Dies ist das Newsblog des Sprachwissenschaftlichen Instituts an der Ruhr-Universität Bochum.





Ruhr-Universität Bochum
Sprachwissenschaftliches Institut



Powered by PivotX - 2.3.11 
XML-Feed (RSS 1.0) 
XML: Atom Feed 

« Vortrag von Sonia Kan… | home | Vortrag von Peter Bos… »

Vortrag von Anette Frank am Dienstag, 8.01.2013, 16-18 Uhr

Montag, 17. Dezember 2012. Aus der Kategorie 'Vortragsreihe'. Das Sprachwissenschaftliche Institut lädt ein zu dem Vortrag von Anette Frank (Heidelberg): Corpus-based Learning of Formal Semantic Concepts: Genericity, Presupposition and more --
Statistical models of semantics are successfully applied to areas such as word sense disambiguation, acquisition of lexical semantic relations, or assignment of predicate-argument structure. This talk will take a closer look at linguistic concepts that have been thoroughly investigated in formal theoretical semantics – Genericity and Presupposition.

Theoretical semantics has long discussed the interpretation of generic vs. non-generic NPs as in (1) and (2). No clear consensus was reached as to the proper definition of ‘genericity’ or ‘generic NP’. Reiter & Frank (2010) develop a corpus-based model that distinguishes generic from non-generic NPs. By investigating the distributional features that permit their classification, we can gain some insights about the linguistic factors that determine the interpretation of NPs as generic.

(1) Lions eat meat.
(2) Lions have been observed in this area.

Presupposition is a relation between propositions that can be triggered by verbs: given (3), people can infer (4), due to their knowledge about the meaning of the verb win. Moreover, we can infer (4) from the negation of (3). Presupposition differs from other semantic relations that do not allow this inference, especially entailment, were ‘persistence under negation’ does not hold. Thus, in automatic inferencing tasks we need to distinguish presupposition from entailment, to avoid illicit inferences.

(3) Spain won/didn’t win the finals of the 2012 World Cup.
(4) Spain played the finals of the 2012 World Cup.

Tremper & Frank (2012) develop a discriminative classification approach for fine-grained semantic relations including presupposition that builds on core linguistic properties of presupposition and other inferential relations, such as entailment, troponoymy, antonymy and synonymy. Using corpus-based statistical methods, we obtain high classification performance for this difficult classification task.

The two phenomena I discuss yield different perspectives on the interplay between formal-theoretical and corpus-based statistical semantic analysis:
- statistical semantics can yield novel insights for formal semantics, and
- formal semantics can deliver crucial insights to guide statistical models of semantics.

In an outlook, I will briefly talk about further phenomena well-known from formal semantics that can feed into corpus-based statistical models of semantics.

Der Vortrag findet in Raum 3/159 statt.