diff options
author | Victor Chahuneau <vchahune@cs.cmu.edu> | 2012-08-10 19:03:38 -0400 |
---|---|---|
committer | Victor Chahuneau <vchahune@cs.cmu.edu> | 2012-08-10 19:03:38 -0400 |
commit | c3d0668c17f45247e1fec6ffe31b807fbbba6674 (patch) | |
tree | e8d673346ddbacfab81de2204be1fea99a0aecbb /python/examples | |
parent | b6474b5cdbf870725371b32670c9dc28671e394c (diff) |
[python] Examples directory including Rampion
Diffstat (limited to 'python/examples')
-rw-r--r-- | python/examples/rampion.py | 77 | ||||
-rw-r--r-- | python/examples/test.py | 70 |
2 files changed, 147 insertions, 0 deletions
diff --git a/python/examples/rampion.py b/python/examples/rampion.py new file mode 100644 index 00000000..66d89a61 --- /dev/null +++ b/python/examples/rampion.py @@ -0,0 +1,77 @@ +import argparse +import logging +from itertools import izip +import cdec, cdec.score + +def evaluate(hyp, ref): + """ Compute BLEU score for a set of hypotheses+references """ + return sum(cdec.score.BLEU(r).evaluate(h) for h, r in izip(hyp, ref)).score + +T1, T2, T3 = 5, 10, 20 # number of iterations (global, CCCP, SSD) +K = 500 # k-best list size +C = 1 # regularization coefficient +eta = 1e-4 # step size +cost = lambda c: 10 * (1 - c.score) # cost definition + +def rampion(decoder, sources, references): + # Empty k-best lists + cs = [cdec.score.BLEU(refs).candidate_set() for refs in references] + # Weight vector -> sparse + w = decoder.weights.tosparse() + w0 = w.copy() + + N = len(sources) + for t in range(T1): + logging.info('Iteration {0}: translating...'.format(t+1)) + # Get the hypergraphs and extend the k-best lists + hgs = [] + for src, candidates in izip(sources, cs): + hg = decoder.translate(src) + hgs.append(hg) + candidates.add_kbest(hg, K) + # BLEU score for the previous iteration + score = evaluate((hg.viterbi() for hg in hgs), references) + logging.info('BLEU: {:.2f}'.format(100 * score)) + logging.info('Optimizing...') + for _ in range(T2): + # y_i^+, h_i^+; i=1..N + plus = [max(candidates, key=lambda c: w.dot(c.fmap) - cost(c)).fmap + for candidates in cs] + for _ in range(T3): + for fp, candidates in izip(plus, cs): + # y^-, h^- + fm = max(candidates, key=lambda c: w.dot(c.fmap) + cost(c)).fmap + # update weights (line 11-12) + w += eta * ((fp - fm) - C/N * (w - w0)) + logging.info('Updated weight vector: {0}'.format(dict(w))) + # Update decoder weights + for fname, fval in w: + decoder.weights[fname] = fval + +def main(): + logging.basicConfig(level=logging.INFO, format='%(message)s') + + parser = argparse.ArgumentParser() + parser.add_argument('-c', '--config', help='cdec config', required=True) + parser.add_argument('-w', '--weights', help='initial weights', required=True) + parser.add_argument('-r', '--reference', help='reference file', required=True) + parser.add_argument('-s', '--source', help='source file', required=True) + args = parser.parse_args() + + with open(args.config) as fp: + config = fp.read() + + decoder = cdec.Decoder(config) + decoder.read_weights(args.weights) + with open(args.reference) as fp: + references = fp.readlines() + with open(args.source) as fp: + sources = fp.readlines() + assert len(references) == len(sources) + rampion(decoder, sources, references) + + for fname, fval in sorted(dict(decoder.weights).iteritems()): + print('{0}\t{1}'.format(fname, fval)) + +if __name__ == '__main__': + main() diff --git a/python/examples/test.py b/python/examples/test.py new file mode 100644 index 00000000..eb9e6a95 --- /dev/null +++ b/python/examples/test.py @@ -0,0 +1,70 @@ +#coding: utf8 +import cdec +import gzip + +weights = '../tests/system_tests/australia/weights' +grammar_file = '../tests/system_tests/australia/australia.scfg.gz' + +# Load decoder width configuration +decoder = cdec.Decoder(formalism='scfg') +# Read weights +decoder.read_weights(weights) + +print dict(decoder.weights) + +# Read grammar +with gzip.open(grammar_file) as f: + grammar = f.read() + +# Input sentence +sentence = u'澳洲 是 与 北韩 有 邦交 的 少数 国家 之一 。' +print ' Input:', sentence.encode('utf8') + +# Decode +forest = decoder.translate(sentence, grammar=grammar) + +# Get viterbi translation +print 'Output[0]:', forest.viterbi().encode('utf8') +f_tree, e_tree = forest.viterbi_trees() +print ' FTree[0]:', f_tree.encode('utf8') +print ' ETree[0]:', e_tree.encode('utf8') +print 'LgProb[0]:', forest.viterbi_features().dot(decoder.weights) + +# Get k-best translations +kbest = zip(forest.kbest(5), forest.kbest_trees(5), forest.kbest_features(5)) +for i, (sentence, (f_tree, e_tree), features) in enumerate(kbest, 1): + print 'Output[%d]:' % i, sentence.encode('utf8') + print ' FTree[%d]:' % i, f_tree.encode('utf8') + print ' ETree[%d]:' % i, e_tree.encode('utf8') + print ' FVect[%d]:' % i, dict(features) + +# Sample translations from the forest +for sentence in forest.sample(5): + print 'Sample:', sentence.encode('utf8') + +# Get feature vector for 1best +fsrc = forest.viterbi_features() + +# Feature expectations +print 'Feature expectations:', dict(forest.inside_outside()) + +# Reference lattice +lattice = ((('australia',0,1),),(('is',0,1),),(('one',0,1),),(('of',0,1),),(('the',0,4),('a',0,4),('a',0,1),('the',0,1),),(('small',0,1),('tiny',0,1),('miniscule',0,1),('handful',0,2),),(('number',0,1),('group',0,1),),(('of',0,2),),(('few',0,1),),(('countries',0,1),),(('that',0,1),),(('has',0,1),('have',0,1),),(('diplomatic',0,1),),(('relations',0,1),),(('with',0,1),),(('north',0,1),),(('korea',0,1),),(('.',0,1),),) + +lat = cdec.Lattice(lattice) +assert (lattice == tuple(lat)) + +# Intersect forest and lattice +assert forest.intersect(lat) + +# Get best synchronous parse +f_tree, e_tree = forest.viterbi_trees() +print 'FTree:', f_tree.encode('utf8') +print 'ETree:', e_tree.encode('utf8') + +# Compare 1best and reference feature vectors +fref = forest.viterbi_features() +print dict(fsrc - fref) + +# Prune hypergraph +forest.prune(density=100) |