I am looking a pretrained deep neural network (programmed in TensorFlow or keras if possible) that given a sequence of text characters could give the probability of appearence of every next character (in english language).
Do you know anything like this?
Andrej Karpathy's Char-RNN is a LSTM based RNN for character prediction. He talks about it and RNN text generators in his blog:
http://karpathy.github.io/2015/05/21/rnn-effectiveness/
He also has a Torch implementation of Char-RNN in his GitHub repository:
https://github.com/karpathy/char-rnn
There are some implementations in TensorFlow or Keras, in the next repository you have one with the pretrained weights:
https://github.com/crazydonkey200/tensorflow-char-rnn
Char-RNN is designed to generate sequences char by char, but you can feed the model with a sequence of characters and predict the next one.
It is not what you asked, but I incidentaly found this ealier today:
https://github.com/aaron-xichen/pytorch-playground
- Badges
- Science topic