# number of unique words we want to use
max_features = 8192

# max number of words in a comment to use (or: number of columns in incoming embedding vector)
max_len = 128

# dimension of the embedding variable (or: number of rows in output of embedding vector)
embedding_dims = 64

# instantiate CNN model
cnn_2cnv = Sequential()

# add embedding layer 
cnn_2cnv.add(Embedding(input_dim=max_features, input_length=max_len,
                        output_dim=embedding_dims))
 
# set the dropout layer to drop out 50% of the nodes
cnn_2cnv.add(SpatialDropout1D(0.5))

# add convolutional layer that has ...
# ... 100 filters with a kernel size of 4 so that each convolution will consider a window of 4 word embeddings
cnn_2cnv.add(Conv1D(filters=100, kernel_size=4, padding='same', activation='relu'))

# add normalization layer
cnn_2cnv.add(BatchNormalization())

# add convolutional layer that has ...
# ... 100 filters with a kernel size of 4 so that each convolution will consider a window of 4 word embeddings
cnn_2cnv.add(Conv1D(filters=100, kernel_size=4, padding='same', activation='relu'))

# add normalization layer
cnn_2cnv.add(BatchNormalization())

# add pooling layer 
cnn_2cnv.add(GlobalMaxPool1D())

# set the dropout layer to drop out 50% of the nodes
cnn_2cnv.add(Dropout(0.5))

# add dense layer to produce an output dimension of 50 and using relu activation
cnn_2cnv.add(Dense(50, activation='relu'))
# finally add a dense layer
cnn_2cnv.add(Dense(1, activation='sigmoid'))

cnn_2cnv.compile(loss='binary_crossentropy',
                  optimizer=Adam(0.01),
                  metrics=['accuracy'])

cnn_2cnv.summary()

Model: "sequential_1"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
embedding_1 (Embedding)      (None, 128, 64)           524288    
_________________________________________________________________
spatial_dropout1d_1 (Spatial (None, 128, 64)           0         
_________________________________________________________________
conv1d_1 (Conv1D)            (None, 128, 100)          25700     
_________________________________________________________________
batch_normalization_1 (Batch (None, 128, 100)          400       
_________________________________________________________________
conv1d_2 (Conv1D)            (None, 128, 100)          40100     
_________________________________________________________________
batch_normalization_2 (Batch (None, 128, 100)          400       
_________________________________________________________________
global_max_pooling1d_1 (Glob (None, 100)               0         
_________________________________________________________________
dropout_1 (Dropout)          (None, 100)               0         
_________________________________________________________________
dense_1 (Dense)              (None, 50)                5050      
_________________________________________________________________
dense_2 (Dense)              (None, 1)                 51        
=================================================================
Total params: 595,989
Trainable params: 595,589
Non-trainable params: 400

cnn_2cnv_hist = cnn_2cnv.fit(X_train, y_train, batch_size=512, 
                         epochs=5, validation_split=0.1)

Train on 3240000 samples, validate on 360000 samples
Epoch 1/8
3240000/3240000 [==============================] - 3219s 994us/step - loss: 0.2760 - accuracy: 0.8817 - val_loss: 0.2206 - val_accuracy: 0.9118
Epoch 2/8
3240000/3240000 [==============================] - 3235s 998us/step - loss: 0.2331 - accuracy: 0.9061 - val_loss: 0.2106 - val_accuracy: 0.9157
Epoch 3/8
3240000/3240000 [==============================] - 3257s 1ms/step - loss: 0.2251 - accuracy: 0.9099 - val_loss: 0.2124 - val_accuracy: 0.9157
Epoch 4/8
3240000/3240000 [==============================] - 3184s 983us/step - loss: 0.2199 - accuracy: 0.9121 - val_loss: 0.2029 - val_accuracy: 0.9193
Epoch 5/8
3240000/3240000 [==============================] - 3294s 1ms/step - loss: 0.2164 - accuracy: 0.9137 - val_loss: 0.2015 - val_accuracy: 0.9201

cnn_2cnv_loss, cnn_2cnv_auc = cnn_2cnv.evaluate(X_test, y_test, batch_size=32)
print('Test Loss:    ', cnn_2cnv_loss)
print('Test Accuracy:', cnn_2cnv_auc)

def scrape_reviews ():
    # For ignoring SSL certificate errors
    ctx = ssl.create_default_context()
    ctx.check_hostname = False
    ctx.verify_mode = ssl.CERT_NONE
    url=input("Enter Amazon Product Url- ")
    html_page = urllib.request.urlopen(url) #Make a get request to retrieve the page
    soup = BeautifulSoup(html_page, 'html.parser')

    reviews=[]
    ratings=[]

    review_row=soup.findAll('div', attrs={'data-hook': 'review'})
    for row in review_row:
        ratings.append(row.find('span',  attrs={'class':'a-icon-alt'}).text.strip()[0])
        reviews.append(row.find('div', attrs={'data-hook': 'review-collapsed'}).text.strip())
    
    print('There are {} reviews in for this product on this page'.format(len(reviews)) )
    #print(reviews)
    
    return reviews, ratings

def punctuationRemover(p):
    '''
    Input: Takes a string. You may have to use str() to force it. 
    Removes all punctuation by checking every single character.
    Output: Returns a string.
    '''
    punctuations = '''!()-[]{};:'"\,<>./?@#$%^&*_~1234567890''' 
    no_punctuations = ''

    for words in p: # You may not have to loop this high
        for char in words:
            if char in punctuations:
                no_punctuations = no_punctuations + ' '
            if char not in punctuations:
                no_punctuations = no_punctuations + char    
    return(no_punctuations)

def no_stopword (p):
    token= ' '.join([word.lower() for word in p.split() if word.lower not in (stop)])
    return token

def lemmatize_verbs(words):
    """Lemmatize verbs in list of tokenized words"""
    lemmatizer = WordNetLemmatizer()
    lemmas = []
    for word in words:
        lemma = lemmatizer.lemmatize(word, pos='v')
        lemmas.append(lemma)
    return lemmas

def web_comments ():
    """Scraping and predicting multiple customer reviews from a website"""
    review_list, ratings = scrape_reviews() 
    review_list_no_punc=[punctuationRemover(p) for p in review_list]
    review_list_no_stop=[no_stopword(p) for p in review_list_no_punc]
    test=review_list_no_stop
    test_vektor=tokenizer.texts_to_sequences(test)
    test_vector=sequence.pad_sequences(test_vektor, maxlen=128)
    prediction=model.predict([test_vector])
    
    neg_com=[]
    pos_com=[]
    label=[int(round(prediction[0][0])) for prediction[0][0] in prediction]
    labels=[]
    for i in label:
        if i==1:
            labels.append('P')
        else: 
            labels.append('N')
        
    for i in label:
        if i==0:
            neg_com.append(review_list[i])
            
            
    for i in label:
        if i==1:
            pos_com.append(review_list[i])
   
    print ("")
    print ("Predictions")
    print(labels)
    print ("Actual Rates")
    print(ratings)
    print ("")
    
    print("List of negative comments")
    print("============================")
    for i in neg_com:
        print (i)
        print ("")