pearsonkyle/gpt2-exomachina

A deep language model, GPT-2, is trained on scientific manuscripts from NASA's Astrophysical Data System pertaining to extrasolar planets and the references therein . The model is trained for 500,000 steps with fine tuning . After training, the model is used to generate new text from scratch and from user input . The study uses the abstracts of each article as training data in order to explore correlations in scientific literature from a language perspective . Letting the scrape run for two hours found articles from these publications: The Astrophysics Journal 3365 - Astronomy and Astronomy . and The Astronomical Journal 1355 - ArXiv e-prints 617 - Icarus 498 -,