Publication Date

5-2021

Date of Final Oral Examination (Defense)

3-12-2021

Type of Culminating Activity

Thesis

Degree Title

Master of Science in Computer Science

Department

Computer Science

Major Advisor

Casey Kennington, Ph.D.

Advisor

Bogdan Dit, Ph.D.

Advisor

Steve Cutchin, Ph.D.

Abstract

An incremental system takes advantage of upcoming data as early as possible. In other words, an incremental system processes received data incrementally. Incremental systems can be useful over non-incremental systems to build spoken dialog systems when we are looking for faster and more human-like behavior. For example, human-to-human conversations are incremental, as a listener does not wait for a speaker to finish speaking to begin understanding. Inspired by the fact that Robot-Ready Spoken Dialog Systems must be incremental and need to work distributedly, and IU framework "breaks" in a distributed architecture, I attempted to use the IU network to fulfill the incremental requirements and be able to extend the IU framework to work flawlessly in a distributed environment. This work aims to answer the question whether we can make a distributed IU network efficiently and consistently. More specifically, I explored the optimal ways to establish a complex IU data store that can facilitate the conservation and accessibility of the total generated IU data network in a distributed environment avoiding the \breaking" of the IU network, and act as a backbone for a final and complete "Robot-Ready" incremental dialog system. We evaluated the HRI response differences happening along with IU store implementation differences in a live, interactive study with robots and found out that humans do notice small performance differences and subconsciously become judgmental of robots' anthropomorphism characteristics in relation to the robots' performance.

DOI

10.18122/td.1804.boisestate

Share

COinS