Actionability and incremental interpretation

Loading...
Thumbnail Image
Authors
Babkin, Petr
Issue Date
2018-05
Type
Electronic thesis
Thesis
Language
ENG
Keywords
Cognitive science
Research Projects
Organizational Units
Journal Issue
Alternative Title
Abstract
Language understanding has traditionally been viewed in AI research as an input preprocessing step, which translates unstructured text into some formal representation — prerequisite for downstream reasoning and action (Bar-Hillel, 1972). By contrast, a substantial amount of psycholinguistic evidence suggests that reasoning and even action happen in tandem with the increments of understanding. This work is pursued within the research program of Language Endowed Intelligent Agents (LEIA), which aims to unify language understanding with reasoning in the framework of a cognitive architecture. In this thesis I explore the avenue of modeling actionability associated with a meaning representation under construction. Incorporating actionability in a language-endowed agent allows to approximate the way humans understand language --- from the standpoint of their plans and goals. It also offers practical benefits namely, it helps optimize the agent's efficiency by eliminating the need for exhaustive processing of the input. I propose a computational model where meaning representation construction is performed incrementally in conjunction with tracking the state of the interlocutor's plan. I hypothesize that the possibility of grounding a partial meaning representation in the plan is a good indicator of its actionability. In addition, I show that tracking the state of the plan, in turn, facilitates disambiguation and resolution of lacunae in the meaning.
Description
May 2018
School of Humanities, Arts, and Social Sciences
Full Citation
Publisher
Rensselaer Polytechnic Institute, Troy, NY
Terms of Use
Journal
Volume
Issue
PubMed ID
DOI
ISSN
EISSN