Acquisition of Unknown Word Paradigms for Large-Scale Grammars

摘要

Unknown words are a major issue for large-scale grammars of natural language. We propose a machine learning based al-gorithm for acquiring lexical entries for all forms in the paradigm of a given un-known word. The main advantages of our method are the usage of word paradigms to obtain valuable morphological knowl-edge, the consideration of different con-texts which the unknown word and all members of its paradigm occur in and the employment of a full-blown syntactic parser and the grammar we want to im-prove to analyse these contexts and pro-vide elaborate syntactic constraints. We test our algorithm on a large-scale gram-mar of Dutch and show that its application leads to an improved parsing accuracy.