def pipeline() -> None:
about = input()
ing = input()
category = input()
points = pickle.load(open('points.pkl', 'rb'), encoding='utf-8')
mdl = models.load_model('model.h5')
about = np.array(np.array(text.hashing_trick(preprocess(about), 91)))
ing = np.array(np.array(text.hashing_trick(preprocess(ing), 56)))
category = np.array(np.array(to_categorical(points[-1][category], 19)))
mdl.predict([ing, about, category])
def hashing_method(self):
# get vocab size
motiv = self.flatten(self.motiv_quotes)
demotiv = self.flatten(self.demotiv_quotes)
self.vocab = set(text_to_word_sequence(motiv + " " + demotiv))
self.vocab_size = len(self.vocab)
# perform hash encoding
self.quotes = self.motiv_quotes + self.demotiv_quotes
before = time.time()
for quote in self.quotes:
self.encoded_quotes.append(
hashing_trick(quote,
round(self.vocab_size * 1.5),
hash_function='md5'))
after = time.time()
diff = (after - before) * 1000
print("hashing trick time: " + str(diff) + " ms")
# PADDED HASH DATA FOR TRAINING
self.padded_encoded_quotes = pad_sequences(self.encoded_quotes,
maxlen=280)
#print(self.encoded_quotes)
#print("----------------------------------------")
print(self.padded_encoded_quotes)
def new() -> None:
with open('dataset.csv', newline='') as csvfile:
spamreader = csv.DictReader(csvfile)
name = [row['name'] for row in spamreader]
with open('dataset.csv', newline='') as csvfile:
spamreader = csv.DictReader(csvfile)
ing = [row['ing'] for row in spamreader]
with open('dataset.csv', newline='') as csvfile:
spamreader = csv.DictReader(csvfile)
about = [row['about'] for row in spamreader]
with open('dataset.csv', newline='') as csvfile:
spamreader = csv.DictReader(csvfile)
category = [row['category'] for row in spamreader]
with open('dataset.csv', newline='') as csvfile:
spamreader = csv.DictReader(csvfile)
score = [row['score'] for row in spamreader]
arr = np.array([name, category, about, ing, score])
about = np.array(
[np.array(text.hashing_trick(arr[2, i], 91)) for i in range(357)])
ing = np.array(
[np.array(text.hashing_trick(arr[3, i], 56)) for i in range(357)])
ing = sequence.pad_sequences(ing, maxlen=56, padding='post')
about = sequence.pad_sequences(about, maxlen=91, padding='post')
hist = {}
category = set(category)
for i in enumerate(category):
hist[i[1]] = i[0]
category = np.array(
[np.array(to_categorical(hist[arr[1, i]], 19)) for i in range(357)])
score = np.array([float(x) for x in score])
pickle.dump((about, ing, category, hist), open('points.pkl', 'wb'))
mdl = new_model(about, ing, category, np.array(score))
mdl = models.load_model('model.h5')
from tensorflow.keras.layers import Input, Embedding, Dense
from tensorflow.keras.models import Model
from tensorflow.keras.optimizers import Adam
import numpy as np
samples = ['너 오늘 이뻐 보인다',
'나는 오늘 기분이 더러워',
'끝내주는데, 좋은 일이 있나봐',
'나 좋은 일이 생겼어',
'아 오늘 진짜 짜증나',
'환상적인데, 정말 좋은거 같아']
labels = [[1], [0], [1], [1], [0], [1]]
# hash 테이블로 문서를 수치화한다.
VOCAB_SIZE = 10 # vocabulary 크기 (hash table)를 10개로 한정한다.
sequences = [hashing_trick(s, VOCAB_SIZE) for s in samples]
sequences = np.array(sequences)
labels = np.array(labels)
print(sequences)
# Embedding layer 내부의 출력층 개수임. 단어의 latent feature 개수
EMB_SIZE = 8
# 딥러닝 모델을 빌드한다.
xInput = Input(batch_shape=(None, sequences.shape[1]))
embed_input = Embedding(input_dim=VOCAB_SIZE + 1, output_dim=EMB_SIZE)(xInput)
embed_input1 = tf.reduce_mean(embed_input, axis=-1)
hidden_layer = Dense(128, activation=tf.nn.relu)(embed_input1)
output = Dense(1, activation='sigmoid')(hidden_layer)
model = Model(xInput, output)