def get_onehot_vector(sent):
"""
convert sentecne to vector
:return: list
"""
try:
return_vector = []
embeddings = np.zeros([30])
idx = ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9', '-', ' ']
num_reg = re.compile("[0-9- ]")
if (type(sent) not in [type('str'), type([])]):
raise Exception("input must be str")
if (type(sent) == type([])):
sent = sent[0]
for char in sent:
vector_a = np.copy(embeddings)
vector_b = np.copy(embeddings)
vector_c = np.copy(embeddings)
vector_d = np.copy(embeddings)
if (num_reg.match(char) == None and hangul.is_hangul(char)):
anl = hangul.separate(char)
vector_a[anl[0] if anl[0] > 0 else 0] = 1
vector_b[anl[1] if anl[1] > 0 else 0] = 1
vector_c[anl[2] if anl[2] > 0 else 0] = 1
elif (num_reg.match(char)):
vector_d[idx.index(char)] = 1
return_vector.append(
np.append(vector_a, [vector_b, vector_c, vector_d]))
return np.array(return_vector)
except Exception as e:
print("error on get_onehot_vector : {0}".format(e))
def String_to_Token_List(string):
#English or special char....
if re.search(r'[가-힣\d\s\.,\?!]+', string) is None or re.search(
r'[가-힣\d\s\.,\?!]+', string).group() != string:
print(string)
return False
regex_DtoS = r'(?:^|[^\d])(\d{4})(?:$|[^\d])'
string = re.sub(
regex_DtoS,
lambda x: re.sub(r'\d{4}', lambda y: Number_to_String(y.group()),
x.group()), string)
regex_CtoS1 = r"([+-]?\d[\d,]*)[\.]?\d*"
regex_CtoS2 = r"(시|명|가지|살|마리|포기|송이|수|톨|통|점|개|벌|척|채|다발|그루|자루|줄|켤레|그릇|잔|마디|상자|사람|곡|병|판)"
string = re.sub(
regex_CtoS1 + regex_CtoS2, lambda x: re.sub(
regex_CtoS1, lambda y: Count_Number(int(y.group())), x.group()),
string)
regex_NtoS = r"([+-]?\d[\d,]*)[\.]?\d*"
string = re.sub(regex_NtoS, lambda x: Read_Number(int(x.group())), string)
token_List = []
token_List.append(0)
#<EOS>
for char in string:
if char == " ":
token_List.append(2)
continue
elif char == ".":
token_List.append(71)
continue
elif char == ",":
token_List.append(72)
continue
elif char == "?":
token_List.append(73)
continue
elif char == "!":
token_List.append(74)
continue
elif hangul.is_hangul(char):
onset, nucleus, coda = hangul.separate(char)
onset += 3
nucleus += 3 + 19
coda += 3 + 19 + 21
token_List.extend([onset, nucleus, coda])
else:
raise Exception("Not handled letter")
token_List.append(1)
#<EOE>
return token_List
def get_jamos(character):
if hangul.is_hangul(character):
character_jamos = decompose_character(character, final_char=True)
elif character in string.punctuation:
character_jamos = ['.', '.', '.']
elif character.isdigit():
character_jamos = ['0', '0', '0']
elif character.isalpha():
character_jamos = ['a', 'a', 'a']
else:
character_jamos = ['x', 'x', 'x']
return character_jamos
def convert_hangul_to_index(string):#, size): #string = unicode(string) list = []#np.ndarray([size, 3]) for i in range(len(string)): #exception if not hangul.is_hangul(string[i]): continue char3 = hangul.separate(string[i]) idx = char3[0] + char3[1] * FirNum + char3[2] * FirNum * SecNum list.append([idx]) if len(list)==0: list.append([ClassNum - 1]) return np.array(list)
def convert_index_to_hangul(list): str = '' #list = list.view(-1, 3) for i in range(len(list)): remain1 = list[i] char3 = remain1 // (FirNum * SecNum) remain2 = remain1 - char3 * FirNum * SecNum char2 = remain2 // FirNum remain3 = remain2 - char2 * FirNum char1 = remain3 hg = hangul.synthesize(char1, char2, char3) if hangul.is_hangul(hg): str = str + hg return str
def get_onehot_vector(self, sent):
"""
convert sentecne to vector
:return: list
"""
try:
return_vector = []
embeddings = np.zeros([40])
idx = ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9', '-', ' ',
'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm',
'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z']
num_reg = re.compile("[a-z0-9- ]")
if (type(sent) not in [type('str'), type([])]):
raise Exception("input must be str")
if (type(sent) == type([])):
sent = sent[0]
for char in sent:
vector_a = np.copy(embeddings)
vector_b = np.copy(embeddings)
vector_c = np.copy(embeddings)
vector_d = np.copy(embeddings)
if (num_reg.match(char) == None and hangul.is_hangul(char)):
anl = hangul.separate(char)
vector_a[anl[0] if anl[0] > 0 else 0] = 1
vector_b[anl[1] if anl[1] > 0 else 0] = 1
vector_c[anl[2] if anl[2] > 0 else 0] = 1
elif (num_reg.match(char)):
vector_d[idx.index(char)] = 1
else :
vector_d[39] = 1
return_vector.append(np.append(vector_a, [vector_b, vector_c, vector_d]))
return np.array(return_vector)
except Exception as e:
print("error on get_onehot_vector : {0}".format(e))
def test_is_hangul():
assert hangul.is_hangul(u'한') == True
assert hangul.is_hangul('A') == False
assert hangul.is_hangul('1') == False
assert hangul.is_hangul(None) == False
def test_is_hangul():
assert hangul.is_hangul(u"한")
assert not hangul.is_hangul("A")
assert not hangul.is_hangul("1")
assert not hangul.is_hangul(None)
# 첫번째 글자는 무조건 한자
# 한글 뜻, 음, 획수, 영어뜻 포함
#d={'hanja':l[0],'meanings':[],'sounds':[], 'eng_meanings':[], 'strokes':int(l[i+1:-1])}
# 한글 뜻, 음, 획수 포함
#d={'hanja':l[0],'meanings':[],'sounds':[], 'strokes':int(l[i+1:-1])}
# 한글 뜻, 음만 포함
d = {'hanja': l[0], 'meanings': [], 'sounds': []}
l = l[2:i].replace(';', ',').split(',') # 영어 뜻을 제대로 분리하려면 수정 필요
for i in range(len(l)):
f = False
for c in l[i]:
if not (hangul.is_hangul(c) or c == ' '):
f = True
break
if f: break
kor = l[:i]
eng = l[i:]
for i in range(len(kor)):
kor[i] = kor[i].strip()
for i in range(len(eng)):
eng[i] = eng[i].strip()
for s in kor:
tmp = s.split()
try:
if tmp[0] == '': raise Exception()
d['meanings'].append(' '.join(tmp[:-1]))
for idx, i in enumerate(all_name):
idx_dict[i] = int(idx)
matrix = make_matrix(raw2)
matrix = pd.DataFrame(matrix)
matrix.columns = all_name
matrix.index = all_name
## 데이터 클렌징.
## 오타로 인한 추가 구분 테이터를 통합.
for i in matrix.columns:
tmp = 0
for j in i:
tmp += hangul.is_hangul(j)
if tmp != 0:
print(i,all_dict[i])
matrix = data_cleansing('심상정',matrix)
matrix = data_cleansing('민병두',matrix)
## 의원직이 박탈된 사람은 제거.
with open(r'C:\Github\Project_Social_Bigdata_Assignment\Data\remove_list.pickle', 'rb') as f:
remove_list = pickle.load(f)
for i in remove_list:
try:
matrix = data_cleansing(i,matrix, delete = True)
except:
print(i)
def test_is_hangul():
assert hangul.is_hangul(u'한') == True
assert hangul.is_hangul('A') == False
assert hangul.is_hangul('1') == False
assert hangul.is_hangul(None) == False
