natural_language_coursework

This worked focused on attempting various kinds of preprocessing to see the impact that they may have on the classification of sentiment for movie reviews. The preprocessing was tested with Multilayer Perceptrons and Support Vector Machines. This included various n-gram levels through BoW and TF-IDF for stemming and lemming. The effect of Latent Semantic Analysis was also assessed, however it seemed that the best performance came through using stemming and uni gram tf-idf. Please read the other read me to run the notebooks.