Next Word Predictor.txt

import tensorflow as tf
from tensorflow.keras.preprocessing.text import Tokenizer
from tensorflow.keras.layers import Embedding,LSTM,Dense
from tensorflow.keras.models import Sequential
from tensorflow.keras.utils import to_categorical
from tensorflow.keras.optimizers import Adam
import pickle
import numpy as np
import os

from google.colab import files
uploaded=files.upload()

files=open('text_data.txt','r',encoding='utf8')
# store file in the list
data_lines=[]
for i in files:
  data_lines.append(i)
# convert the list to string
data=""
for i in data_lines:
  data=' '.join(data_lines)

# replace unnecessary stuff with space
data=data.replace('\n','').replace('\r','').replace('\ufeff','').replace('"','')
data=data.split()
data=' '.join(data)
data[:500]

len(data)
tf.test.gpu_device_name()
tokenizer=Tokenizer()
tokenizer.fit_on_texts([data])

pickle.dump(tokenizer,open('token.pkl','wb'))

sequence_data=tokenizer.texts_to_sequences([data])[0]
sequence_data[:10]
# tokenizer.word_index

len(sequence_data)

sequences=[]

for i in range(2,len(sequence_data)):
  words=sequence_data[i-2:i+1]
  sequences.append(words)
print("The length of sequences are: ",len(sequences))
sequences=np.array(sequences)
sequences[:20]

x=[]
y=[]

for i in sequences:
  x.append(i[0:2])
  y.append(i[2])

X=np.array(x)
Y=np.array(y)

vocab_size=len(tokenizer.word_index)+1
Y=to_categorical(Y,num_classes=vocab_size) # converting to matrix
Y[:8]

model=Sequential()
model.add(Embedding(vocab_size,10,input_length=2))
model.add(LSTM(1000,return_sequences=True))
model.add(LSTM(1000))
model.add(Dense(1000,activation="relu"))
model.add(Dense(vocab_size,activation="softmax"))

model.summary()

from tensorflow.keras.utils import plot_model

plot_model(model,to_file='plot.png',show_layer_names=True)

from tensorflow.keras.callbacks import ModelCheckpoint

checkpoint=ModelCheckpoint("next_words.h5",monitor='loss',verbose=1,save_best_only=True)
model.compile(loss='categorical_crossentropy',optimizer='Adam')

model.fit(X,Y,epochs=80,batch_size=55,callbacks=[checkpoint])


from tensorflow.keras.models import load_model
model=load_model('next_words.h5')
tokenizer=pickle.load(open('token.pkl','rb'))

def predict_next_word(model,tokenizer,text):
  sequence=tokenizer.texts_to_sequences([text])
  sequence=np.array(sequence)
  pred=np.argmax(model.predict(sequence))
  predicted_word=""

  for key,value in tokenizer.word_index.items():
    if value==pred:
      predicted_word=key
      break
  print(predicted_word)
while(True):
  text=input("Enter your line: ")
  if text=="0":
    print("execution completed...")
    break
  else:
    try:
      text=text.split(" ")
      text=text[-3:]
      print(text)
      predict_next_word(model,tokenizer,text)
    except Exception as e:
      print("Error occured: ",e)
      continue