def main():
# loading training feature data
dir_path = os.path.dirname(os.path.realpath(__file__))
data = json.load(open(dir_path + "/../../../feats_all_ng_dep_prior.json"))
output_file = open("window_size_output_file.tsv", "w")
query_folder = "/home/smsarwar/PycharmProjects/civilian_killing/data/query_configurations/iterative_query_lr_dir"
# allfeats will hold all the features in the training data and their weights
onlyfiles = [
os.path.join(query_folder, file) for file in os.listdir(query_folder)
if os.path.isfile(os.path.join(query_folder, file))
]
# print onlyfiles
#results = {}
number_of_files = 50
#query_terms = np.arange(10, 41, 10)
window_sizes = np.asarray([2, 3, 5, 10, 15, 20, 25, 30])
res = np.zeros((4, len(window_sizes)))
for idx, file in enumerate(onlyfiles[:number_of_files]):
#results[file] = {}
print file
lines = open(file).readlines()
lines = [line.strip() for line in lines]
line = lines[20]
sentence_ids = line.strip().split()
allfeats = []
#loading all the features
for item in data.keys():
if item in sentence_ids:
feats = {}
for feature in data[item]:
feats[feature] = 1
allfeats.append(feats)
#prior feats will contain all the features in the training dataset
prior_feats = {}
for item in data.keys():
prior_feats = merge_two_dicts(prior_feats, data[item])
allfeats.append(prior_feats)
features, weights = transform_data_and_train(allfeats)
name_set = load_gold()
id_dict, train_id_dict = load_dictionary()
#query_terms = np.arange(1, 1002, 100)
#number_of_query_terms = min(len(features), 200)
#query_terms = np.arange(1, 1002, 100)
terms = []
p10 = []
p20 = []
p30 = []
num_civilian = []
query_number = 0
number_of_query_terms = min(len(features), 200)
for window_size in window_sizes:
file_temp = open("indri_lr_query_file.xml", "w")
file_temp.write(get_query_prefix())
query = prepare_indri_query_top_k_features(
features,
weights,
query_number,
window_size=window_size,
topk=number_of_query_terms,
prf=False)
file_temp.write(query + '\n')
query_number += 1
file_temp.write(get_query_suffix())
file_temp.close()
print "now querying"
cmd = "IndriRunQuery indri_lr_query_file.xml"
output = sp.check_output(cmd.split())
# #print output
precs10, precs20, precs30, number_of_civilians_found, name_of_civilians = get_metric_values(
output.split("\n"), id_dict, name_set, train_id_dict)
# #results.append((number_of_query_terms, precs10, precs20, precs30, number_of_civilians_found))
terms.append(window_size)
p10.append(precs10)
p20.append(precs20)
p30.append(precs30)
num_civilian.append(number_of_civilians_found)
res[0] += np.asarray(p10)
res[1] += np.asarray(p20)
res[2] += np.asarray(p30)
res[3] += np.asarray(num_civilian)
# print res[0]/(idx+1)
# print res[1]/(idx+1)
# print res[2]/(idx+1)
# print res[3]/(idx+1)
res_P10 = ''
res_P20 = ''
res_P30 = ''
res_num_civilian = ''
for index in np.arange(len(window_sizes)):
res_P10 += str(res[0][index] / (idx + 1)) + "\t"
res_P20 += str(res[1][index] / (idx + 1)) + "\t"
res_P30 += str(res[2][index] / (idx + 1)) + "\t"
res_num_civilian += str(res[3][index] / (idx + 1)) + "\t"
print window_sizes
print res_P10.strip()
print res_P20.strip()
print res_P30.strip()
print res_num_civilian.strip()
output_file.write(str(window_sizes) + "\n")
output_file.write(res_P10.strip() + "\n")
output_file.write(res_P20.strip() + "\n")
output_file.write(res_P30.strip() + "\n")
output_file.write(res_num_civilian.strip() + "\n")
output_file.write(
