Frequently Asked Questions

Twitter's publicly available data is used via API.

Non-representative social media data is made representative with the Multi-level Regression with Post Stratification model.

For example, while the marriage rates of Twitter users are between 15% and 20% for different cities, thanks to the MRP analysis that helps to make the data representative, the marriage rates at the city level can be estimated very close to the real figures (55%-65%).

Politus Analytics results are published after evaluating the correlation with official data or other subject-based surveys.

Social scientists who are experts in their fields create detailed definition manuals about the concepts. Tweets are flagged by at least two markers according to these manuals. All tweets marked as common or different by the markers are again evaluated within the scope of conceptualizations developed by experts and finalized. By studying these conceptualizations in great detail, artificial intelligence models are trained to distinguish even rare difficult concepts.

17 emotions are reported in this category. Emotional tweets are marked according to the expressions of emotions in Turkey, according to manuals prepared by experts in the field of psychology. For example, while the feeling of gratitude is not included in the content of the general emotion models, this emotion is also included by evaluating the sociological structure of Turkey.

The segmentation criteria are designed as modular so that they can be created completely according to the user's needs. For example; It may be preferable to rank the different feelings of "environmentalist women aged 19-29" about the economy or "conservative men over 40 in Istanbul". Thanks to this ease of segmentation provided by the data platform, the user can evaluate the data he sees on the platform together with his own data. Thus, Politus Analytics allows the user to gain a broader and more interactive insight.