Last active
February 12, 2019 08:07
-
-
Save prakhar21/8ca372422c67cbd8be7efeddeaa81871 to your computer and use it in GitHub Desktop.
Neural Langauge Model
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
#!/usr/bin/env python | |
""" | |
@uthor: Prakhar Mishra | |
date: Dec, 12 2017 | |
""" | |
# importing packages | |
from numpy import array | |
from keras.preprocessing.text import Tokenizer | |
from numpy import array | |
from keras.utils import to_categorical | |
from keras.models import Sequential | |
from keras.layers import Dense, LSTM, Embedding | |
# source text | |
data = """My name is prakhar mishra . prakhar mishra writes blog on medium .""" | |
# tokenization and encoding into sequences | |
tokenizer = Tokenizer() | |
tokenizer.fit_on_texts([data]) | |
encoded = tokenizer.texts_to_sequences([data])[0] | |
#print encoded | |
# [8, 5, 7, 1, 2, 1, 2, 9, 3, 6, 4] | |
# vocabulary size | |
vocab_size = len(tokenizer.word_index) + 1 | |
# creating i/o sequence pairs | |
sequences = list() | |
for i in range(1, len(encoded)): | |
sequence = encoded[i-1:i+1] | |
sequences.append(sequence) | |
#print sequences | |
# [[8, 5], [5, 7], [7, 1], [1, 2], [2, 1], [1, 2], [2, 9], [9, 3], [3, 6], [6, 4]] | |
# coenverting pairs to input(X) and Output(y) arrays to feed to NN | |
sequences = array(sequences) | |
X, y = sequences[:,0],sequences[:,1] | |
#converting output to one-hot representation | |
y = to_categorical(y, num_classes=vocab_size) | |
# NN define | |
embedding_size = 10 | |
def model(vocab_size): | |
model = Sequential() | |
# input_length = 1 (one word at a time) | |
model.add(Embedding(vocab_size, embedding_size, input_length=1)) | |
model.add(LSTM(50)) | |
model.add(Dense(vocab_size, activation= "softmax")) | |
model.compile(loss="categorical_crossentropy" , optimizer= "adam" , metrics=["accuracy"] ) | |
model.summary() | |
return model | |
model = model(vocab_size) | |
# training starts | |
model.fit(X, y, epochs=500, verbose=2) | |
# testing the model | |
seed = "writes" | |
encoded = tokenizer.texts_to_sequences([seed])[0] | |
encoded = array(encoded) | |
y_pred = model.predict_classes(encoded, verbose=0) | |
for word, index in tokenizer.word_index.items(): | |
if index==y_pred: | |
print word |
Thank you for the explanation and the code. It was helpful.....
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
very nice