FoodIntent ML

Experimental TF-IDF intent classifier. Every prediction requires human review.

Evaluation

Selected model: multinomial_nb
Curated test macro F1: 1.0000
SID test macro F1: 0.9580

Data and limitations

Both the SID source and the project-authored delivery supplement are synthetic. These metrics do not estimate performance on real customers. Confidence values are not calibrated. Do not automate customer-impacting decisions.

Dataset: https://huggingface.co/datasets/luigicfilho/sid

Downloads last month: -