Observable Subjective Well-Being Inferred from Tweets in Russian

This repo contains code for measuring Observable Subjective Well-Being based on tweets in Russian.

Twitter Data

We used the Twitter Stream Grab¹ as a data source of tweets in Russian. Twitter Stream Grab is a publicly available historical collection of JSON grabbed from the general Twitter “Spritzer” API stream.

Sentiment Classification Model

We fine-tuned RuRoBERTa-Large² on RuSentiTweet³ and achibed new state-of-the-art results of macro $F_1=0.7229$ and weighted $F_1=0.7281$, surpassing the previous SOTA results achieved by the RuBERT model.

Subjective Well-Being Data

We used survey-based VCIOM Happiness Index⁴, which is regularly reported by Russia Public Opinion Research Center.

Footnotes

1. https://archive.org/details/twitterstream ↩

2. https://huggingface.co/sberbank-ai/ruRoberta-large ↩

3. https://peerj.com/articles/cs-1039/ ↩

4. https://wciom.ru/analytical-reviews/analiticheskii-obzor/indeks-schastja ↩