This project adapts the deep learning model develop by mkroutikov (https://github.com/mkroutikov/tf-lstm-char-cnn), which predict words, to detect spelling mistakes in english. It converts words in arrays of numbers by character embedding. These are transfered to a convolutionnal layer and then passed to a LSTM.
The RapportINF8225.pdf in rapport/ describes the results of the experiment (in french).