Last Tuesday, I finished my PhD and became a dr. in Computer Science Engineering at @ugent_fea. For my public defense, I created a children's storybook to explain: #KnowledgeGraph Creation and Embedding to Realize Explainable #HybridAI Applications 🧵👇

🚨Thrilled to announce that we are hosting the #ISWC2024 Anomaly Detection for Dynamic Knowledge Graphs Challenge! https://t.co/sqKZIggRUl Use ML/AI/Reasoning/... to detect anomalies in evolving kgs. @pimoens @pernischova @em_alam @psbonte @svhoecke @fongenae @iswc_conf

Our software package, pyRDF2Vec, was recently presented by @bsteenwin at @eswc_conf 2023. https://t.co/g4edG7cDiq

We have added a reference to the new RDF2Vec book ( https://t.co/VQ4Bkq5kil) in the README of pyRDF2Vec today. The examples in there are very useful for new users. @heikopaulheim @pRistoski @JanPortisch @bsteenwin @fongenae 👉

Trending in #Automation: https://t.co/KTvMLooFEw 1) Extending Life of Lithium-Ion Battery Systems 2) Data-Driven Predictive Control for Electrical Drive Systems 3) TALK: Tracking Activities by Linking Knowledge 4) PyES: computation of solution & precipitation equilibria

Superproud promotor!! @MathijsNoort wins the best PhD symposium paper at @eswc_conf Congrats 🎉🎉🎉 @IDLabResearch @ResearchUGent @imec_int @ugent_fea @psbonte

.@bsteenwin presenting pyRDF2Vec, a python implementation and extension of RDF2Vec to create a 2D feature matrix of a Knowledge Graph for downstream ML tasks #hybridAI @Gillesvdwiele @IDLabResearch @imec_int

Visit @ma_tzy & @mathiasdbr at our @imec_int ITF booth where we present our real-world evidence framework for disease follow-up powered by explainable & causal AI @IDLabResearch @ELISUGent @ugent_fea

This large ambient living dataset, with all relevant knowledge, is made available both in raw and in #rdf format on

🥳Our TALK paper is now available. In this work, we use #KnowledgeGraph #embeddings to predict human activities in an ambient living context. @imec_int @IDLabResearch @mathiasdbr @ma_tzy @fongenae @svhoecke @fdeturck

Doe mee aan onze studie om inzicht te krijgen in je triggers voor je rookgedrag! #rookstop #mobileapp #behealth #ehealth

Great to see that our pyRDF2Vec repository (python implementation of #rdf2vec) has been used in 23 other open-source projects so far!

Trivia: FLAGS, INK & TALK,... many titles of my papers refer to a particular band. Even in the created storybook: the king's castle was built upon pillars of salt, pillars of sand & birds came flying from the underground. If you could see it then you'd understand @coldplay ❤️

Thank you @fdeturck & @fongenae to guide me through this PhD journey. Many thanks to @svhoecke @jlijffijt @erikmannens @ejimenez_ruiz & @em_alam to evaluate and judge my work. @ResearchUGent @IDLabResearch

In the end of my story, the river and forest was clean again and the 🦫 could still decide what they found interesting, with less arguing & with the ability to adapt the system in the future. They were also able to explain all the decisions made by the ◼️ to their children

𝐓𝐀𝐋𝐊 & 𝗜𝗡𝗞 were used together to predict the human activity in this new DAHCC dataset. Evaluations showed already very promising results. Follow me for future updates about these 𝐓𝐀𝐋𝐊 publications as more is still to come!

To evaluate 5️⃣, a large human activity recognition #dataset was created: https://t.co/8JRl5mBUjL This #OpenSource dataset provide both #sensor data (wearable, smartphone, environmental sensors, etc.) & #semantic #metadata about possible activities, sensor placements & much more

5️⃣ Sensor + metadata Not only the data can change over time, knowledge also evolves. Therefore we created an approach 𝐓𝐀𝐋𝐊 to link smaller event graphs to each other and make predictions based on current knowledge and knowledge from the past.

4️⃣ 𝗜𝗡𝗞 as #ML input Hmm, the new KG representation in 3️⃣ looks really like features. Why not use them directly as KG embeddings for #ML tasks! Compared to R-GCN & #RDF2Vec, 𝗜𝗡𝗞 makes accurate classifications + keeps the KG explainable aspects https://t.co/fkiDLHWKCi

3️⃣ Optimize rule mining #Semantic rule mining is not new, but not originally designed for #HybridAI use cases... I transformed the #KnowledgeGraph into a 2D tabular format such that might be easier and faster to mine semantic rules. This new KG representation called 𝗜𝗡𝗞

The FLAGS #HybridAI architecture provides controls to an expert to verify unseen cases & uses #semantic rule mining to extends the functionality of the KG. Contribution 1️⃣ also fits in this architecture. Evaluations within the #PredictiveMaintenance domain were made.