Hey Yuuto, thanks for your reply! Indeed and I actually think that Watson wouldn't be able to measure the quality of a project. Hence we started by actually measuring the hype by analysing news and forums (from pre ICO periods) of top performing ICOs and we compare it with bad ICOs. With a first dataset and very first criteria analysed in 24h, we reach a 54% accuracy of differentiating good from the bad on 40 ICOs. Still a lot of work to do!!