def pre_processing_pipeline_text(X_train, X_test):
x_train_process = []
for text in X_train:
tmp = pp.to_lower_case(text)
tmp = pp.to_lower_case(tmp)
tmp = pp.substitute_thousands(tmp)
tmp = pp.fix_common_mistakes(tmp)
tmp = pp.unstack(tmp)
tmp = pp.remove_white_space(tmp)
tmp = pp.remove_punctuation(False, tmp)
tmp = pp.clean_text(False, tmp)
tmp = pp.stemming(tmp) # TODO try without it
x_train_process.append(tmp)
x_test_process = []
for text in X_test:
tmp = pp.to_lower_case(text)
tmp = pp.to_lower_case(tmp)
tmp = pp.substitute_thousands(tmp)
tmp = pp.fix_common_mistakes(tmp)
tmp = pp.unstack(tmp)
tmp = pp.remove_white_space(tmp)
tmp = pp.remove_punctuation(False, tmp)
tmp = pp.clean_text(False, tmp)
tmp = pp.stemming(tmp) # TODO try without it
x_test_process.append(tmp)
return x_train_process, x_test_process
def read_unstructure_texts(files):
texts = []
for file in files:
with open(file, encoding='utf8') as f:
data = f.read()
data = stemming(remove_stop_words(tokenize(data)))
texts.append(data)
return texts
def pre_processing_pipeline(df):
for index, row in df.iterrows():
df.loc[index, 'text'] = pp.to_lower_case(row['text'])
df.loc[index, 'text'] = pp.substitute_thousands(row['text'])
df.loc[index, 'text'] = pp.fix_common_mistakes(row['text'])
df.loc[index, 'text'] = pp.unstack(row['text'])
df.loc[index, 'text'] = pp.remove_white_space(row['text'])
df.loc[index, 'text'] = pp.remove_punctuation(False, row['text'])
df.loc[index, 'text'] = pp.clean_text(False, row['text'])
df.loc[index, 'text'] = pp.stemming(row['text']) # TODO try without it
return df
def read_to_texts(files):
texts = []
for file in files:
with open(file, encoding='utf8') as f:
data = f.read()
parsed_result = parse_paper(data)
text = ''
for section, content in parsed_result['structure'].items():
text += content
text = stemming(remove_stop_words(tokenize(text)))
texts.append(text)
return texts
def load_file(path):
texts = []
num_files = 0
num_methods = 0
num_abstract = 0
paper_attributes = {}
country_counter = []
for folder in os.listdir(path):
for file in os.listdir(os.path.join(path, folder)):
num_files += 1
if num_files <= NUM_START:
continue
parsed_result = file2text(os.path.join(path, folder, file))
method_text = get_methods(parsed_result['text'])
if method_text:
token_text = stemming(remove_stop_words(tokenize(method_text)))
texts.append(token_text)
num_methods += 1
author_countries = get_country(parsed_result['text'][:1000])
if len(author_countries) > 1:
num_abstract += 1
paper_id = file.split('.')[0]
paper_attributes[paper_id] = {'countries': author_countries}
for c in author_countries:
if c in country_counter:
country_counter[c] += 1
else:
country_counter[c] = 1
if num_files % 1000 == 0:
print(num_files, num_methods, num_abstract)
if num_files > NUM_STOP:
with open(os.path.join(os.getcwd(), 'metadata', OUTPUT_FILE), 'w') as f:
json.dump(paper_attributes, f)
print(country_counter)
return
print("finished extraction")
print(num_files, num_methods, num_abstract)
data = json.load(instances)
word_list = [] #create an array with all the words
tags = []
xy = []
for i in data['data']:
tag = i['tag']
tags.append(tag)
for user_response in i['user_responses']:
normalized = normalization(user_response)
words = tokenization(normalized)
word_list.extend(words)
xy.append(
(words, tag)) #array of user responses with the respective tags
word_list = [stemming(word) for word in word_list]
word_list = sorted(set(word_list))
print(tags)
print(word_list)
print(xy)
x_train = []
y_train = []
for (tokenized, tag) in xy:
bag = bag_of_words(tokenized, word_list)
x_train.append(bag)
tag_label = tags.index(tag)
y_train.append(tag_label)
x_train = np.array(x_train)
word_list = [] #create an array with all the words
tags = []
xy = []
for i in data['data']:
tag = i['tag']
tags.append(tag)
for user_response in i['user_responses']:
normalized = normalization(user_response)
words = tokenization(normalized)
word_list.extend(
words) # not apeend becous dont want a arraylist in a array
xy.append(
(words, tag)) #array of user responses with the respective tags
word_list = [stemming(word) for word in word_list] # to remove the symbols
word_list = sorted(set(word_list)) # to remove duplicate elements
print(tags)
print(word_list)
print(xy)
# train the data
x_train = []
y_train = []
for (tokenized, tag) in xy:
bag = bag_of_words(tokenized, word_list)
x_train.append(bag)
tag_label = tags.index(tag)
y_train.append(tag_label)
x_train = np.array(x_train)
def tokenize_text(text):
token_text = stemming(remove_stop_words(tokenize(text)))
filetered_text = remove_lf_words(token_text, 2)
return filetered_text
