Konferenzvortrag: Searching for negative polarity items in DGS


Datum
13. Januar 2023 14:45 — 15:15
Ort
Göttingen, Deutschland
Dieser Vortrag ist nur auf Englisch verfügbar.

Zusammenfassung

Negative Polarity Items (NPIs) are expressions such as any that cannot occur in positive assertions with simple past (*Mary bought any books), but become acceptable under the scope of negation (Mary didn’t buy any books). While NPIs have been claimed to occur in virtually every language, evidence for their use in signed languages is still very limited. Our study seeks to identify potential NPI candidates in German Sign Language (DGS) by performing a collocation analysis of negation signs in the DGS Corpus dataset.

To this end, we compiled an initial list of NPI candidates in DGS by identifying signs that commonly occur in the context of negation, as indicated by their cooccurrence with licensers such as lexical negation signs, morphologically negated signs and headshake negation. The signs with the highest context ratio were then manually inspected to determine whether their observed uses actually indicate a use as NPIs.

To implement this approach, we addressed various limitations imposed by the available corpus data and lack of automatic tools for signed languages. These include a lack of syntactic structural information, e.g., to determine licensor scopes, a lack of explicit clause or sentence boundaries, and limitations of sample size (at >660,000 tokens, the DGS Corpus is very large for a sign language corpus, but quite small compared to many spoken language text corpora).

Our collocation analysis revealed several signs worth investigating further for their potential as NPIs. It also highlights challenges, such as identifying loan expressions from sign-supported German, and the need to include further types of licensors, such as downward-entailing operators and question markers, as well as complex-phrase NPIs. Nevertheless, our findings show that collocation analysis can be an important tool in the search for NPIs in signed languages.

Marc Schulder
Marc Schulder
Wissenschaftlicher Mitarbeiter für Computerlinguistik

Meine Forschungsinteressen umfassen Gebärdensprachen, Computerlinguistik und Open Science.