Formalized Common Knowledge

Home

Semantic Web

Projects

FAQ

Documentation

Releases

Downloads

Discussion

News

License

Last Update: April 8, 2009

Release 1.0

New to Release 1.0.2 of OpenCyc:

• The Cyc ontology now contains virtually all of Cyc's hundreds of thousands of terms, along with millions of assertions relating the terms to each other, forming an upper ontology whose domain is all of human consensus reality.
• English strings corresponding to all concept terms, to assist with search and display.
• Links between Cyc concepts and WordNet synsets.

Release 0.9

New to release 0.9 of OpenCyc:

• MUCH more KB content. Went from approx. 10,000 concepts to approx. 47,000 concepts.

• MANY more facts involving those concepts. Went from 60,000 assertions to 300,000.

• Inference is, on average, twice as fast despite the increased KB size.

• Various changes to the KB Browser, such as collapsible detail in the index pane.

• A functioning Time dimension (of Context Space).

• New inference parameters, including a "probably done" cutoff setting.

• The ability to save queries and query test specifications.

• Template-based OE from queries, which lets you use bindings from queries to drive template-based OE work (modify en masse a batch of facts that match a pattern).

• Sample common sense tests (type CST* in the Search box). You can ask these stored queries and view their justifications.

Known problems:

• The Java API is flaky under Windows, but mostly functional under Linux. This will be addressed in 0.9.1.

• There are nearly 1,000 assertions of the form (#$quotedIsa TERM #$Collection) that should just be deleted. They won't show up in the next version.

• The Planner has not been tested on this version.

Please let us know about other bugs you find by emailing opencyc@cyc.com.

A NOTE ABOUT FUTURE RELEASES OF OPENCYC:

There have been two major changes in approach that should lead to more frequent releases.

1. 1. It was the case, prior to this release, that we were holding back definitional assertions on terms until they were reviewed and found to be essentially complete across terms included in the KB.

For example, for all of the terms 1,2, . . . n in the OpenCyc KB, we would attempt to ensure that, for all i1 and i2 in n, (disjoint i1 i2) was known to be either true or false. This takes far too long to do manually across all definitional predicates, especially as the ontology fans out further down the hierarchy. Some automated methods have been identified to speed this process, but it is not a problem that will be solved any time soon.

This attempt at completeness was motivated, in part, by the belief that the OpenCyc ontology might be used, in its entirety, as a basis for a Semantic Web ontology. We now believe that the OpenCyc ontology will more likely be treated as a library from which a subset of terms, perhaps a large one, will be drawn upon for various Semantic-Web-like uses. For that usage, defintional completeness is not necessary, and we are not attempting it. As a result, we're including all of the definitional assertions for all of the terms that happen to appear in ResearchCyc (see research.cyc.com), the no cost (but not open source) version of the Cyc KB aimed at the academic and commercial research community.

We look forward to help from the OpenCyc community in addressing the incompleteness issue.

2. Since OpenCyc is a mostly unfunded effort, we have determined that we must limit our ambitions with respect to new product features. In short, we can't be innovators here (but you, of course, can be innovative in how you use OpenCyc!). Instead, we need to include whatever we can from features that happen to be developed for Full Cyc or for ResearchCyc. This will result in changes to the release roadmap.