This is a bot that asks the user for a word (or a sentence), and then replies with wether that word/sentence sounds more English or Swedish. The bot gives sarcastical answers, and the level of sarcasm depends on how certain the bot is of its answer.
The project was created with TensorFlow using a FFN neural network.
The network has 1 layer consisting of 4096 nodes, and uses the optimizer "RMSprop" and the loss-function "sparse_categorical_crossentropy". It was trained on a data set consisting of 80 000 words for 10 epochs. It achieved 95% accuracy on a different dataset used for testing.
In the future, I will also implement a LSTM neural network. I will also add support for more languages than just Swedish and English.