WG 1: Lexicon-Grammar Interface

Internal organization

Leader: This email address is being protected from spambots. You need JavaScript enabled to view it. (DE)
Vice-leaders: Carla Parra Escartín (ES), Gyri Smørdal Losnegaard (NO)

Objectives

  • Better understanding of linguistic properties of MWEs, in particular at the lexical and syntactic level
  • Enhancing the usability of MWE lexicons and valence dictionaries in parsing
  • Paving the way towards interoperability of lexicons and the reduction of their production cost.

Expected outcomes

  • Reports on the contrastive analysis of lexical and syntactic properties of MWEs in different European languages
  • Already existing lexicons and valence dictionaries enhanced and enriched with MWEs, for several European languages
  • Design proposals for cost-saving abstract models of MWEs' properties, such as meta-grammars that could be automatically mapped to different lexicon and grammar formalisms; these models would apply to different languages in question.

Get more information on WG 1.

WG 2: Parsing Techniques for MWEs

Internal organization

Leader: Yannick Parmentier (FR)
Vice-leader: Jakub Waszczuk (FR)

Objectives

  • A better understanding of the potential of different linguistic frameworks with respect to parsing MWEs,
  • Enhancing parsing efficiency,
  • Reducing the cost of grammar production.

Expected Outcomes

  • Recommendations of best practices for MWE representation and treatment in parsing within different theoretical frameworks. The resulting designs should be maximally interoperable.
  • Extensions of existing grammars for several European languages with rules handling MWEs.
  • Design proposals for abstract compact models of MWE-specific grammar rules that might be automatically mapped to different lexicon and grammar formalisms, thus reducing the production cost of particular grammars. These models would apply to different languages in question.

Get more information on WG 2.

WG 3: Statistical, Hybrid and Multilingual Processing of MWEs

Internal organization

Leader: Michael Rosner (MT)
Vice-leader: Matthieu Constant (FR)

Objectives

  • Elaborate ways of combining data-driven and knowledge-based methods to yield various hybrid processing models.
  • Increase the efficiency and accuracy of hybrid parsing methods.
  • Improve our understanding of how these may be applied to the processing of MWEs.
  • Make better use of widely accessible unannotated data in order to improve grammars and models based on annotated data.
  • Investigate relation between hybrid processing methods and multilingual applications.

Expected Outcomes

  • Recommendations of best practices for enhancing data-driven parsing with linguistic resources such as MWE lexicons and valence dictionaries, e.g. by MWE-oriented reranking of state-of-the-art parsers’ results;
  • Recommendations of best practices for enhancing knowledge-based parsing of MWEs with probabilistic scores, in order to avoid spurious syntactic ambiguities while parsing MWEs,
  • Guidelines for extracting probabilistic scores from treebanks and for encoding them in lexicons (cf. WG1 and WG4).

Get more information on WG 3.

WG 4: Annotating MWEs in Treebanks

Leader: Victoria Rosén (NO) 
Vice-leader: Petya Osenova (BG)

Objectives

  • Enhanced MWE-aware methodologies of treebank construction. 
  • Optimal usability of MWEs in parsing.

Expected outcomes

  • Annotation guidelines for representing MWEs in constituency and dependency treebanks.
  • Recommendations on how to use current and future treebanks to automatically extract lexicons and probability scores addressed in other WGs.

 Get more information on WG 4.