Description
Recent progress in generating natural language text catches media attention. Are we about to get flooded by autogenerated fake news? Let's learn about approaches to machine-generated text. Get a high level idea of how can you apply basic approaches like N-grams, HMMs as well as advanced ones such as RNNs and VAEs. We'll apply those methods to real-world datasets of Polish articles from Wikipedia.
Recent progress in generating natural language text sparks controversy and catches global media attention. Are we about to get flooded by machine- generated fake news? Are we on the edge of a completely new level of troll farms about to emerge? In this talk I will go over approaches to machine- generated text. You will get a high level idea of how can you apply basic approaches like N-grams, Hidden Markov Model as well as advanced ones such as RNNs and Variational Autoencoders. We will cover the main challenges like methods of evaluation, and potential use cases. We will also have fun applying aforementioned methods into real world datasets of Polish articles from Wikipedia.