Decision-Making During Language Understanding by Intelligent Agents

8 years 6 months ago

Download agi-conf.org

In cognitive modeling and intelligent agent design, a widely accepted architectural pipeline is Perception–Reasoning–Action. But language understanding, while a type of perception, involves many types of reasoning, and can even involve action, such as asking a clariﬁcation question about the intended meaning of an utterance. In the ﬁeld of natural language processing, for its part, the common progression of processing modules is Syntax–Semantics–Pragmatics. But this modularization lacks cognitive plausibility and misses opportunities to enhance eﬃciency through the timely application of knowledge from multiple sources. This paper provides a high-level description of semantically-deep, reasoningrich language processing in the OntoAgent cognitive agent environment, which illustrates the practical gains of moving away from a strict adherence to traditional modularization and pipeline architectures.

Marjorie McShane, Sergei Nirenburg

Real-time Traffic