def bleu_evaluation(self, candidate):
abstractive_score = None
extractive_score = None
"""first elemnt in refrences is the abstractive resume and the second one is the extractive resume"""
# if the list is not empty
if self.resume_abstractive:
abstractive_score = sentence_bleu([
tools.tokenize(
tools.text_to_string(
'./manual_resume_abstractive/' + text + '.txt',
is_resume_abstract=True))
for text in self.resume_abstractive
], candidate)
# if the list is not empty
if self.resume_extractive:
#self.clean_extractive(self.corpus)
extractive_score = sentence_bleu([
tools.tokenize(
tools.text_to_string('./axilaury_extractive/' + text +
'.txt'))
for text in self.resume_extractive
], candidate)
return (abstractive_score, extractive_score)
['meet_116', 'meet_117', 'meet_118', 'meet_119'],
['meet_120', 'meet_121', 'meet_122', 'meet_123'],
['meet_124', 'meet_125', 'meet_126', 'meet_127'],
['meet_128', 'meet_129', 'meet_130', 'meet_131'],
['meet_132', 'meet_133', 'meet_134', 'meet_135'],
['meet_136', 'meet_137', 'meet_138', 'meet_139'],
['meet_140', 'meet_141', 'meet_142', 'meet_143'],
['meet_144', 'meet_145', 'meet_146', 'meet_147'],
['meet_148', 'meet_149', 'meet_150', 'meet_151']
]
meet_list = ["block_" + str(i + 1) for i in xrange(34)]
print 'meet_list', meet_list
adress_ = "./manual_blocks/"
all_documents = [tools.text_to_string(adress_ + i + '.txt') for i in meet_list]
example = tf_idf_class(all_documents)
tfidf_representation = example.tfidf()
counter = 0
for meet in meet_list:
print '***********'
print 'meet', meet
meet_block = classification_scenario_based[counter]
file.write(str(meet_block) + '\n')
file.write('*****************' + '\n')
# file.flush()
#
# file.close()
file = open('./result/Row_results/result_relative_1.txt', 'w')
adress_ = "./manuel_corpus/"
file.write(
'this file includes the tf-idf scores for 138 scenario-based meetings' +
'\n' +
'For each meeting, the tf-idf scores are computed by comparing this meeting w.r.t other scenario-based '
+ '\n' +
'meetings esxcept the meetins with the same order in their bloks.' + '\n')
all_documents_group_1 = [
tools.text_to_string(adress_ + i + '.txt') for i in m1.itervalues()
]
all_documents_group_2 = [
tools.text_to_string(adress_ + i + '.txt') for i in m2.itervalues()
]
all_documents_group_3 = [
tools.text_to_string(adress_ + i + '.txt') for i in m3.itervalues()
]
all_documents_group_4 = [
tools.text_to_string(adress_ + i + '.txt') for i in m4.itervalues()
if i is not None
]
all_documents__1 = all_documents_group_2 + all_documents_group_3 + all_documents_group_4
all_documents__2 = all_documents_group_1 + all_documents_group_3 + all_documents_group_4
all_documents__3 = all_documents_group_2 + all_documents_group_1 + all_documents_group_4
def blocks_4_4_ami(n_words, length= None):
"""
this function extracts the first
:return:
"""
file = open(path + 'result_ami/result_4_4.txt', 'w')
adress = path + '/manual_corpus/'
if length is None:
for meet_list in classification_scenario_based:
file.write('************************' + str(meet_list) + '*************************************\n')
all_documents = [tools.text_to_string(adress + i + '.txt') for i in meet_list]
example = tf_idf_class(all_documents) # , document_0)
tfidf_representation = example.tfidf()
for meet in meet_list:
file.write(meet+ '\n')
file.write('*****************\n')
c = compare_with_AMI_results(meet)
c.get_resumes()
candidate = c.get_best_k_tfidf(n_words, tfidf_representation[meet_list.index(meet)])
file.write('candidate = ' + str(candidate)+ '\n')
file.flush()
bleu_measure = c.bleu_evaluation(candidate)
file.write('score_abstractive = ' + str(bleu_measure[0])+ '\n')
file.write('score_extractive = ' + str(bleu_measure[1])+ '\n')
