Towards automated recipe genre classification using semi-supervised learning.

Nazmus Sakib1, G M Shahariar2, Md Mohsinul Kabir1

  • 1SSL Lab, Dept. of CSE, Islamic University of Technology, Dhaka, Bangladesh.

Plos One
|January 28, 2025
PubMed
Summary

A new dataset of two million cooking recipes, the Assorted, Archetypal, and Annotated Two Million Extended (3A2M+) dataset, aids in recipe classification and named entity recognition (NER). This dataset achieved 98.6% accuracy in genre classification, highlighting the importance of recipe titles.