Note ReciFine is a large-scale annotated recipe dataset containing over 2.2 million recipes and more than 97 million fine-grained entity mentions.

Note ReciFineGold is a manually annotated gold-standard dataset of 500 real-world recipes with fine-grained entity annotations for recipe understanding. Used for training, evaluation, and benchmarking traditional and knowledge-augmented recipe NER models.

Note ReciFineGen is a human-annotated evaluation dataset for automatically generated recipes. Designed to support benchmarking of controllable recipe generation using entity-aware prompts.

nuhuibrahim/recifinegold-recipebert-ka-question

Token Classification • 0.1B • Updated 12 days ago • 211

Note BERT-base-uncased knowledge-augmented and entity-specific model trained on ReciFineGold. Uses natural-language questions as the knowledge prefix to extract a single recipe entity type at a time.

nuhuibrahim/recifinegold-recipebert-ka-entity-type

Token Classification • 0.1B • Updated 12 days ago • 13 • 1

Note BERT-base-uncased knowledge-augmented and entity-specific model trained on ReciFineGold. Uses the entity type name as the knowledge prefix to guide entity extraction.

nuhuibrahim/recifinegold-recipebert-ka-definition

Token Classification • 0.1B • Updated 12 days ago • 11

Note BERT-base-uncased knowledge-augmented and entity-specific model trained on ReciFineGold. Uses formal entity definitions as the knowledge prefix to support precise entity extraction.

nuhuibrahim/recifinegold-recipebert-ka-all

Token Classification • 0.1B • Updated 12 days ago • 15

Note BERT-base-uncased knowledge-augmented and entity-specific model trained on ReciFineGold. Uses the combination of the 5 knowledge types as the knowledge prefix to guide entity extraction.

nuhuibrahim/recifinegold-recipebert-ka-example

Token Classification • 0.1B • Updated 12 days ago • 8

Note BERT-base-uncased knowledge-augmented and entity-specific model trained on ReciFineGold. Uses the entity examples as the knowledge prefix to guide entity extraction.

nuhuibrahim/recifinegold-recipebert-trad

Token Classification • 0.1B • Updated 12 days ago • 103

Note BERT-base-uncased model trained on ReciFineGold using the traditional BIO tagging formulation. Predicts all recipe entity types jointly from raw recipe text.

nuhuibrahim/recifinegold-reciperoberta-ka-question

Token Classification • 0.1B • Updated 12 days ago • 10

Note RoBERTa-base knowledge-augmented and entity-specific model trained on ReciFineGold. Uses natural-language questions as the knowledge prefix to extract a single recipe entity type at a time.

nuhuibrahim/recifinegold-reciperoberta-ka-entity-type

Token Classification • 0.1B • Updated 12 days ago • 3

Note RoBERTa-base knowledge-augmented and entity-specific model trained on ReciFineGold. Uses the entity-type name as the knowledge prefix to extract a single recipe entity type at a time.

nuhuibrahim/recifinegold-reciperoberta-ka-definition

Token Classification • 0.1B • Updated 12 days ago • 9

Note RoBERTa-base knowledge-augmented and entity-specific model trained on ReciFineGold. Uses the entity-type definition as the knowledge prefix to extract a single recipe entity type at a time.

nuhuibrahim/recifinegold-reciperoberta-ka-example

Token Classification • 0.1B • Updated 12 days ago • 6

Note RoBERTa-base knowledge-augmented and entity-specific model trained on ReciFineGold. Uses the entity examples as the knowledge prefix to extract a single recipe entity type at a time.

nuhuibrahim/recifinegold-reciperoberta-ka-all

Token Classification • 0.1B • Updated 12 days ago • 11

Note RoBERTa-base knowledge-augmented and entity-specific model trained on ReciFineGold. Uses the combination of the 5 knowledge types as the knowledge prefix to extract a single recipe entity type at a time.

nuhuibrahim/recifinegold-reciperoberta-trad

Token Classification • 0.1B • Updated 12 days ago • 13

Note RoBERTa-base model trained on ReciFineGold using the traditional BIO tagging formulation. Predicts all recipe entity types jointly from raw recipe text.