CafeteriaSA corpus

CafeteriaSA corpus provides a manually annotated corpus of 500 scientific abstracts with 6,407 Hansard taxonomy, 4,299 FoodOn, and 3,623 SNOMED-CT food-entity annotations to support NLP extraction and semantic mapping of food-related information.

Key Features:

• Corpus size: 500 scientific abstracts annotated for food entities.
• Annotation counts: 6,407 annotations mapped to the Hansard taxonomy, 4,299 to FoodOn, and 3,623 to SNOMED-CT.
• Vocabularies: Annotations explicitly reference three biomedical vocabularies: Hansard taxonomy, FoodOn, and SNOMED-CT.
• Entity focus: Annotations target food-related entities extracted from scientific literature.
• Multi-vocabulary mapping: Food-entity annotations are aligned across multiple semantic frameworks to enable cross-ontology analyses.
• Intended analytical use: Prepared to facilitate development and evaluation of natural language processing methods for food information extraction.
• Relevant domains: Coverage intended for applications in health informatics, nutrition science, and environmental health research.

Scientific Applications:

• NLP model development: Support training and evaluation of NLP models for entity recognition and semantic linking of food items.
• Health informatics and nutrition research: Enable integration of food-entity data into studies in nutrition science and health informatics.
• Environmental and public health analysis: Provide annotated data for analyses of food-related factors in environmental and public health contexts.
• Predictive modeling: Furnish semantically enriched inputs for predictive modeling that require accurate food entity representations.

Methodology:

Manual curation and annotation of food entities in 500 scientific abstracts with mappings to Hansard taxonomy, FoodOn, and SNOMED-CT.