CafeteriaFCD Corpus: Food Consumption Data Annotated with Regard to Different Food Semantic Resources.

Area of Science:

• Computational linguistics
• Bioinformatics
• Food science

Background:

• The food and nutrition domain is notably low-resourced despite extensive research.
• Annotated corpora are crucial for domain experts, researchers, and data scientists.

Purpose of the Study:

• To present the annotation process of food consumption data (recipes) using diverse semantic resources.
• To create an enhanced, interoperable food consumption dataset for NLP applications.

Main Methods:

• Annotation of the FoodBase corpus using Hansard taxonomy, FoodOn ontology, SNOMed CT, and FoodEx2.
• Utilized the NCBO annotator for FoodOn and SNOMed CT, and the StandFood method for FoodEx2.
• Developed the Cafeteria Food Consumption Data (CafeteriaFCD) corpus as a new gold standard.

Main Results:

• The CafeteriaFCD corpus contains recipes annotated with semantic tags from four distinct resources.
• Achieved data interoperability across five semantic resources from different domains.
• Established a valuable resource for training information extraction pipelines.

Conclusions:

• The CafeteriaFCD corpus significantly improves food data interoperability and accessibility.
• This resource facilitates the development of NLP tools for food safety knowledge extraction.
• Enables advanced data analysis and knowledge discovery in the food domain.