Additional and better data.

The second fieldwork is now ended. It is the same setting, but the ad-hoc script has been improved. Hopefully we did find solutions for some of the (very) recurrent problems.

Four intensive days of recordings allowed us to gather 14 additional hours of data. We made some bets inside the team : around 350 more interactions are expected, these will be added to the 538 previous ones. All the recordings will help us to identify particular interactional problems and to produce an annotated corpus of data for Machine Learning.