Lexicon Acquisition for NLP: A Consumer Report
Sergei Nirenburg · 1994
Abstract Current natural language processing systems typically operate in a ‘demo’ mode—they sometimes feature sizeable grammars but seldom sizeable lexicons containing information about meaning. This difficulty of going beyond toy systems is one of the main bottlenecks of artificial intelligence in general. Scaling up the dictionaries (and other knowledge bases) of a knowledge-based system is, however, essential for the overall success of the field. There are several ways in which the indispensable massive knowledge acquisition program can in principle be conducted. One can envisage developing a machine learning system for automatic acquisition of vast quantities of knowledge through experimentation on a (sub)world.