Text this: C-BiLDA extracting cross-lingual topics from non-parallel texts by distinguishing shared from unshared content.