An Empirical Comparison of Unknown Word Prediction Methods

摘要: We compare two types of methods which deal with unknown words in the context computational grammars. Methods first type are based on idea supertagging and use a tagger to predict lexical descriptions for tokens given input. The second perform acquisition (LA) which, this paper, refers automatic new entries lexicon grammar. compared effect their application has parsing coverage accuracy GG grammar German (Crysmann, 2003). In particular, we adapt LA method Cholakov van Noord (2010) was originally developed Dutch Alpino system be used GG. Its impact test corpus newspaper texts is results reported previously same employed tagger. Furthermore, smaller experiment, show that linguistic knowledge provides can also sentence realisation.