Unilever challenge

12 09 2019

Door Josien Kapma

FarmHack Kennismobilisator. Voormalig melkveehouder en specialist in online interactieve leertrajecten.

Unilever challenge

The WUR Life Sciences hackathon on the 25 & 26th of October is a 32-hour long event to ignite data- and tech-driven collaborations across boundaries. Students, tech entrepreneurs, researchers, and domain experts work jointly on targeted challenges, side by side with representatives of leading Dutch agri-food businesses. In this blogpost Unilever presents the challenge they invite you to.


Taking stock of literature for future food flavours



Jan Koek: ‘we need better search for literature review’

"It was a somewhat obscure publication that I happened to run into, but it meant a breakthrough." Unilever’s challenge “Taking stock of literature for future food flavours” relates to a personal experience of food chemist Jan Koek. Jan Koek acts as the spokesperson for the Unilever challenge. 


In 2017, by mere coincidence, Koek came across a 1998 scientific publication that offered essential clues to a major project he worked on. The extensive literature reviews by the project had completely missed this article. Koek: "What else did we miss? There is an almost exponential growth of scientific publications, even in our in-company experimental and report databases. Scifinder and Google are not good enough, we need more specific search engines."

There are not only too many complete misses (false negatives) as well as too many false positives. The idea of this challenge is to train AI to search better.

Koek, who has over 25 years of Unilever experience, will mentor and guide the team. His experience as a secondary school chemistry teacher might come in handy. He can evaluate the quality of search results as he is a Maillard specialist. There will also be data science expertise. Wendy van Herpen, Digital R&D director at Unilever, says: ”We want to strengthen these kind of collaborations beyond the boundaries of Unilever. This team and this challenge are not a one-off activity, but rather an open invitation to getting to know each other. Feel welcome to join our challenge!”


Taking stock of literature for future food flavours

Within foods research at Unilever, we look for healthy natural flavours for our brands. Unilever believes food can be a force for good and looks for ways in which its food brands can be better for people and better for the planet. The trend is towards vegan, to reduce the impact on the planet. Flavours that people appreciate and that can be produced from non-animal food ingredients often make use of an assembly of specific chemical reactions, collectively named as the Maillard reaction. The Maillard reaction brings the authentic colours, aromas and tastes resulting from familiair home cooking processes (boiling, grilling, roasting etc). The Maillard reaction has been widely researched in thousands of experiments in different conditions all over the world, but despite all the research, the Maillard reaction remains unpredictable due to its complexity.

Although being unpredictable and not amenable for modelling, the huge amount of publications & patents with specific data on the Maillard reaction out in the open domain might give new insights and potential if efficiently processed (read, relevant data extracted and analysed). Will this hold clues to better modelling of the reaction? Can we make the future stock cube by taking stock of large volumes of the current literature on Maillard?


Better learning 

New insights from literature can lead to better, more sustainable and natural processes for current flavour preparation, while making use of AI to search the literature could guide to less and more critical experiments to get to these processes, i.e. experimentation in a more sustainable way. Learnings from this could easily be applied to other food industrial areas having already loads of publications and patents.  


Access to a well-prepared subset of research literature

For this challenge our information specialists have compiled a subset of relevant literature. No time will be lost organizing access. For setting up the rules for selecting relevant passages, you will work with chemists and food experts as well as data scientists.



To test if automated analysis of existing scientific literature, helps us better predict the outcome of Maillard experiments. 

Can we create a model that predicts if an article or paragraph is relevant when looking for a specific outcome of the Maillard reaction? Can we automatically create an overview of all experimental conditions under which the Maillard reaction was tested and link these conditions to outcomes?


Who are we looking for?

We could use as team members: Aspiring text-mining or search specialists, food technologists and chemists who like adding some flavour to their skills! We have experts with mentoring abilities.


About Unilever

Unilever is one of the world’s leading suppliers of Foods & Refreshment, Beauty & Personal Care and Home Care products with sales in over 190 countries and reaching 2.5 billion consumers a day. Unilever has around 400 brands found in homes all over the world, including Knorr, Dove, Dirt Is Good, Rexona, Hellmann’s, Lipton, Wall’s, Lux, Magnum, Axe, Sunsilk and Surf. Since 2010 Unilever has been taking action through the Unilever Sustainable Living Plan to help more than a billion people improve their health and well-being, halve their environmental footprint and enhance the livelihoods of millions of people as they grow their business. 

Geef een reactie

Het e-mailadres wordt niet gepubliceerd. Vereiste velden zijn gemarkeerd met *

Deze site gebruikt Akismet om spam te verminderen. Bekijk hoe je reactie-gegevens worden verwerkt.

Deel dit bericht


KPI Live 2022

Vraaggestuurde sessies voor provinciemedewerkers rondom Kritische Prestatie Indicatoren voor duurzame landbouw Leren van elkaar in online bijeenkomsten DOELGROEP: Specifiek voor […]

Maritieme Hack

Er zijn belangrijke stappen te zetten op weg naar toekomstbestendige visserij en de maritieme sector staat dan ook bol van […]

Results NIVA hack

The results of the NIVA Hackathon on environmental data for farming on the 16th and 17th of June 2022. Participants from Italy, France, the Netherlands and Ireland met in a Dutch farm and worked for 28h on four hackathon challenges.

NIVA Hack (16&17JUNE)

How do we open up environmental data to farmers to help make them more informed land management decisions and improve their environmental performance? This is the key goal for the 32 hour NIVA hackathon, which will take place on a farm on the 16th and 17th June.