Knowledge engineering in agrifood

Knowledge engineering in agrifood

From developing new products to making processes sustainable: expert knowledge is crucial to success for agrifood companies. Wageningen Food & Biobased Research takes organizational knowledge and learning and makes it concrete, while revealing hidden, sometimes limiting, assumptions. We pull all this together to create algorithms which enable companies to make scientifically-substantiated claims and create decision-support systems that employ the latest insights.

Knowledge models

Most knowledge is, in fact, not to be found on paper, but in people's minds. Think of the common sense of a nutritionist who tells someone that they should - in principle - adhere to specific dietary guidelines, but that it is okay to eat fries occasionally. Or the fruit trader who can see at a glance whether a batch of apples will survive being transported abroad. This often involves valuable knowledge and experience; combining both and drawing from every source and employee can be incredibly difficult for an organization. Experts do change companies, different disciplines often do not understand each other well or share data and learning; for some it is simply seen as too expensive or time-consuming to consult experts. By linking meaningful data to knowledge models, smart applications can offer continuous support.

Model-based reasoning

Wageningen Food & Biobased Research translates knowledge into algorithms, the ‘thinking’ engines of software. These enable the predictions, on which companies base their decisions ("what if ..."), to be more accurately and better substantiated. We combine a deep understanding of knowledge engineering with broad expertise in agrifood - from fresh food chains and product development to nutrition and health.
We structure systematic models of expert knowledge, providing insight into implicit assumptions, differences in jargon and inconsistencies in patterns of thinking. We also reveal where thinking has become completely stuck, where it appears there is no good solution. Integrating all of the above creates a foundation upon which we create original smart software systems. For example, to support wholistic, integrated product development, or for apps that provide personalized nutritional advice to consumers. This way of making expert knowledge usable by software is complementary to machine learning, where large data sets form the basis for new knowledge models.

Anticipate potential food safety risks. This is now possible with the (in development) risk detection system that we have developed for the Chinese dairy company Yili. We created a model of all the knowledge and experience resident in a broad group of experts from Wageningen University & Research (Wageningen Food & Biobased Research and Rikilt) about food safety in the chains. The company can now better anticipate potential food safety risks and thus prevent expensive product recalls.
Precisely determining the quality of roses enhances informed decisions about storage, transport and sales. A new knowledge model, we developed within GreenCHAINge (2016-2019) - a project with 15 partners from the entire chain – has made this a reality. The system is based on a Bayesian Belief Network.