode 100755 example/cdec2json.py diff --git a/example/cdec2json.py b/example/cdec2json.py new file mode 100755 index 0000000..cd2a846 --- /dev/null +++ b/example/cdec2json.py @@ -0,0 +1,100 @@ +#!/usr/bin/env python2 + +import cdec +import sys, argparse +import json +import gzip + + +#FIXME new format +# strings? +# map? +def hg2json(hg, weights): + """ + output a JSON representation of a cdec hypegraph + """ + res = '' + res += "{\n" + 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" + res += '"nodes":'+"\n" + res += "[\n" + a = [] + a.append( '{ "id":0, "symbol":"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])) + res += ",\n".join(a)+"\n" + res += "],\n" + 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)))) + # f + #xs = ' "f":{' + #b = [] + #for j in i.feature_values: + # b.append( '"%s":%s'%(j[0], j[1]) ) + #xs += ", ".join(b) + #xs += "}," + # tails + if len(list(i.tail_nodes)) > 0: + s += ', "tails":[ %s ],'%(",".join([str(n.id+1) for n in i.tail_nodes])) + else: + s += ', "tails":[ 0 ],' + #s += xs + s += ' "score":%s }'%(i.prob) + a.append(s) + res += ",\n".join(a)+"\n" + res += "]\n" + res += "}\n" + return res + +def main(): + 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) + + 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) ) + 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)) ) + + print hg2json(hg, decoder.weights) + + +if __name__=="__main__": + main() + -- cgit v1.2.3