"-----------------------------------------------------\n")
output_file.close()
def main():
plotting = False
# loading training feature data
dir_path = os.path.dirname(os.path.realpath(__file__))
data = json.load(open(dir_path + "/../../../feats_all_ng_dep_prior.json"))
query_folder = "/home/smsarwar/PycharmProjects/civilian_killing/data/query_configurations/iterative_query_lr_dir"
# allfeats will hold all the features in the training data and their weights
onlyfiles = [
os.path.join(query_folder, file) for file in os.listdir(query_folder)
if os.path.isfile(os.path.join(query_folder, file))
]
# print onlyfiles
#results = {}
number_of_files = 50
results = np.zeros((30, 4))
for idx, file in enumerate(onlyfiles):
#results[file] = {}
query_number = 1
file_temp = open("indri_lr_query_file.xml", "w")
file_temp.write(get_query_prefix())
for line in open(file):
sentence_ids = line.strip().split()
allfeats = []
#loading all the features
for item in data.keys():
if item in sentence_ids:
feats = {}
for feature in data[item]:
feats[feature] = 1
allfeats.append(feats)
#prior feats will contain all the features in the training dataset
prior_feats = {}
for item in data.keys():
prior_feats = merge_two_dicts(prior_feats, data[item])
#print 'length of the feature set ' + str(len(prior_feats))
allfeats.append(prior_feats)
features, weights = transform_data_and_train(allfeats)
name_set = load_gold()
id_dict, train_id_dict = load_dictionary()
#query_terms = np.arange(1, 1002, 100)
#query_terms = np.arange(1, 10, 10)
number_of_query_terms = min(len(features), 200)
#query_terms = [5]
terms = []
p10 = []
p20 = []
p30 = []
num_civilian = []
query = prepare_indri_query_top_k_features(
features,
weights,
query_number,
topk=number_of_query_terms,
prf=False)
file_temp.write(query + '\n')
#print(query + "\n")
query_number += 1
file_temp.write(get_query_suffix())
file_temp.close()
# print "now querying"
# cmd = "IndriRunQuery indri_lr_query_file.xml"
# output = sp.check_output(cmd.split())
# run_file = open("../../../data/runs/iterative_lr_ir_run_dir/" + os.path.basename(file) + ".run", "w")
# run_file.write(output)
# run_file.close()
# #print output
# precs10, precs20, precs30, number_of_civilians_found, name_of_civilians = get_metric_values(output.split("\n"), id_dict, name_set, train_id_dict)
# #results.append((number_of_query_terms, precs10, precs20, precs30, number_of_civilians_found))
# terms.append(number_of_query_terms)
# p10.append(precs10)
# p20.append(precs20)
# p30.append(precs30)
# num_civilian.append(number_of_civilians_found)
# results[index][0]+=precs10
# results[index][1]+= precs20
# results[index][2]+= precs30
# results[index][3]+= number_of_civilians_found
# print results[index]/(idx+1)
# index+=1
# file = open("2.xml.run", 'w+')
# for i in np.arange(30):
# file.write(output)
# file.close()
#
# cmd = "mv 2.xml.run /home/smsarwar/PycharmProjects/civilian_killing/data/runs/iterative_lr_ir_run_dir/"
# output = sp.check_output(cmd.split())
# #print "querying done"
# #print output
#
# cmd = "python /home/smsarwar/PycharmProjects/civilian_killing/src/eval/entity_level_evaluation.py"
# output = sp.check_output(cmd.split())
# print "evaluation done"
# print output
#print p10
#print p20
#print p30
#print num_civilian
print results
def main():
# loading training feature dataa
dir_path = os.path.dirname(os.path.realpath(__file__))
data = json.load(open(dir_path + "/../../../feats_all_ng_dep_prior.json"))
