Skip to main content
European Commission logo print header
Zawartość zarchiwizowana w dniu 2022-12-23

A Georgian, Russian, English, German Multilingual valency lexicon for natural language processing

Cel

Language engineering (LE) needs lexicons. Ideally, they specify the subcategorisation possibilities and thematic roles for verbs, and these specifications match the verbs' patterns of use in text corpora. Multilingual LE needs multilingual resources, where information for a number of languages is expressed in a co-ordinated way.

Currently there are no lexicons suitable for language engineering ('computational lexicons') for Georgian, and very few in the public domain for Russian. For English and German, much more is available, but still little that ties the lexicons to corpus evidence. Multilingual resources are very rare outside specialised domains, substantially because of the lack of theoretical understanding of how they might be coherently organised.

The GREG project will address all three issues. The first, by producing, and placing in the public domain, substantial computational lexicons of Georgian and Russian verbs (alongside their English and German counterparts). The second, by extending current work on the automatic acquisition of valency information from text corpora. And the third, by establishing the GREG lexicon as a multilingual resource and using it as a testbed for current theoretical research into multilingual representation.

The project brings together partners with complementary expertise. The Georgian Academy of Sciences and Tbilisi State University contribute in-depth knowledge of Georgian, in itself and in contrast to the other languages. The University of Stuttgart is a world leader in corpus processing for lexicographical purposes, and the University of Brighton, in formalisms for lexical information.

The languages for the project have been selected because of their salience to Georgia: English, for its international role; German, for the longstanding special relationship between Germany and Georgia; and Russian, because it is Georgia's largest trading partner and the most widely-spoken foreign language.

The scientific motivations to the project are fourfold: to deepen our understanding of the Georgian verbal system, and its relation to Indo-European systems, through detailed, formal, lexicological analysis; to explore relations between text corpora and lexical entries; to explore semantic case, and its relations to syntactic subcategorisation, using computer-based methods across large numbers of verbs; and to explore models of organisation for multilingual lexicons.

The work done in the project will be firstly, selecting a set of 1,000 Georgian verbs and their Russian, English and German counterparts; then identifying the subcategorisation patterns each verb appears in, associating the syntactic constituents with semantic cases (such as SOURCE, AGENT, GOAL), and linking translation equivalents across the languages.

The work will use a combination of existing dictionaries, corpus-analysis techniques and corpus-based and traditional lexicography. The lexicography will take place in Georgia, with Stuttgart guiding the corpus-based work. Brighton will work particularly on a formal framework for the project that makes it possible to express generalisations that apply to classes of verbs, across languages. This will ease the burden of the lexicography and produce a more succinct, usable resource.

Zaproszenie do składania wniosków

Data not available

System finansowania

Data not available

Koordynator

Universität Stuttgart
Wkład UE
Brak danych
Adres
Breitwiesenstr. 20-22
70565 Stuttgart
Niemcy

Zobacz na mapie

Koszt całkowity
Brak danych

Uczestnicy (3)