Constructing a Multilingual Database of Verb Valence
Lars Hellan, Tore Bruland · 2013
We show the initial stage of an incremental on-line multilingual valence pattern demo, presently populated with two languages, Norwegian and Ga.The procedure for establishing the Norwegian part of the valence database resides in reusing material available in the computational HPSGgrammar Norsource, which has a rich array of lexical information, in part developed from earlier existing lexical resources for Norwegian.The procedure used for Ga is based on a Toolbox lexicon for Ga, with a first stage of processing enabling its data to join the conversion strategy used for Norwegian.A common template is used for the valence information display, although neither source fills in the template completely, reflecting their original differences in content.Essential among these is the availability of example sentences illustrating each valence option for each verb -this is available for Ga, but not for Norwegian.The results are implemented but not yet widely published, serving at the moment partly for self-improvement through exhibiting weaknesses in the resources from which they were derived, and partly for development of the multilingual design.