def loadTruth(template_data_path, exclude_ria=[], targets=None):
'''
If the ground truth matches aren't already saved, extract the ground truth from all prior RIA template2 and
create the target matches dictionary. Save this for future use. If a certain RIA is to be excluded, create
a new dictionary that won't be saved.
Args:
template_data_path (string) : Directory to the RIA templates documents.
exclude_ria (list[string]) : List of files to be excluded from RIA data extraction.
targets (dict) : Target matches dictionary of just target descriptions.
Returns:
target_matches (dict) : Dictionary of ground truth matches for each target.
'''
if targets:
target_matches = targets
elif len(exclude_ria) > 0:
development_matches = parseundp.extract_template_data(
template_data_path, exclude_ria)
target_matches = parseundp.create_target_dictionary(
development_matches)
else:
try:
shelf = shelve.open('undp')
target_matches = shelf['targets']
shelf.close()
except:
shelf.close()
development_matches = parseundp.extract_template_data(
template_data_path)
target_matches = parseundp.create_target_dictionary(
development_matches)
shelf = shelve.open('undp')
shelf['targets'] = target_matches
shelf.close()
return target_matches
vec = 1
for par_vec in par_vecs:
for i in range(len(all_policy_documents)):
exclude_ria = all_exclude_ria[i]
policy_documents = all_policy_documents[i]
target_matches = getTargetDoc(template_data_path, exclude_ria)
targs, targ_vecs, sents = getInfo(par_vec, target_matches)
print(exclude_ria[0][:-5]+str(vec))
score_dict = ria(documents_path, policy_documents, par_vec, sents, targ_vecs, targs)
combined_prior_matches[exclude_ria[0][:-5]+str(vec)] = [score_dict]
vec += 1
i = 0
for key, val in combined_prior_matches.items():
exclude_test = [file for file in os.listdir(template_data_path) if file not in all_exclude_ria[i]]
test_development_matches = parseundp.extract_template_data(template_data_path, exclude_test)
test_target_matches = parseundp.create_target_dictionary(test_development_matches)
print(key, all_exclude_ria[i])
match_by_sent = evaluateByTarget(val[0], test_target_matches, 301)
combined_prior_matches[key].append(match_by_sent)
avg_new = avgMatches(match_by_sent, test_target_matches, 301)
combined_prior_matches[key].append(avg_new)
i+=1
if i % 5 == 0:
i = 0
combined_prior_matches['liberia_google'] = combined_prior_matches.pop('liberia1')
combined_prior_matches['liberia_nbow'] = combined_prior_matches.pop('liberia2')
combined_prior_matches['liberia_tfidf'] = combined_prior_matches.pop('liberia3')