def main():
logging.basicConfig(format='%(asctime)s : %(levelname)s : %(message)s', level=logging.INFO)
data_directory = '../../experiments/ptr/'
# Load data
logging.info('Loading ptr_data')
with open(os.path.join(data_directory, 'ptr_data.p')) as f:
ptr_data = cPickle.load(f)
logging.info('Loading ptr_params')
with open(os.path.join(data_directory, 'ptr_params.p')) as f:
ptr_params = cPickle.load(f)
pfst_params = PTRExperiment.PFSTParams(ptr_data, ptr_params)
# Copy new parameters
#logging.info('Creating new_ptr_params')
#new_ptr_params = copy.deepcopy(ptr_params)
#doc_ct = 0
logging.info('Loading sentence counter')
with open(os.path.join(data_directory, 'ptr_sentence_counter.p')) as f:
sentence_counter = cPickle.load(f)
sorted_sentence_counts = sorted(sentence_counter.items(), key=lambda x:x[1], reverse=True)
del sentence_counter
alignment_library = {}
assignment_library = {}
# Inference
logging.info('Starting inference')
sentence_ct = 0
for wordids, count in sorted_sentence_counts:
sentence_ct += 1
if sentence_ct % 1000 == 0:
logging.info('%s sentences done, current count: %s' % (sentence_ct, count))
with open(os.path.join(data_directory, 'ptr_assignment_library.p'), 'w') as f:
cPickle.dump(assignment_library, f)
with open(os.path.join(data_directory, 'ptr_alignment_library.p'), 'w') as f:
cPickle.dump(alignment_library, f)
if wordids in assignment_library:
pass
elif wordids in ptr_params.ideas.wordids_idx_dict:
assignment_library[wordids] = ptr_params.ideas.wordids_idx_dict[wordids]
else:
top_ideas = metrics.get_top_candidates(wordids, ptr_params, num_top_candidates=20, jaccard_cutoff=0.5)
assignment = metrics.get_assignment(wordids, top_ideas, ptr_data, ptr_params, pfst_params, alignment_library)
assignment_library[wordids] = assignment
logging.info('Writing new_ptr_params')
with open(os.path.join(data_directory, 'ptr_assignment_library.p', 'w')) as f:
cPickle.dump(assignment_library, f)
def main():
logging.basicConfig(format='%(asctime)s : %(levelname)s : %(message)s', level=logging.INFO)
data_directory = '../../experiments/ptr/'
# Load data
logging.info('Loading ptr_data')
with open(os.path.join(data_directory, 'ptr_data.p')) as f:
ptr_data = cPickle.load(f)
logging.info('Loading ptr_params')
with open(os.path.join(data_directory, 'ptr_params.p')) as f:
ptr_params = cPickle.load(f)
pfst_params = PTRExperiment.PFSTParams(ptr_data, ptr_params)
# Copy new parameters
logging.info('Creating new_ptr_params')
new_ptr_params = copy.deepcopy(ptr_params)
doc_ct = 0
logging.info('Loading assignment and alignment libraries')
with open(os.path.join(data_directory, 'ptr_assignment_library.p')) as f:
assignment_library = cPickle.dump(assignment_library, f)
with open(os.path.join(data_directory, 'ptr_alignment_library.p')) as f:
alignment_library = cPickle.dump(alignment_library, f)
# Inference
logging.info('Starting inference')
for docid, wordids in ptr_data.docid_wordids.iteritems():
doc_ct += 1
if doc_ct % 1000 == 0:
logging.info('%s documents done' % doc_ct)
#with open(os.path.join(data_directory, 'ptr_params.greedy.001.p', 'w')) as f:
# cPickle.dump(new_ptr_params, f)
for p, assignment in ptr_params.docid_partitions[docid].iteritems():
sent_wordids = tuple(wordids[p[0]:p[1]])
if sent_wordids in new_ptr_params.ideas.wordids_idx_dict:
new_ptr_params.docid_partitions[docid][p] = new_ptr_params.ideas.wordids_idx_dict[sent_wordids]
else:
top_ideas = metrics.get_top_candidates(sent_wordids, new_ptr_params, num_top_candidates=20, jaccard_cutoff=0.5)
assignment = metrics.get_assignment(sent_wordids, top_ideas, ptr_data, new_ptr_params, pfst_params, alignment_library)
new_ptr_params.docid_partitions[docid][p] = assignment
logging.info('Writing new_ptr_params')
with open(os.path.join(data_directory, 'ptr_params.greedy.001.p', 'w')) as f:
cPickle.dump(new_ptr_params, f)
