User Tools

Site Tools


Working Group 3: Multilingual and cross-lingual language technology


Unified modelling helps solve NLP tasks with higher accuracy and better awareness of diversity. Therefore, this WG will be dedicated to NLP coordinating the development of tools leveraging universality and promoting diversity:

  1. Multilingual and cross-lingual syntactic parsers which:
    • pay attention to hard and underrepresented phenomena (unbounded dependencies, MWEs,…),
    • leverage transfer of annotations or models in order to cope with data scarceness;
  2. Prototypes of multilingual and cross-lingual semantic parsers which:
    • derive bi-lexical semantic dependencies from syntactic trees,
    • resolve idiosyncrasies in the syntax-semantics interface;
  3. Multilingual MWE discovery tools which:
    • exploit large non-annotated data to compensate the sparseness of MWEs in annotated corpora,
    • are coupled both with lexicons and MWE identifiers;
  4. Multilingual MWE identifiers which:
    • are coupled with MWE discovery and lexica to better handle unseen data,
    • pay attention to underrepresented phenomena, e.g., discontinuity/variability of MWEs;
  5. Prototypes of tools for automatic identification of idiosyncratic constructions.

The tools themselves will be funded at the national level. WG3 will bring the federating effect to these activities, notably by organizing multilingual evaluation campaigns on parsing and MWE identification. Diversity-based evaluation measures from WG4 will be promoted. The outcomes should validate the computational tractability of the terminologies unified in WG1.


Translations of this page:
  • en
wg3/wg3.txt · Last modified: 2023/03/21 16:43 by gulsen.eryigit