Natural language understanding: Why we are here

Natural Language Understanding: why we are here


We are terribly excited in announcing the opening of a new Innoradiant unit, namely IrradiantLabs. While progressing with the commercial and technical development, at Innoradiant we realized the importance of being beyond the state-of-the-art research in Natural Language Processing. The tradition of our team is already a tradition of research, with 100+ research papers and dozens of project in cooperation with prestigious European institutions.

With IrradiantLabs we go a little further to create our own research department. The reason for such a creation was rooted in the fact that given our conception of AI and NLP applications, it was necessary to perform our own base research and not only use current state-of-the-art advances. The reasons for this choice are described along the different pages/posts of this site. This is just a bird's-eye view:

  • The application customers demand to us are different each time: we cannot rely on standard NLP task popular in the academic community (even though we do not deny their importance for assessing SOTA).
  • The applications we deliver evolve quite rapidly in time, and sometimes necessitate 24H max delay intervention: we have no way of relying on the construction of new datasets to modify their behaviors.
  • We must always be capable to explain to our customers why a system gave a certain prediction: popular “black box” models are extremely unsuitable for giving such explanations.

As practical and business driven as these requirements might appear, they motivate our research activity, which, in consequence, must take a quite unconventional path, developing technologies that, while taking advantage of recent development, are based on a cognitive comprehension of language and its syntactic and semantic structures: we call it Natural Language Understanding.