Honors Capstone Project
1
Advisor(s)
Khalid Al-Olimat, PhD
Ohio Northern University
Electrical and Computer Engineering
k-al-olimat@onu.edu
Xiangyi Cheng, PhD
Ohio Northern University
Mechanical Engineering
x-cheng@onu.edu
Confirmation
1
Document Type
Paper
Location
ONU McIntosh Center; Dean's Heritage Room
Start Date
18-4-2023 3:00 PM
End Date
18-4-2023 5:00 PM
Abstract
Large language models have been proven to be powerful tools for information retrieval, content summarization, data manipulation, and even holding conversations. In the past, conversational robots relied on decision trees to relay information to the user, which limited both the creativity and the total knowledge of these tools. Now, large language models such as GPT-3 have the entire internet at their disposal and creativity in the form of non-deterministic neural networks, generating natural, human-like conversations. The Greeting Robot Engineering Capstone uses large language models in its greeting robot to interact with both campus personnel and guests, while also providing accurate information about the university. This honors capstone enhancement builds upon this effort by completing two main tasks: Implementing a collection of various statistical metrics in the interaction system to track engagement, and performing a series of interviews with campus tour guides from the College of Engineering in order to ensure that the robot has exceptional knowledge of the questions that are asked by prospective students. These tasks culminate to establish a comprehensive system to interact with users effectively, while also collecting information that can shape future tour guide training.
Recommended Citation
Reichling, Logan, "Ensuring Accuracy and Engagement in Robots using Large Language Models" (2023). ONU Student Research Colloquium. 20.
https://digitalcommons.onu.edu/student_research_colloquium/2023/papers/20
Level of Access
Restricted to ONU Community
Restricted
Available to ONU community via local IP address and ONU login.
Ensuring Accuracy and Engagement in Robots using Large Language Models
ONU McIntosh Center; Dean's Heritage Room
Large language models have been proven to be powerful tools for information retrieval, content summarization, data manipulation, and even holding conversations. In the past, conversational robots relied on decision trees to relay information to the user, which limited both the creativity and the total knowledge of these tools. Now, large language models such as GPT-3 have the entire internet at their disposal and creativity in the form of non-deterministic neural networks, generating natural, human-like conversations. The Greeting Robot Engineering Capstone uses large language models in its greeting robot to interact with both campus personnel and guests, while also providing accurate information about the university. This honors capstone enhancement builds upon this effort by completing two main tasks: Implementing a collection of various statistical metrics in the interaction system to track engagement, and performing a series of interviews with campus tour guides from the College of Engineering in order to ensure that the robot has exceptional knowledge of the questions that are asked by prospective students. These tasks culminate to establish a comprehensive system to interact with users effectively, while also collecting information that can shape future tour guide training.