Chapter 17 Knowledge Processing on an Extended Wordnet 17.1 Very Large Knowledge Bases and Wordnet 17.1.1 Desirable Features and What Wordnet Can Ooer

Commonsense reasoning requires extensive knowledge. In order to be useful for practical artiicial intelligence problems, a knowledge base needs to have millions of concepts and relations. The nature of knowledge is such that even solutions to relatively small, narrow artiicial intelligence problems require a great deal of knowledge. Unfortunately, knowledge cannot be modularized or partitioned. WordNet contains almost all English words, grouped into a hundred thousand synsets, which in turn are linked by semantic relations. But most importantly, each synset has a gloss that, when disambiguated, may increase the number of semantic relations by an order of magnitude. It takes a large eeort to build a very large knowledge base. Because of this, it is highly desirable that the knowledge base be generally applicable, such that it can be adaptated to a particular problem with as little eeort as possible. If the knowledge base is to be generally 684