Nalantis: The next-generation language technology company that will help Flying Forward 2020 make UAM legislation and regulation machine-readable, interpretable and executable

By Nathy Ercol

8 July 2021


Flying Forward 2020 (FF2020) is a three-year project funded by the European Union, working on the development of an entire state-of-the-art geospatial Urban Air Mobility (UAM) ecosystem. By deploying and testing this Urban Air Mobility geospatial ecosystem in five living labs, FF2020 will be able to deliver a best-in-class drone infrastructure, autonomous monitoring and last-mile delivery.

Nalantis, a next generation language technology company, is one of the 12 organisation comprising the FF2020 consortium. We sat down with Stephen Lernout, Chief Strategy and Innovation Officer at Nalantis, to find out what the organisation’s specific role is in FF2020 and how their technology will contribute to the success of the project.


About Stephen Lernout 

"I’m Stephen Lernout and I work for Belgium-based language technology company, Nalantis. My role is about strategy and innovation. I continuously look for exciting new opportunities to which the company can add value."

Can you tell us a bit more about Nalantis?

Nalantis is a Belgium-based next generation language technology company. We started out a small decade ago with the idea of bridging the gap between processing language and actually understanding it. In the last few years our team has developed an Artificial Intelligence (AI) engine that understands the meaning of human language as humans do. We have modelled the software to several domains, including the legal space. Using semantic pattern recognition and machine learning – software that can automatically generate context from unstructured language – we constructed a pre-trained language model for a few of our customers, which can be applied to legislation and case law. 

How did Nalantis become involved with FF2020?

We decided to join the FF2020 consortium because we felt we could also create value for the European Union by developing an interoperable semantic regulatory framework for Urban Air Mobility – an all-encompassing semantic (language) model that automatically understands and contextualises the meaning of all relevant Urban Air Mobility legislation, as well as regulation of the European Union and its Member States.

What is Nalantis’ specific role in the project?

Nalantis is helping to build a ‘Digital Law Pipeline’ to which we can apply our language technology software. This will make legislation and regulation – which is written for and by humans – machine-readable, interpretable, and executable. Our goal is to make drones operate themselves with the use of Natural Language Understanding technology.

Currently, there is a huge multilingual Urban Air Mobility legal data stack in the European Union at Member State and local level. Our automatic semantic annotation software deconstructs (legal) language to meaning. It enables machines to read, understand and interpret a big bulk of this data stack. In collaboration with other FF2020 technical partners, we assess the semantic output and create a technical format that can be easily integrated with other software vendors in the consortium.

We know that Nalantis is working closely with another FF2020 technical partner, Maastricht University. How do you add value to each other’s models?   

What really got me excited about this partnership is working closely with Maastricht University. They have merged their data science and law departments exceptionally well. Nalantis’ machine learning and AI expertise, combined with Maastricht University’s domain expertise in the legal space, enable us to improve the model we are creating. 

Nalantis is also tasked with ontology development. Can you describe what this means and why it’s an important part of your technology?   

One of the central elements in Nalantis’ technology stack is the ontology part. It is a semantic network consisting of nodes which represent words or short phrases of natural language and the labelled relationships between them. Semantic pattern matching recognition is used to combine these concepts into meaningful entities. In other words, we extract meaning to find other meaning.

This is very similar to how human beings understand language and communicate with each other. We don’t think in words but in concepts. This is how our software works. It converts language to meaning using a language-independent ontology tailored towards a specific (vertical) market. The language-independent character of the ontology creates a unique interoperable feature because it understands legislation and regulation in different languages. In terms of governance, this means that legislation and regulation won’t have to be adapted to fit our model. On the contrary, FF2020 products and solutions will adapt to the data cities provide.

“The language-independent character of the ontology creates a unique interoperable feature because it understands legislation and regulation in different languages.”

Stephen Lernout, Chief Strategy & Innovation at Nalantis

Ontology is, thus, interoperable. Will it also be scalable and sustainable?   

Nalantis is striving to build a key engine for natural language understanding with a library of highly scalable components. We have incorporated a self-learning process to the ontology through we can track past, new or real-time legislation and regulation. Because that process happens on our end, it unburdens the customer, end-user or developer of this task, making the model sustainable and scalable in deploying it over time.

How will Nalantis contribute to the success of the FF2020 living labs?

