The query hashtags from each tweet were extracted and the total number of pro-Ukrainian (ending in Ukraine or Zelenskyy) and pro-Russian (ending in Russia or Putin) hashtags were counted and used to establish the national lean of a tweet. If the number of pro-Ukranian query hashtags exceeded that of the pro-Russian hashtags, the tweet was labelled as ‘ProUkraine’, and labeled as ‘ProRussia’ conversely. If the counts were balanced, the tweet was labelled ‘Balanced’. Where applicable, the lean of an account was taken to be the most commonly occurring national lean across all tweets from that account. We found that 90.16% of accounts fell into the ‘ProUkraine’ category, while only 6.80% fell into the ‘ProRussia’ category. The balanced category contained 3.04% of accounts, showing that accounts exhibiting mixed behaviour are present in the dataset. We explored other methods for categorising accounts, e.g., labelling accounts as ‘ProUkraine’ or ‘ProRussia’ if they use only those types of hashtag. However, as we were primarily concerned with aggregated activity, we elected to prioritise labelling each account by their ‘usual’ behaviour.
____
Oh wise one - Please let me know how you would interpret this.
