The notebook aimed to perform and demonstrate a binary classification problem - spam detection, on a dataset of Polish emails. To complete the task we developed a LSTM model which is a specific type of recurrent neural network (RNN). We proved the high effectiveness of its application for natural language processing (NLP) tasks and achieved an overall 97.06% accuracy. Before training the neural network we used word2vec technique to convert descriptions into embeddings with a pre-trained model specifically for Polish, developed by the Polish Academy of Science. The evaluation of the neural network was done with standard plots.
