See the problem, that is incorrect. The term has been in use long before 2022. Originally it was coined during the 2015, and meant to be used for both sides due to how miserable and incompetent the entire thing looked (and was).
That is the problem with the paper in question as well - authors don't seem to be familiar with the topic they're trying to research, thinking it's a single event. The timeframe in the dataset is 2015 to mid-2023, which makes very little sense. The use of Telegram for war reporting and the language have been completely different at various points of this timeframe.
To add insult to injury, they are labeling various channels as pro-R or pro-U based on recent messages, but certain channels literally switched sides. They (and many others as well) wiped their message history multiple times, came back with slightly or completely different narratives, and their actual history can only be found in one of the Telegram-related cache services, if at all, as some of these services are either long dead or the info didn't survive. Some people who have been trying to profit from the war started multiple pro-R and pro-U media, including the Telegram channels, although 2022 quickly made them choose sides.
So much happened in 8 years they tried to shove into an LLM and do a primitive sentiment analysis. Gathering the full picture on this timeframe should have been their main thing, as it's not trivial. Just like with anything on the internet and in real life across 8 years, especially if you don't speak any of the languages. These results are not going to be accurate.
See the problem, that is incorrect. The term has been in use long before 2022. Originally it was coined during the 2015, and meant to be used for both sides due to how miserable and incompetent the entire thing looked (and was).
That is the problem with the paper in question as well - authors don't seem to be familiar with the topic they're trying to research, thinking it's a single event. The timeframe in the dataset is 2015 to mid-2023, which makes very little sense. The use of Telegram for war reporting and the language have been completely different at various points of this timeframe.
To add insult to injury, they are labeling various channels as pro-R or pro-U based on recent messages, but certain channels literally switched sides. They (and many others as well) wiped their message history multiple times, came back with slightly or completely different narratives, and their actual history can only be found in one of the Telegram-related cache services, if at all, as some of these services are either long dead or the info didn't survive. Some people who have been trying to profit from the war started multiple pro-R and pro-U media, including the Telegram channels, although 2022 quickly made them choose sides.
So much happened in 8 years they tried to shove into an LLM and do a primitive sentiment analysis. Gathering the full picture on this timeframe should have been their main thing, as it's not trivial. Just like with anything on the internet and in real life across 8 years, especially if you don't speak any of the languages. These results are not going to be accurate.