a set of texts that human expert coders have already classified (populist/not populist, or a continuous populism score). 2. **Feature extraction** — converting text into numerical features (word frequencies, n-grams, sentiment scores, structural features). 3. **Model training** — fitting a classifie