Project ELSNET

European Network in Speech and Natural Language Processing

Activities of IMS in the framework of ELSNET

Term

1995-2000

PI

Ulrich Heid

Short Description

ELSNET is a Europe-based forum dedicated to human language technologies. It operates in an international context, and will consider, across discipline boundaries, all human communication research areas related to language and speech.

Long Description

IMS is the national German managing node and participates actively inthe ELSNET task group on linguistic resources.

The following is a short summary of the relevant activities of IMSStuttgart (``STR'') as a Managing Node of ELSNET, in the period comprised between January and end December 1995.

The activities of STR concern the following areas:

  • acting as Managing Node for ELSNET in Germany and Austria;
  • contribution to the work of the ELSNET Executive Board;
  • preparation of a small morphosyntactically annotated referencecorpus for German (60,000 word forms), in the framework of the ELSNET Reusable Resources Task Group(RRTG).

Executive Board Activities

  • Contribution to the ELSNET Executive Board Meetings held in Stockholm, Leuven and London;
  • contributions to the proposal for a second ELSNET (time frame1996/1998);
  • several contributions to ELSNET's actions in the field ofevaluation:
    • ELSNET brainstorming meeting on evaluation and assessment (17-7-95);
    • contribution (in December 1995 and January 1996) to the preparation of a project proposal on the evaluation and best practice in the field of Spoken Languages Dialogue Systems and Components (DISC), submitted in January 1996 to ESPRIT/LTR.

Reusable Resources Task Group

Much workload has been invested in the preparation of a referencecorpus, in the framework of the ELSNET Reusable Resources Task Group.

As proposed and scientifically organized by STR, in 1995 a joint action between ELSNET's RRTG and the EAGLES / Computational Lexicons Workgroup has taken place. On the basis of proposals for lexical description and annotation, worked out in EAGLES, two small text samples (newspaper text from different fields) one for Italian (carried out by Pisa) and one for German (carried out by STR) have been morpho-syntactically annotated and manually corrected. Thus, the Resources Task Group has produced small ``reference corpora'' which will be disseminated in early 1996, in a joint package, most likely through ELRA. The package will include EAGLES documents and possibly tools from STR and PSA.

In this framework, the following actions have been carried out and documented:

  • selection of a 60,000 words sample from newspaper text material of German; material comes from the ECI MC-1;
  • manual annotation, in two steps, of the text material, with a10% overlap between the individuals carrying out the encoding work;
  • regular monitoring of the manual annotation work, to ensure consistency of the output;
  • preparation of information documents about the corpusannotation, for the ELSNET Executive Board, the ELSNET and EAGLES communities;

ELSNET

Ulrich Heid

Apl. Prof. PD Dr.
To the top of the page