Export categorized text corpus

[turicas]

NLTK supports categorized corpora: a method categories is created in the corpus object and all other methods like sents, paras etc. accept a categories option to filter by category.
We should export this corpus as a categorized corpus where categories are Wikipedia Portals. This task includes:

• Retrieving portal information from Portuguese Wikipedia (which page belongs to which portal?);
• Create a custom corpus reader (or modify an existing one) to export.