Am I reading it wrong, or does the article say that 90% of _accounts_ are pro-Ukrainian, 7% pro-Russian? Section 3.0 describes how they got the dataset (all tw

Am I reading it wrong, or does the article say that 90% of _accounts_ are pro-Ukrainian, 7% pro-Russian? Section 3.0 describes how they got the dataset (all tweets in a timeframe with a "#IStandWith[Ukraine|Russia|Zelynskyy|Putin]" hashtag). Section 3.1 describes how they assign national 'lean' to the accounts in their entire dataset; that's the section that has the 90%/7% numbers. Section 3.2 describes how they quantify bot activity in their dataset, using a program called BotoMeter to apparently run regression on accounts in the dataset. Section 4 gives the average botometer scores by national lean. This is the oddest of all, since the botometer scores for both sides are roughly equal. It seems to imply that Ukraine and Russia both had bot armies that were roughly equal in proportion to the total number of accounts favouring their national camp, despite the huge disparity in the size of the national camps (90% vs 7%). Which seems odd. [Edit: to clarify: I may have missed it but the paper doesn't seem to give a clear statement on %bot/country anywhere.] All this article really does is make me curious about botometer accuracy and the usefulness of a 'IStandWith*' filter for doing a study like this.