Capturing Real World Knowledge in ErgoAI

-- A Tutorial --

Vinay Chaudhri and Michael Kifer

The goal of this tutorial is to illustrate the process of capturing real world knowledge in Ergo, with the goal of enabling reasoning and question answering. The tutorial is based on the lecture notes of a course on knowledge representation that the first author taught at Stanford in 2011. We focus mainly on how to approach the problem of representing real world knowledge rather than on how to exercise the various features of Ergo. Therefore, for the most part, the tutorial assumes only basic familiarity with Ergo at the level of the first three lessons of the ErgoAI tutorial (Getting Started, The Basics, and More Advanced Features) plus the lessons Multi-file and Multi-module Knowledge Bases and User-defined Types and Default Values in the Frame Syntax. When more advanced features are used, we explain them inline.

This tutorial uses the object-oriented frame syntax (based on F-logic) of Ergo, which is geared towards the users who prefer object-oriented design to the more limited predicate-oriented syntax used in Prolog. For a comparison of frame vs predicate syntax, see Using Frame Syntax in Ergo. (Ergo supports the predicate syntax as well, but it is not used in this tutorial.) The reader is thus expected to be familiar with Ergo frames.

Technical questions? We have a user forum!

We use the California Driver’s Handbook as the source of the knowledge to be represented. The focus of the reasoning task will be on knowledge that can be objectively and operationally used by the driver of a vehicle (examples of this will be given in the tutorial). We will test the knowledge by using two sets of questions: a set of basic tests that verify each rule individually, a set of additional advanced tests, which exercise several rules at once, and a third set of questions created later, after the knowledge base has been tested with the first two sets. The third set is a hedge against "over-fitting" of representation to specific questions, which encourages making the representation as general as necessary.

Coherent Knowledge is the maker of the ErgoAI software, please visit our website for more information.