Research

AI-based language teaching

The Revita project aims at developing state-of-the-art tools for language teaching, to assist learners and teachers:

  • for learners:
    • personalized learning — exercises are tailored exactly for each learner, based on prior performance,
    • exercises based on arbitrary, authentic texts, chosen by the learner (or by the teacher),
    • immediate assistance with any topic or any problem,
    • continual assessment, to make sure that the exercises fit optimally to the learner’s current level of skills
  • for teachers:
    • manage heterogeneous groups of learners,
    • provide learning content and materials for the students optimally suited to their level,
    • continual assessment of each learner’s performance — individual, and as a group.

AI research challenges

We address these objectives through research in a large number of areas in language technology, including:

  • fundamental problems in language analysis, on all levels — morphology, syntax, semantics, etc.,
  • student modeling and psychometrics — Item-Response Theory (IRT), Knowledge Space Theory, etc.,
  • language modeling,
  • question answering and question generation,
  • grammatical error detection and error correction,
  • translation,
  • development and use of linguistic resources,
  • software engineering — user experience, gamification.

Rapidly advancing deep-learning technology opens new possibilities for more effective language teaching and learning. In addition to learning “major” languages, Revita works on revitalization of smaller endangered languages. Initial focus was on endangered Finno-Ugric languages, and Turkic languages. In the context of endangered languages we also work on low-resource language technology.