We selected a random subsample E of 1.3M dyads of Yahoo Mail users worldwide who have significant interactions with each other, sending at least five replies in each direction in the time span of undisclosed number of months. For privacy and policy reasons, the dataset includes messages belonging exclusively to users who voluntarily opted-in for such studies. Consequently, considering a pair of users who exchanged more than 5 emails, both users need to be opt-in users to be included in the study. These pairs comprise a set N containing 2M unique users exchanging 187M emails overall. We refer to the full sequence of emails flowing within the dyad as a conversation. Next, we gathered all incoming and outgoing emails of users in N over the same time period, a total of 16B emails
