Knowledge Retrieval With Functional Object-Oriented Networks

TL;DR

This paper compares three search algorithms for extracting task trees from a large knowledge graph called FOON, demonstrating that iterative deepening performs well overall, with heuristics providing benefits in specific cases.

Contribution

The study implements and compares multiple search algorithms for task tree retrieval in FOON, highlighting the effectiveness of iterative deepening and heuristic functions.

Findings

Iterative deepening performs strongly overall.

Heuristics improve search in certain situations.

Performance varies with different heuristics.

Abstract

In this experiment, three different search algorithms are implemented for the purpose of extracting a task tree from a large knowledge graph, known as the Functional Object-Oriented Network (FOON). Using a universal FOON, which contains knowledge extracted by annotating online cooking videos, and a desired goal, a task tree can be retrieved. The process of searching the universal FOON for task tree retrieval is tested using iterative deepening search and greedy best-first search with two different heuristic functions. The performance of these three algorithms is analyzed and compared. The results of the experiment show that iterative deepening performs strongly overall. However, different heuristics in an informed search proved to be beneficial for certain situations.