i collected three point seven million twets by these accounts and the cons tafollow them. And then i translated all of these into english using ggl translate. I then manually read two thousand 211 of these posts and classified them as either supporting russia, supporting ukraine or being neutral. And then used a machine learning algridom to classify all of the twets,. And then we tried to use that data as a sort of sheriff of tweed stuffer pro cremlin to figure out which users were natural and which were decidedly not. So looking at the data that we had at that time alone, this is not a question you could really answer.