Russian sentiment analysis evaluation datasets from past competitions are posted in this repository.
| Dataset | Train Volume | Test Volume |
|---|---|---|
| ROMIP-2012 | 4260 | 5500 |
| SentiRuEval-2015 Telecom | 5000 | 5322 |
| SentiRuEval-2015 Banks | 5000 | 5296 |
| SentiRuEval-2016 Telecom | 8643 | 2247 |
| SentiRuEval-2016 Banks | 9392 | 3313 |
Opinions in direct or indirect speeches extracted from news articles.
Twitter datasets annotated for the task of reputation monitoring. The goal of the competition was to find sentiment-oriented opinions or positive and negative facts about two types of organizations: banks and telecom companies.
Similar to SentiRuEval-2015, but the training datasets in both domains were constructed by uniting of the training and test data of the SentiRuEval-2015 and, therefore they were much larger in size.