# allfeats will hold all the features in the training data and their weights
allfeats = []
print len(data.keys())
#loading all the features
for item in data.keys():
feats = {}
for feature in data[item]:
feats[feature] = 1
allfeats.append(feats)
#prior feats will contain all the features in the training dataset
prior_feats = {}
for item in data.keys():
prior_feats = merge_two_dicts(prior_feats, data[item])
print 'length of the feature set ' + str(len(prior_feats))
allfeats.append(prior_feats)
features, weights = transform_data_and_train(allfeats)
name_set = load_gold()
id_dict, train_id_dict = load_dictionary()
query_terms = np.arange(1, 1002, 100)
terms = []
p10 = []
p20 = []
p30 = []
num_civilian = []
for number_of_query_terms in query_terms:
file = open("indri_lr_query_file.xml", "w+")
file.write(prepare_indri_query_top_k_features(features, weights, topk=number_of_query_terms))
file.close()
print "now querying"
cmd = "IndriRunQuery indri_lr_query_file.xml"
output = sp.check_output(cmd.split())
#print output
precs10, precs20, precs30, number_of_civilians_found, name_of_civilians = get_metric_values(output.split("\n"), id_dict, name_set, train_id_dict)
#results.append((number_of_query_terms, precs10, precs20, precs30, number_of_civilians_found))
terms.append(number_of_query_terms)
p10.append(precs10)
p20.append(precs20)
p30.append(precs30)
num_civilian.append(number_of_civilians_found)
# file = open("2.xml.run", 'w+')
# for i in np.arange(30):
# file.write(output)
# file.close()
#
# cmd = "mv 2.xml.run /home/smsarwar/PycharmProjects/civilian_killing/data/runs/iterative_lr_ir_run_dir/"
# output = sp.check_output(cmd.split())
# #print "querying done"
# #print output
#
# cmd = "python /home/smsarwar/PycharmProjects/civilian_killing/src/eval/entity_level_evaluation.py"
# output = sp.check_output(cmd.split())
# print "evaluation done"
# print output
print p10
print p20
print p30
print num_civilian
f, axarr = plt.subplots(2, 2)
# axarr[0, 0].plot(terms, p10)
# axarr[0, 0].set_title('Precision@10')
# axarr[0, 1].scatter(terms, p20)
# axarr[0, 1].set_title('Precision@20')
# axarr[1, 0].plot(terms, p30)
# axarr[1, 0].set_title('Precision@30')
# axarr[1, 1].scatter(terms, number_of_civilians_found)
# axarr[1, 1].set_title('\#civilians (top-1000)')
# # Fine-tune figure; hide x ticks for top plots and y ticks for right plots
# plt.setp([a.get_xticklabels() for a in axarr[0, :]], visible=False)
# plt.setp([a.get_yticklabels() for a in axarr[:, 1]], visible=False)
#print terms
axarr[0, 0].plot(terms, p10, marker='o', markerfacecolor='black', markersize=6, color='darkgray', linewidth=2,
label='P@10', linestyle='-.')
axarr[0, 0].set_title('Precision@10')
axarr[0, 1].plot(terms, p20, marker='v', markerfacecolor='blue', markersize=6, color='gray', linewidth=2,
label='P@20', linestyle='--')
axarr[0, 1].set_title('Precision@20')
axarr[1, 0].plot(terms, p30, marker='x', markerfacecolor='red', markersize=6, color='black', linewidth=2,
label='P@30', linestyle=':')
axarr[1, 0].set_title('Precision@30')
axarr[1, 1].plot(terms, num_civilian, marker='s', markerfacecolor='red', markersize=6, color='black', linewidth=2,
label='P@30', linestyle=':')
axarr[1, 1].set_title('Relevant entities')
f.text(0.5, 0.005 , 'Number of Features in Query', ha='center', fontsize=10)
#plt.xlabel('Number of Examples in Query')
# plt.ylabel('Precision@100')
# plt.legend()
# plt.savefig("p@100")
# plt.show()
# f.subplots_adjust(hspace=0.3)
plt.show()