Skip to content

prashant-kikani/text-generator-by-LSTM

Repository files navigation

text-generator-by-LSTM

It predicts next word and generate the whole play in shakespeare style given the seed sequence.
Simple LSTM of 128 units is used with one Dense layer.
You can cut the dataset to half if your laptop or PC run out of memory while creating the train data for neural network.
Minimimum 20 to 25 epoch is required to generate a human like play. Which requires a great amount of time and computantional power (many GPUs)

You can feed any dataset to it. It will learn from it. And will generate further.

Alternatives to do the same task

We can use Markov chains to do the same. But it will only rely on the previous word, not on the whole previous line. And that will make output artificial rather than natural.
LSTM model also considers previous 15 words to predict next word. That's why it's a better choice.

About

Generates shakespeare style play by LSTM (RNN) network.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published