dataset generation

From a 'root' Wikipedia article, a tree is constructed, where the child nodes are random subset of the articles linked in the parent node article. Used to generate datasets of varying size and similarity to the root article for NLG models.