A Python Tensorflow/Keras LSTM neural network generates new adjectives that sound generally cromulent as real words. The words it creates are saved and the user is prompted to provide two inputs for each word: a definition and an example of its use. The script then puts this information together into a full dictionary entry for each word that the user defined, saved into another text file.
The loop of word definition/example inputs can currently be broken early by entering the word "exit" when prompted for a definition. A word can be skipped by entering "skip" as the definition. The main word generating/training function will run for a long time (maybe an hour) if not stopped prematurely by stopping the cell. But, since the model has already been trained, it should create good quality outputs straight away.
Model was trained on a collection of real adjectives from this dataset. Many thanks to the creator of the dataset