Swiss German Language Detection

Welcome to the first shared task on Swiss German Language Detection

Proceedings are available here. We thank all participants for their contributions.

We invite researchers and practitioners to participate in our shared task on Swiss German Language Detection. The task is held as part of the 2020 Germeval evaluation campaign, collocated with the 2020 joint conferences SwissText & KONVENS.

The goal is to build a system that can automatically identify whether a snippet of text is written in Swiss German. A successful system will need to be able to handle (noisy) inputs from multiple domains.

We will provide participants with samples of swiss german texts from a variety of sources, including tweets, news comments etc. We encourage participants to use any additional resources to achieve high-quality results.

Data:

The data has been released here.

The training data consists of 2000 Swiss German Tweets. We encourage you to use any additional resources.

The test data consists of 5374 Tweets to be classified.

Some additional corpora that you might want to use:

• NOAH

• SB-CH

• SwissCrawl

• Sms4Science

If there are any problems or questions please contact: vode@zhaw.ch

Submission:

Send your submission file to vode@zhaw.ch until Friday March 27th 2020 Midnight anywhere on earth.

Evaluation:

We will provide a test set of tweets in Swiss German and a variety of other languages. We will evaluate Precision, Recall and F1 of binary - gsw vs. not_gsw - class predictions as well as Average Precision based on classifier scores.

Results:

Timeline:

Contact: vode@zhaw.ch

Organization:

Pius von Däniken, ZHAW InIT

Manuela Hürlimann, ZHAW InIT

Mark Cieliebak, ZHAW InIT