BHKAI: Turkish news category classifier

BHKAI assigns a category to each incoming story. When it is confident it answers on its own; otherwise the story goes to the AI model.

Model

Method: TF-IDF (300,000 features, min_df 3) with an SGD classifier (alpha 2e-06).

Version: bhkai-20261101, retrained nightly and promoted only if it passes its exam.

Training data: 454,564 stories labelled by the teacher model. Data cut-off: 2026-11-01T03:25:32.

Exam

Last 2 days, 3517 stories.

Threshold	Coverage	Agreement
0.5	0.857	0.901
0.7	0.676	0.952
0.8 (live)	0.561	0.965

Files

• agirlik.npz: weights
• sozluk.json: vocabulary
• bilgi.json: version, exam and comparison details

License

CC BY-NC 4.0 (non-commercial). For commercial use, contact https://bultenhub.com.