25th Belgium NLP Meetup

A new year has begun, and 2025 promises to be an interesting one, to say the least! Will GPT-5 see the light, or is it stuck in training purgatory? Will AI agents take our jobs, or just our coffee orders? Can NLP finally decode what my cat is thinking?

We invite all NLP enthusiasts to join us at the next Belgium NLP Meetup, on Thursday February 20th in the offices of TechWolf in Ghent. You can count on an enjoyable evening of talks, discussions and, let’s face it, probably no definitive answers to the burning questions above.

As always, doors open at 7pm for pizza and drinks. The first talk starts around 7.30pm, and after 9pm there's time for drinks and networking.

From Zero to One with NLP
Mathias Colpaert (Trensition)
In his talk, Mathias will discuss how Trensition automates trend research with NLP. He will introduce Trensition's data processing pipeline, which starts with the identification of high-quality data sources, data ingestion and a Kafka pipeline for (re)processing. Next up are various NLP steps with spaCy, supported by a Lambda architecture and various data analytics methods, such as time series analysis and frequency-based algorithms. Mathias' talk will conclude with a look ahead to future developments, including the integration of large language models (LLMs) and knowledge graphs.

Meet the Lightning-Fast ModernBERT (and its surprise Belgian Edition)
François Remy (Parallia)
ModernBERT is ready to revolutionize your NLP tasks! A group of European researchers revamped good old BERT, with a Belgian twist. In this presentation, we'll explore the technological advancements that have propelled ModernBERT well past its predecessor. You will also get to learn how the four Belgian languages were brought to the ModernBERT era, and how you can leverage this to facilitate the seamless cross-lingual transfer of NER annotations, or any other type of span label.

Taking the N out of NLP: Using NLP for things that are not language
Marijn De Kerpel (TechWolf)
TechWolf tackles the challenge of working with skills by treating it as an NLP problem. Using NLP techniques, they build a skill ontology where all possible skills form the vocabulary. This approach goes beyond traditional language processing, showing how anything can be modeled as "language" to solve complex problems — not just in HR, but across any domain that requires structured understanding of relationships.
While large language models (LLMs) are transforming NLP, ontologies remain critical as the grounding framework for data. LLMs amplify this process, enabling the creation of even more effective ontologies and unlocking new possibilities for structuring and interpreting unstructured information in various fields.