file.flush()
file.write('\n')
file.flush()
else:
for meet_list in classification_scenario_based[0:length]:
file.write('************************' + str(meet_list) + '*************************************\n')
all_documents = [tools.text_to_string(adress + i + '.txt') for i in meet_list]
example = tf_idf_class(all_documents) # , document_0)
tfidf_representation = example.tfidf(length)
for meet in meet_list:
file.write(meet+ '\n')
file.write('*****************\n')
c = compare_with_AMI_results(meet)
c.get_resumes()
candidate = c.get_best_k_tfidf(n_words, tfidf_representation[meet_list.index(meet)])
file.write('candidate = ' + str(candidate)+ '\n')
file.flush()
bleu_measure = c.bleu_evaluation(candidate)
file.write('score_abstractive = ' + str(bleu_measure[0])+ '\n')
file.write('score_extractive = ' + str(bleu_measure[1])+ '\n')
file.flush()
file.write('\n')
file.flush()
file.close()
return
def blocks_ami(n_words, length = None):
#print('we are inside')
file = open(path + 'result_ami/result_4_block_scen.txt', 'w')
adress_ = path + "manual_blocks/"
meet_list = ["block_" + str(i + 1) for i in xrange(34)]
all_documents = [tools.text_to_string(adress_ + i + '.txt') for i in meet_list]
example = tf_idf_class(all_documents)
counter = 0
if length is None:
tfidf_representation = example.tfidf()
for meet in meet_list:
#print '***********'
#print 'meet', meet
meet_block = classification_scenario_based[counter]
file.write(str(meet_block) + '\n')
file.write('*****************' + '\n')
file.flush()
c = compare_with_AMI_results(meet_block)
c.get_resumes()
candidate = c.get_best_k_tfidf(n_words, tfidf_representation[counter])
file.write('candidate = ' + str(candidate) + '\n')
file.flush()
bleu_measure = c.bleu_evaluation(candidate)
file.write('score_abstractive = ' + str(bleu_measure[0]) + '\n')
file.write('score_extractive = ' + str(bleu_measure[1]) + '\n')
file.write('\n')
file.flush()
counter += 1
else:
#print('length is not none')
tfidf_representation = example.tfidf(length)
for meet in meet_list[0:length]:
#print '***********'
#print 'meet', meet
meet_block = classification_scenario_based[counter]
file.write(str(meet_block) + '\n')
file.write('*****************' + '\n')
file.flush()
c = compare_with_AMI_results(meet_block)
c.get_resumes()
candidate = c.get_best_k_tfidf(n_words, tfidf_representation[counter])
file.write('candidate = ' + str(candidate) + '\n')
file.flush()
bleu_measure = c.bleu_evaluation(candidate)
file.write('score_abstractive = ' + str(bleu_measure[0]) + '\n')
file.write('score_extractive = ' + str(bleu_measure[1]) + '\n')
file.write('\n')
file.flush()
counter += 1
file.close()
['meet_116', 'meet_117', 'meet_118', 'meet_119'],
['meet_120', 'meet_121', 'meet_122', 'meet_123'],
['meet_124', 'meet_125', 'meet_126', 'meet_127'],
['meet_128', 'meet_129', 'meet_130', 'meet_131'],
['meet_132', 'meet_133', 'meet_134', 'meet_135'],
['meet_136', 'meet_137', 'meet_138', 'meet_139'],
['meet_140', 'meet_141', 'meet_142', 'meet_143'],
['meet_144', 'meet_145', 'meet_146', 'meet_147'],
['meet_148', 'meet_149', 'meet_150', 'meet_151']
]
for meet_list in classification_scenario_based:
file.write('************************' + str(meet_list) +
'*************************************\n')
all_documents = [
tools.text_to_string(adress + i + '.txt') for i in meet_list
]
example = tf_idf_class(all_documents) # , document_0)
tfidf_representation = example.tfidf()
for meet in meet_list:
file.write(meet + '\n')
file.write('*****************\n')
c = compare_with_AMI_results(meet)
c.get_resumes()
candidate = c.get_best_k_tfidf(
20, tfidf_representation[meet_list.index(meet)])
file.write('candidate = ' + str(candidate) + '\n')