def save_value(self, path, text, coverage_list, relative_matrix,
clues_list, entities_list):
ftools.check_filename(path)
save_dict = {}
save_dict['#$#'.join(text)] = [
coverage_list, relative_matrix, clues_list, entities_list
]
tools.save_object(save_dict, path)
def train(traindata, savepath=Dir.res + "/parameter/words_vector/w2v.model"):
ftools.check_filename(savepath)
model = Word2Vec(sentences=traindata,
size=200,
window=5,
min_count=3,
workers=4)
model.save(savepath)
def save_value(self, path, key, coverage_list, relative_matrix, clues_list,
entities_list):
ftools.check_filename(path)
save_dict = {}
save_dict[key] = [
coverage_list, relative_matrix, clues_list, entities_list
]
tools.save_object(save_dict, path)
def PythonROUGE(guess_summ_list, ref_summ_list, ngram_order=2):
""" Wrapper function to use ROUGE from Python easily. """
# even though we ask that the first argument is a list,
# if it is a single string we can handle it
if type(guess_summ_list) == str:
temp = list()
temp.append(ref_summ_list)
guess_summ_list = temp
del temp
# even though we ask that the second argument is a list of lists,
# if it is a single string we can handle it
# if type(ref_summ_list[0]) == str:
# temp = list()
# temp.append(ref_summ_list)
# ref_summ_list = temp
# del temp
# this is the path to your ROUGE distribution
ROUGE_path = Dir.res + "/RELEASE-1.5.5/ROUGE-1.5.5.pl"
data_path = Dir.res + "/RELEASE-1.5.5/data"
# these are the options used to call ROUGE
# feel free to edit this is you want to call ROUGE with different options
options = '-a -t 1 -m -l 140 -n ' + str(ngram_order)
# this is a temporary XML file which will contain information
# in the format ROUGE uses
xml_path = Dir.res + '/Temp/temp.xml'
tools.check_filename(xml_path)
xml_file = open(xml_path, 'w')
xml_file.write('<ROUGE-EVAL version="1.0">\n')
for guess_summ_index, guess_summ_file in enumerate(guess_summ_list):
xml_file.write('<EVAL ID="' + str(guess_summ_index + 1) + '">\n')
create_xml(xml_file, guess_summ_file, ref_summ_list[guess_summ_index])
xml_file.write('</EVAL>\n')
xml_file.write('</ROUGE-EVAL>\n')
xml_file.close()
# this is the file where the output of ROUGE will be stored
ROUGE_output_path = Dir.res + '/Temp/ROUGE_result.txt'
# this is where we run ROUGE itself
exec_command = ROUGE_path + ' -e ' + data_path + ' ' + options + ' -x ' + xml_path + ' > ' + ROUGE_output_path
# print(exec_command)
os.system(exec_command)
# here, we read the file with the ROUGE output and
# look for the recall, precision, and F-measure scores
recall_list = list()
precision_list = list()
F_measure_list = list()
with open(ROUGE_output_path, 'r') as ROUGE_output_file:
for n in range(ngram_order):
ROUGE_output_file.seek(0)
for line in ROUGE_output_file:
match = re.findall(
'X ROUGE-' + str(n + 1) + ' Average_R: ([0-9.]+)', line)
if match != []:
recall_list.append(float(match[0]))
match = re.findall(
'X ROUGE-' + str(n + 1) + ' Average_P: ([0-9.]+)', line)
if match != []:
precision_list.append(float(match[0]))
match = re.findall(
'X ROUGE-' + str(n + 1) + ' Average_F: ([0-9.]+)', line)
if match != []:
F_measure_list.append(float(match[0]))
# with open(ROUGE_output_path, 'r') as ROUGE_output_file:
# content = ROUGE_output_file.read()
# match = re.findall('X ROUGE-L Average_R: ([0-9.]+)', content)
# if match != []:
# recall_list.append(float(match[0]))
# match = re.findall('X ROUGE-L Average_P: ([0-9.]+)', content)
# if match != []:
# precision_list.append(float(match[0]))
# match = re.findall('X ROUGE-L Average_F: ([0-9.]+)', content)
# if match != []:
# F_measure_list.append(float(match[0]))
ROUGE_output_file.close()
# remove temporary files which were created
# os.remove(xml_path)
# os.remove(ROUGE_output_path)
return (recall_list, precision_list, F_measure_list)
