Building a Corpus for Personality-dependent Natural Language Understanding and Generation