Scientific data seem like double Dutch? You need a food dictionary!

Scientific knowledge is increasing rapidly and scientific data is being generated at an equally brisk pace. It’s a wonder the scientific community can keep up with all this new information, let alone fully access, interpret and apply the data.

With this in mind, a group of Dutch companies and universities is developing a text analysis tool that will systematically analyse scientific literature for food-related knowledge. The Development of Food Ontologies project involves the development of a large food dictionary (ontology) that will be used to search a large number of texts such as books, patents and journal articles for all things food related.

An initiative of Radboud University Nijmegen, Vrije Universiteit Amsterdam, Wageningen UR Food & Biobased Research and NIZO food research, the developers anticipate that combining these search results will help the food and beverage industry in its search for new product concepts.

“Instead of having to go through enormous amounts of information by hand, computer software now enables us to screen over 20 million documents in a very short period of time,” said NIZO project leader Wynand Alkema. “Furthermore, computer models will come up with combinations that are not easily thought of by scientists, resulting in refreshing and unexpected outcomes.

“For example, this technology can bring together the knowledge for a particular food product or ingredient from the seemingly unrelated fields of agricultural, medical and social sciences to provide leads for new applications or new markets that had not been considered before.”

Although a number of generic knowledge discovery and knowledge management (KDKM) and text mining (TM) tools already exist, their application in food research is limited. A major reason is the absence of structured vocabularies that are of interest to specific food research applications.

The Development of Food Ontologies project will develop structured vocabularies covering the food domain, which will be incorporated in existing KDKM and TM tools to link potentially related research findings. These relations can be used to generate hypotheses addressing important areas of food research.

Each member of the consortium brings valuable experience and skill to the project. VU Amsterdam and Radboud University Nijmegen are specialists in the development of ontologies and data mining software. Wageningen UR and NIZO are able to put their expert food knowledge to use by applying the ontologies and interpreting the results.

“The collaboration between research companies and universities ensures a rapid implementation for use by industry,” said Jan Top, project leader at Wageningen UR and professor at VU Amsterdam.