NextKB: An Integrated, Linguistically, Visually, and Inferentially Rich Open-license Knowledge Base

Our group has used contents from ResearchCyc and OpenCyc in almost all of our projects for several decades now. That is, we use our own reasoning engine, running with knowledge extracted semi-automatically from a Cyc KB. While we admire the Cyc reasoning engine in many ways, we find it useful to build our own reasoning engine, FIRE, optimized for our purposes. For example, FIRE has analogical reasoning and learning built in more deeply than backchaining. SME is already available as open-source software, and FIRE, along with the rest of CogSketch, will be available as open-source software before long.

We think the AI commmunity could more broadly benefit from an open-license KB resource that integrates natural langauge information, visual and spatial representations, and supports reasoning. To this end, we are making available a cleaned up version of the open-license knowledge base we are using in-house. It incorporates and integrates

The OpenCyc ontology, from Cycorp
The NuLex lexicon, developed by our group
Material from FrameNet, VerbNet, and WordNet, integrated with NuLex and the OpenCyc ontology
Visual and spatial representations our group developed in creating CogSketch and other projects
Representations to support qualitative reasoning, including QP theory. We note that a number of the predicates are implemented in FIRE via procedural attachments (aka outsourced predicates). We make the FIRE manual available (see below) for additional documentation.

Data License

The complete data in NextKB is available under the Creative Commons Attribution 4.0 license. In addition to contributions from members of Northwestern University's Qualitative Reasoning Group, this resource relies on the following other resources:

OpenCyc KB contents, from Cycorp. Their heroic and persistent efforts spanning decades have produced a clean ontology and ideas that are extremely usable for the AI and Cognitive Science communities, and we hope this resource increases interest in Cyc itself.
WordNet contents, from the Princeton WordNet project. This project, started by the late George Miller, provides useful lexical information that can be nicely combined with the others.
VerbNet contents, from the University of Colorado VerbNet project.
FrameNet contents, from the Berkeley FrameNet project. This project, started by the late Charles Fillmore, provides a rich language-centric view on semantics that we have been integrating with the OpenCyc ontology.

We express our gratitude to those who contributed to the above resources. We have filtered, transformed, and combined them, along with making our own contributions, so errors and limitations may well be introduced by us rather than by them. The QRG contributions include the linkages between the four resources above, plus support for qualitative, visual, spatial, and analogical reasoning and learning.

Funding for the Qualitative Reasoning Group's work represented here includes the National Science Foundation, via the Spatial Intelligence and Learning Center, the Office of Naval Research, the Air Force Office of Scientific Research, the Defense Advanced Projects Research Agency, and IBM.

In academic papers where NextKB was used, we ask that you please cite the following paper:

Forbus, K. & Hinrichs, T. (2017). Analogy and Qualitative Representations in the Companion Cognitive Architecture. AI Magazine 38(4):34-42.

To give credit to NextKB in software that uses it, please put this text in a user-visible place:

This work includes data from NextKB, which was compiled by the Qualitative Reasoning Group at Northwestern University. NextKB is freely available under the Creative Commons Attribution 4.0 license from http://qrg.northwestern.edu/nextkb/index.html. The included data was created by contributors to the Qualitative Reasoning Group, contributors to Cycorp's; OpenCyc, University of California at Berkeley's; FrameNet project, the VerbNet project, and Princeton University's; WordNet project. For details of attributions, please see http://www.qrg.northwestern.edu/nextkb/license.html

Downloading NextKB

There are three ways to download the knowledge base:

Flat files. This zip archive contains one text file per microtheory. Each in-microtheory statement indicates that every statement until the next such directive is explicitly stored in the given microtheory.
A FIRE-style KB. This zip archive contains such a knowledge base. It can be used with Case Mapper for easy browsing, simply unzip it into a new directory and open it from Case Mapper. The KB browser built into CaseMapper simplifies getting familiar with the contents. NextKB is also used in our CogSketch sketch understanding system, which includes the same browser.
An OWL subset. This is very much a work in progress, what we have so far are the microtheories below (4 out of roughly 690). These files do not include higher-order predicates, predicates of arity higher than two, and microtheory structure implemented via named graphs.

NextKB is a work in progress. We plan on expanding the reasoning capabilities and OWL translations considerably. Collaborators who wish to help are welcome, and we will probably move this to github shortly to facilitate participation.

Documentation

You may find the following materials useful:

Back to Resources page | Back to QRG Home Page