The Cyc Project. Cyc [Lenat et al. 1990] is a huge compendium of ``consensus knowledge'' about the world, crafted by hand by a relatively small team working for a period of over a decade. In contrast to our introductory comments, the Cyc project has proven that if a Knowledge Engineer can enter 50 axioms a day, then 10 Knowledge Engineers can enter a million axioms in 10 years. However, this is the exception which proves the rule: There have been no other such efforts and the success of Cyc as a general purpose store of knowledge still needs demonstation. The ``upper model'' of the Cyc knowledge base is now accessible to other researchers and we regard it as worthy of investigation.
The ISI Sensus. A team at ISI has constructed an ontology of over 50,000 nodes by aligning a variety of available linguistic assets. In particular, the core of Sensus is the WordNet lexical database [Miller 1990]; to this were merged the Penman Upper Model [Bateman 1990], the Longman's Dictionary of Contemporary English, the ONTOS [Carlson and Nirenburg 1990] ontology from CMU and at various times a variety of bilingual dictionaries. The merged ``ontology'' preserves the various formal labels from each of the contributing sources and these are useful for natural language processing. Techniques similar to Davis' work on consensus knowledge acquisition were developed to align the different contributing sources; in both cases the assumption is that the underlying semantics (in the case of Sensus this primarily means taxonomic relationships) are very similar (or identical) and that this similarity will be reflected in structural similarity with only a modest number of ``near misses.'' This technique seems most powerful for acquiring structural (e.g. taxonomic) information but less so for acquiring more general content. We believe that both the techniques and the combined lexicon produced in Sensus may of great value to our work.