The dialogues are extracted from the online books of the Gutenberg Project, which can even produce multilingual data. I present a detailed analysis of data and errors, as well as some results. Then I would like to brain-storm about how to improve data quality and make effective use of multilingualism.