External solution using using Graph Neural Networks (GNNs)

One external possible solution (#explainer) using Graph Neural Networks, however this is just one possible way and not a classical solution.

Discerning Odors Using Machine Learning (Fhel Dimaano)


Dear @Tobi, several groups have tested methods to model this olfactive description with small to middle datasets, Google is one of them.
You can also look at the DREAM challenge for over methods tested in the past. But current reported accuracies are still very limited (around 40-45%)
We carefully build this dataset to give the change to crack this challenging question.

Let’s hope that someone will have a better idea.

Best regards,



Thank you for the explanation!

there is one paper that further uses NLP to analyse the distance of the tag in the ground truth … e.g. what is the similarity between scent “rose” and “flower”, etc … sometimes such description are subjective. so the paper see if we can better results by processing the label, e.g. clustering/hierarchy, etc

@hengck23: we do have some broad ontology of how all the labels are related to each other, and we will be releasing the same data in the subsequent round. But until then, it is indeed a clever idea to try and build an ontology of the odor words just by using pre existing language models. Would love to see how they compare with the ontology we will finally release.

If you come up with some exploratory notebooks in that direction, we will strongly encourage you to share some colab links for others to jump on exploring the idea !

Cheers !