To me, automated governance is the key. The living labs possess key expertise and competences within their domain, but not necessarily in understanding legislation and regulation. Nalantis’ software is already highly efficient and pre-trained, so the living labs won’t be required to manually gather large amounts of structured sample data to train the AI language model. This is a huge leap forward in terms of automated governance, which will speed up the adoption of AI in government in general.

Additionally, we apply a no-black-box or open methodology. Black box AI is any artificial intelligence system used, whose inputs and operations are not visible to the user or another interested party. This means that we’ll be able to open the ontology to explain how legislation and regulation will be handled in the project through software. Thus, Nalantis will not only be able to help the living labs pragmatically, but also in terms of explainability.

What do you think is FF2020’s biggest hurdle in realising its vision?

Speaking from our own point of view, we are still confronted with the fact that legislation and regulation is written by and for humans. It makes it very hard for machines to read, understand and execute this format. The way that legislation and regulation is written today, leaves a window open to interpretation for humans and machine. Thus, not all data is quantifiable by machines.

How will Nalantis contribute to overcoming these hurdles?

We will need to take a phased approach in which we will automate that what actually can be automated. Together with our FF2020 partners, we are seeking to install a ‘decade roadmap’ that contains milestones in how to create fully autonomous systems (drones) over time. The roadmap will, in turn, create a set of recommendations to the EU.

One example is how future legislation and regulation can be written so machines will get better at understanding and interpreting language. Right now, we know which data is quantifiable by machines and which data is not. In the first part of the project, we are focusing on the data that is quantifiable. How to improve quantifiability of the data and how to create recommendations for the European Commission will have our focus in the second stage of the project. 

What kind of impact do you wish Nalantis and FF2020 to have on the quality of life of European citizens and industries?

I think that we’re able to contribute to a system that enables automation in a safe and sustainable way. It’s exciting to be part of a project that is able to develop how passenger and cargo drones will operate themselves over time. Creating a back end of trustworthy AI systems enables front end engineers to build trustworthy products in the Urban Air Mobility space, which will help to increase public acceptance of these emerging technologies.

“When it comes down to safety and sustainability, you need AI to make trustworthy decisions. The only way to do that is to open the black box.”

Stephen Lernout, Chief Strategy & Innovation at Nalantis

Earlier in the interview you mentioned that Nalantis uses a no black box or open methodology. Can you also elaborate on how the transparency and openness of your technology will help citizens accept AI in their daily lives?

Recently, the European commission announced their regulatory framework in terms of AI adoption. One of their key messages is that trustworthy AI is a must. Trust is a must. If you look at the inherent characteristics of machine and deep learning, developers and end-users are uncertain of what happens between the input and output process, since there are no explainable algorithms being generated.

By having a semantic component central to the software that we’re delivering in this project, Nalantis can build explainable input and output algorithms, resulting in explainable and trustworthy AI. I think it’s dangerous if we apply deep learning black boxes to regulatory aspects, because when it comes down to safety and sustainability, you need AI to make trustworthy decisions. The only way to do that is to open up the black box. In my opinion, Nalantis’ no black box mantra is going to be key in the Flying Forward project and Urban Air Mobility development. 

What do wish to accomplish for Nalantis through FF2020?

Trust is a must. We want to emphasise that AI should not necessarily be a ‘dark secret’. Black box AI is typically used for deep learning modelling. In other words, the algorithm takes millions of data points as inputs and correlates specific data features to produce an output. That process is largely self-directed and is generally difficult for data scientists, programmers, and users to interpret. This is very relevant for the FF2020 project, given its regulatory and safety aspects. Nalantis, with the help of other technical partners, is aiming to prove during this project that a no black box approach is possible to create explainability and transparency.

“Trust is a must.”
Stephen Lernout, Chief Strategy & Innovation at Nalantis

To finalise the interview, would you like to make a final statement?

Entering the spatial era, one can only image how cyber-physical societies will look like in the next decades. Together with other FF2020 partners, we feel a sense of contribution in trying to facilitate the road ahead within the regulatory domain. We already talked about helping to create a regulatory roadmap that contains several milestones and recommendations for the next few years.

Nalantis wants to help create ‘safe passage’ across these markers by introducing a phased approach, guiding us from a human to man-machine environment and eventually to a machine and algorithm-driven ecosystem. This is very often accompanied by social acceptance. It also requires a ‘whole-of-society’ approach to integrate safety, security and sustainability. Regulating an entirely new industry is a big challenge in an already highly regulated market. It will require a common (interoperable) recipe between Member States. We think Nalantis and FF2020 partners can play an essential role in this.

© FF2020