classifying reviews

I used tensor in python to create a model classifying if a review was positive or negative. I started with loaded in previously created reviews, but then I created a text file and added in my own reviews. I went on instagram to find some reviews of college board's ap tests, and then I added them into a text file. The closer the first number in the array was to 0, that meant that the review was positive, while the closet it was to 1 meant that the review was negative. The 2nd number in the array was the confidence of the model. The closer the number was to one, the higher the confidence of the model.

Code:

from tensorflow import keras

import numpy as np

data = keras.datasets.imdb

(train_data, train_labels), (test_data, test_labels) = data.load_data(num_words=10000) #only take most frequent words

word_index = data.get_word_index()

word_index = {k:(v+3)for k, v in word_index.items()} #keys and values, three special charcters

word_index["<PAD>"] = 0 #padding

word_index["<START>"] = 1 #start

word_index["<UNK>"] = 2 #unknown

word_index["<UNUSED>"] = 3 #unused

reverse_word_index = dict([(value, key) for (key, value) in word_index.items()]) #swap all values and keys

#trim the data,

train_data = keras.preprocessing.sequence.pad_sequences(train_data, value=word_index["<PAD>"],padding="post", maxlen=250)

test_data = keras.preprocessing.sequence.pad_sequences(test_data, value=word_index["<PAD>"],padding="post", maxlen=250)

def decode_review(text):

return " ".join([reverse_word_index.get(i, "?") for i in text]) #return readable words

# model

# embedding layer to group words that are similar and different

model = keras.Sequential()

model.add(keras.layers.Embedding(88000, 16)) #creating word vectors

model.add(keras.layers.GlobalAveragePooling1D()) #gets all the word vectors

model.add(keras.layers.Dense(16, activation="relu"))

model.add(keras.layers.Dense(1, activation="sigmoid")) #give probabilty

model.summary()

model.compile(optimizer="adam", loss="binary_crossentropy", metrics=["accuracy"])

#split training data

x_val = train_data[:10000]

x_train = train_data[10000:]

y_val = train_labels[:10000]

y_train = train_labels[10000:]

fitModel = model.fit(x_train, y_train, epochs=40, batch_size=512, validation_data=(x_val, y_val), verbose=1) #how many movie reviews

results = model.evaluate(test_data, test_labels)

print(results)

model.save("model.h5")

def review_encode(s):

encoded = [1] #starting point

for word in s:

if word.lower() in word_index:

encoded.append(word_index[word.lower()])

else:

encoded.append(2)

return encoded

model = keras.models.load_model("model.h5")

with open("test.txt", encoding="utf-8") as f:

for line in f.readlines():

nline = line.replace(",","").replace(".","").replace("(", "").replace("(", "").replace(":", "").replace("\"", "").strip().split(" ")

encode = review_encode(nline)

encode = keras.preprocessing.sequence.pad_sequences([encode], value=word_index["<PAD>"], padding="post", maxlen=250)

predict = model.predict(encode)

print(line)

print(encode)

print(predict[0])

Model output: