Naive Physics Perplex

AI Magazine

The "Naive Physics Manifesto" of Pat Hayes (1978) proposes a large-scale project to develop a formal theory encompassing the entire knowledge of physics of naive reasoners, expressed in a declarative symbolic form. The theory is organized in clusters of closely interconnected concepts and axioms. More recent work on the representation of commonsense physical knowledge has followed a somewhat different methodology. The goal has been to develop a competence theory powerful enough to justify commonsense physical inferences, and the research is organized in microworlds, each microworld covering a small range of physical phenomena. In this article, I compare the advantages and disadvantages of the two approaches.


An Interface for Crowd-Sourcing Spatial Models of Commonsense

AAAI Conferences

Commonsense is a challenge not only for representation and reasoning but also for large scale knowledge engineering required to capture the breadth of our "everyday" world. One approach to knowledge engineering is to "outsource" the effort to the public through games that generate structured commonsense knowledge from user play. To date, such games have focused on symbolic and textual knowledge. However, an effective commonsense reasoning system will require spatial and physical reasoning capabilities. In this paper, I propose a tool for gathering commonsense information from ordinary people. It is a user-friendly 3D sculpting tool for modeling and annotating models of physical objects and spaces.


Commonsense Causal Reasoning Using Millions of Personal Stories

AAAI Conferences

The personal stories that people write in their Internet weblogs include a substantial amount of information about the causal relationships between everyday events. In this paper we describe our efforts to use millions of these stories for automated commonsense causal reasoning. Casting the commonsense causal reasoning problem as a Choice of Plausible Alternatives, we describe four experiments that compare various statistical and information retrieval approaches to exploit causal information in story corpora. The top performing system in these experiments uses a simple co-occurrence statistic between words in the causal antecedent and consequent, calculated as the Pointwise Mutual Information between words in a corpus of millions of personal stories.


Choice of Plausible Alternatives: An Evaluation of Commonsense Causal Reasoning

AAAI Conferences

Research in open-domain commonsense reasoning has been hindered by the lack of evaluation metrics for judging progress and comparing alternative approaches. Taking inspiration from large-scale question sets used in natural language processing research, we authored one thousand English-language questions that directly assess commonsense causal reasoning, called the Choice Of Plausible Alternatives (COPA) evaluation. Using a forced-choice format, each question gives a premise and two plausible causes or effects, where the correct choice is the alternative that is more plausible than the other. This paper describes the authoring methodology that we used to develop a validated question set with sufficient breadth to advance open-domain commonsense reasoning research. We discuss the design decisions made during the authoring process, and explain how these decisions will affect the design of high-scoring systems. We also present the performance of multiple baseline approaches that use statistical natural language processing techniques, establishing initial benchmarks for future systems.


A Model-Based Approach to Visual Reasoning on CNLVR Dataset

AAAI Conferences

Visual Reasoning requires an understanding of complex compositional images and common-sense reasoning about sets of objects, quantities, comparisons, and spatial relationships. This paper presents a semantic parser that combines Computer Vision (CV), Natural Language Processing (NLP) and Knowledge Representation & Reasoning (KRR) to automatically solve visual reasoning problems from the Cornell Natural Language Visual Reasoning (CNLVR) dataset. Unlike the data-driven approaches applied to the same dataset, our system does not require any training but is guided by the knowledge base that is manually constructed. The system demonstrates robust overall performance which is also time and space efficient. Our system achieves 87.3% accuracy, which is 17.6% higher over the state-of-the-art method on raw image representations.