X_test, y_test = load_data(in_path + 'yelp-2014-seg-20-20.test.ss')
X_train = [paragraph.split(' <sssss> ') for paragraph in X_train]
X_test = [paragraph.split(' <sssss> ') for paragraph in X_test]
X_train = [[split_words(sent) for sent in paragraph]
for paragraph in X_train]
X_test = [[split_words(sent) for sent in paragraph]
for paragraph in X_test]
W2V_corpus = W2V_corpus_iter(X_train)
w2vModel = train_W2V(W2V_corpus, in_path + 'w2vModel')
word2idx, embedMatrix = build_word2idx_embedMatrix(
w2vModel) # 制作word2idx和embedMatrix
X_train_idx = make_X_train_idx(X_train, word2idx, MAX_SENT_NUM,
MAX_SENT_LEN)
X_test_idx = make_X_train_idx(X_test, word2idx, MAX_SENT_NUM,
MAX_SENT_LEN)
y_train_oneHot = make_y_train_oneHot(y_train, is_cate_dict=True)
y_test_oneHot = make_y_train_oneHot(y_test, is_cate_dict=True)
print(len(X_train_idx), len(X_test_idx), len(y_train_oneHot),
len(y_test_oneHot))
yelp_2014_data = {}
yelp_2014_data['X_train_idx'] = X_train_idx
yelp_2014_data['X_test_idx'] = X_test_idx
yelp_2014_data['y_train_oneHot'] = y_train_oneHot
yelp_2014_data['y_test_oneHot'] = y_test_oneHot
yelp_2014_data['embedMatrix'] = embedMatrix
print('——————————————load data——————————————')
(X_train, y_train), (X_test, y_test) = imdb.load_data()
X_all = (list(X_train) + list(X_test))[0:]
y_all = (list(y_train) + list(y_test))[0:]
print(len(X_all), len(y_all))
imdb_word2idx = imdb.get_word_index()
imdb_idx2word = dict((idx, word) for (word, idx) in imdb_word2idx.items())
X_all = [[imdb_idx2word.get(idx - 3, '?') for idx in sen][1:]
for sen in X_all]
w2vModel = train_W2V(X_all, in_path + 'w2vModel')
word2idx, embedMatrix = build_word2idx_embedMatrix(
w2vModel) # 制作word2idx和embedMatrix
X_all_idx = make_X_train_idx(X_all, word2idx, MAX_SEQ_LEN)
y_all_idx = np.array(y_all) # 一定要注意，X_all和y_all必须是np.array()类型，否则报错
X_tra_idx, X_val_idx, y_tra_idx, y_val_idx = train_test_split(
X_all_idx,
y_all_idx,
test_size=0.2,
random_state=0,
stratify=y_all_idx)
y_tra_oneHot = make_y_train_oneHot(y_tra_idx)
y_val_oneHot = make_y_train_oneHot(y_val_idx)
print('——————————————模型的训练和预测——————————————')
start = time()
model = textCNN_train_test(embedMatrix)
model.train([X_tra_idx, y_tra_oneHot]) #不知道为什么，验证非常非常慢！！但keras非常快，很奇怪！！！
y_pred_idx = model.test([X_val_idx, y_val_oneHot])
X_tra_c, X_tra_t, y_tra = load_data(in_path + 'train.raw')
X_test_c, X_test_t, y_test = load_data(in_path + 'test.raw')
print(len(X_tra_c), len(X_tra_t), len(y_tra))
print(len(X_test_c), len(X_test_t), len(y_test))
if os.path.exists(in_path + 'embedMatrix.pkl'):
embedMatrix = pickle.load(open(in_path + 'embedMatrix.pkl', 'rb'))
word2idx = pickle.load(open(in_path + 'word2idx.pkl', 'rb'))
else:
all = X_tra_c + X_tra_t + X_test_c + X_test_t
word_set = set([w for sent in all for w in sent])
w2vModel = load_W2V(in_path + 'glove.42B.300d.txt', word_set=word_set)
word2idx, embedMatrix = build_word2idx_embedMatrix(w2vModel)
pickle.dump(embedMatrix, open(in_path + 'embedMatrix.pkl', 'wb'))
pickle.dump(word2idx, open(in_path + 'word2idx.pkl', 'wb'))
X_tra_c_idx = make_X_train_idx(X_tra_c, word2idx, MAX_SEQ_LEN)
X_tra_t_idx = make_X_train_idx(X_tra_t, word2idx, MAX_SEQ_LEN)
X_test_c_idx = make_X_train_idx(X_test_c, word2idx, MAX_SEQ_LEN)
X_test_t_idx = make_X_train_idx(X_test_t, word2idx, MAX_SEQ_LEN)
y_tra_oneHot = make_y_train_oneHot(y_tra)
y_test_oneHot = make_y_train_oneHot(y_test)
print('——————————————train model——————————————')
model = IAN_train_test(embedMatrix)
model.train([X_tra_c_idx, X_tra_t_idx, y_tra_oneHot])
y_pred = model.test([X_test_c_idx, X_test_t_idx, y_test_oneHot])
print(y_pred)
