About me

I am Louis van Beurden, a French software engineer (M.eng. from ENSIMAG) specialized in natural language processing (M.Sc. in NLP from Université de Montréal). I am deeply passionate about formal and semantic codes, and I have been working for the past 3 years on the construction of a code to represent meaning: IEML.

Research interest in AI :

Here is a list of the research topics I am following, that I believe are essential to the development of an human-like intelligent AI:

  • Achieving a hybrid system that makes the best use of the two main paradigms: Symbolic (meaning is intrinsically discrete) vs. connectionism (statistical inference yields a continuous representation). The symbolic approach is reliable but does not cover the diversity of languages, where statistics are unable to produce a meaningful model of the world.
  • Clarifying the distinction between designation and reference. That is, between linguistic knowledge (common sense, knowledge of the dictionary) and knowledge of the world (properties, facts, object theories). This distinction is fundamental to define the right apriori to be encoded in the model.
  • Finding a way to interpret and explain the results of statistical machine learning models. Without innovation in this direction, statistical machine learning will not be able to be used in critical systems, such as autonomous cars.
  • Bridging the gap between the AI and the digital humanities. The question of mind, intelligence, meaning and perception is complex, but has been studied by philosophers for thousands of years.
  • Modeling pragmatics as formal language games. Our social world is governed by implicit codes that allow specific acts of speech in specific situations that produce concrete actions. Language is not only a descriptive but also a performative tool (semantics is not a value of truth about the world).

My commitments

Since 2016, I have been collaborating with the philosopher Pierre Lévy on his artificial language IEML. I agree with his vision that innovation in our language is necessary to exploit the opportunities that the Internet still has to offer. IEML is unique in its computable nature, the fact that syntax represents semantic as well as the construction of IEML expressions is made by composition over a small set of primitive meaning symbols. I participate in the conception of the grammar by giving my point of view as a computer scientist. I also design the workflow for writing and editing IEML, by defining how to visualize, write and store IEML expressions in a reusable and standardized way. I also propose and open up areas of research in IT that I consider essential to bring IEML closer to machine learning.

Here are the main tools I have built to make semantic computation with IEML a reality:

  • The ieml library : a python package with an IEML parser and serializer, and the IEML grammar logic.
  • Intlekt : a WYSIWYG writing tool for IEML expressions.
  • The IEML database: a github repository that stores every versions of the IEML vocabulary.

I also wrote a thesis on comparing many automatic translation systems to see if we are able to bypass the human post-editor effort for automatic translation.

Follow me

Montreal / Paris

Get in Touch

Create your website with WordPress.com
Get started
%d bloggers like this: