Confidences for Commonsense Reasoning

Abstract

Commonsense reasoning has long been considered one of the holy grails of artificial intelligence. Our goal is to develop a logic-based component for hybrid – machine learning plus logic – commonsense question answering systems. A critical feature for the component is estimating the confidence in the statements derived from knowledge bases containing uncertain contrary and supporting evidence obtained from different sources. Instead of computing exact probabilities or designing a new calculus we focus on extending the methods and algorithms used by the existing automated reasoners for full classical first-order logic. The paper presents the CONFER framework and implementation for confidence estimation of derived answers.

GK: Implementing Full First Order Default Logic for Commonsense Reasoning (System Description)

“…GK integrates the exception-handling algorithms described in the previous chapter with the algorithms designed for handling inconsistent KB-s and numeric confidences assigned to clauses, previously presented as a CONFER framework in [18]. The framework is built on the resolution method.…”

“…The described algorithms are implemented by the first author as a software system GK available at https://logictools.org/gk/. GK is written in C on top of our implementation of the CONFER framework [18] which is built on top of a high-performance resolution prover GKC [17] (see https://github.com/tammet/ gkc) for conventional first order logic. Thus GK inherits most of the capabilities and algorithms of GKC.…”

“…Some of these uncertainties, like fuzziness and confidence, can be represented numerically, while others, like rules with exceptions, are discrete. In [18] we present the design and implementation of the CONFER framework for extending existing automated reasoning systems with confidence calculation capabilities. In the current paper we present the implementation called GK for default logic [13], built by further extending the CONFER implementation.…”

Waste Reduction: Experiments in Sharing Clauses between Runs of a Portfolio of Strategies (Experimental Paper)

“…We conduct experiments on a modified version of the conventional high-performance resolutionbased FOL prover GKC [18] implemented by the author and available at https://github.com/tammet/gkc. The prover was built to serve as an underlying system for the commonsense reasoner GK, which implements numeric confidences [20] and default logic [19]. The GK system is a critical component of the pipeline [23] for commonsense reasoning using natural language.…”

“…The prover was built to serve as an underlying system for the commonsense reasoner GK, which implements numeric confidences [20] and default logic [19]. The GK system is a critical component of the pipeline [23] for commonsense reasoning using natural language.…”