02.05.2014 Views

Proceedings - Österreichische Gesellschaft für Artificial Intelligence

Proceedings - Österreichische Gesellschaft für Artificial Intelligence

Proceedings - Österreichische Gesellschaft für Artificial Intelligence

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Preface<br />

Lexical-semantic contents serve as valuable sources of linguistic information for research and<br />

applications in natural language processing. Recent years have seen different kinds of efforts to<br />

enhance the usability of these resources. Successful attempts to integrate and merge lexical resources<br />

have led not only to the mapping of lexical databases of the WordNet and FrameNet type but also to<br />

the creation of online dictionary networks of different sorts. Enhancing usability also means to<br />

provide data for the human user by new sophisticated access interfaces as well as to make these data<br />

available for different kinds of NLP applications. In addition, with the increase of statistical methods<br />

in linguistics, these resources have also become more and more interesting for various research<br />

issues. The methods to achieve this rise in functionality and usability of lexical resources is a major<br />

topic of the workshop.<br />

The workshop was conceived as a forum for discussing recent developments and applications of<br />

various kinds of lexical-semantic resources including WordNets, ontologies, FrameNets, electronic<br />

dictionaries, internet lexicons, and other collaboratively built lexical semantic resources. It was a<br />

follow-up to a series of thematically related events: GSCL workshops in 2003, 2005, 2007, 2008, and<br />

a DGfS workshop in 2006.<br />

The six papers presented at the workshop focus partly on the creation of new lexical resources and<br />

partly on applications in language processing that are connected to lexical resources.<br />

Godoy and colleagues report on the construction of a lexical resource of Brazilian Portuguese verbs.<br />

Based on Levin’s hypothesis on the dependency of a verb’s argument structure on the semantic class<br />

it belongs to, the authors supply verbs with predicate decompositions which in turn define semantic<br />

verb classes. The decompositions are linked to syntactic structures exemplified by example<br />

sentences. The resulting resource will be available in printed form as well as a lexical database.<br />

Klyueva’s paper focuses on the detection of dissimilarities between Czech and Russian verb<br />

valencies. Building on two existing resources, a Czech-Russian bilingual dictionary and a Czech<br />

valency dictionary, she explores how far verb semantics and semantic class membership might<br />

account for the differences. The study will result in a digital lexical resource that contains those verb<br />

pairs in Czech and Russian that differ in their valency frames. This resource is supposed to support<br />

applications in automatic language processing such as machine translation.<br />

The paper by Lyngfelt and colleagues presents the development of a constructicon for Swedish as<br />

part of the Swedish FrameNet. Based on principles of Construction Grammar, it contains partially<br />

schematic multi-word units. The constructicon can be used as a lexicographic resource per se, is<br />

supposed to support language technology applications (e.g., facilitation of automatic syntactic<br />

analysis), and shall stimulate L2 acquisition research. The resources described in this paper are the<br />

result of activities carried out within Språkbanken, the national center for research on Swedish<br />

language resources at the University of Gothenburg.<br />

Pado and Utt present ongoing work on the creation of a Distributed Memory resource for German,<br />

aiming at the determination of lexical meaning from the word’s distributional behavior in corpora.<br />

Based on a German web-corpus, the study follows Baroni and Lenci’s original proposal (dealing with<br />

434

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!