Faculty Collaborator: Rebecca Nedostup
About:
Naphat has been has been collaborating with Professor Rebecca Nedostup to develop a pipeline for comparing corpora extraction between Generative AI and Library Databases. This project is designed as an assignment for students in a Chinese History class to explore the concept of corpora extraction from both traditional and modern sources. The project is divided into two parts: researching the differences between the two sources using various types of questions about Chinese History (e.g., factual questions, reading suggestions) and creating a pipeline for future researchers or students to use. They utilized Jupyter Notebook to begin aggregating APIs for Generative AI (such as ChatGPT, Bard, Claude) and combined this with web scraping on the library database.
Agenda
Future Plans
Current Position of Projects
Process
Previous Attempts and Growth
Start
Main Project Goals and Initial Thoughts
Start
Quote: “The journey of a thousand miles begins with one step” - Lao Tsu
Project Goals
Objective: Crafting searches and queries depending on the platform.
Goals:
Professor's interest in understanding what kind of information is used to build different AI models (e.g., Google Books, Open AI, other models) and how that influences their outputs/performance.
Final Expectation: Develop a robust workflow for testing AI models, repeatable by students in future classes to continue the research.
Class Assignment: Students of Chinese history conduct directed search queries in library databases, search engines, and generative AI tools/platforms, comparing results and understanding strategies for crafting searches and queries depending on the platform and the corpora making up selected databases or training selected LLMs.
Tools
Platforms Used:
Library Database (e.g., BruKnow)
Search Engines (e.g., Google Search)
Generative AI (e.g., ChatGPT)
Initial Thought
Focus: “Robust workflow for testing AI models”
Question: Assignments?
Process
Quote: “The biggest room in the world is the room for improvement.” – Helmut Schmidt
Learning Outcomes
Library Database: Understanding strategies for crafting searches and queries depending on the platform.
Search Engine: Understanding strategies for crafting searches and queries depending on the platform.
Generative AI: Understanding strategies for crafting searches and queries depending on the platform.
Objective: Students compare query results and venture explanations for differences.
Research on Generative AI
Collaborators:
Chinese History Knowledge from Prof. Rebecca
Technical Knowledge in ChatGPT from me
Research: Initial comparison of ChatGPT answers to Google Search and Databases.
Critical AI Discussion
Discussion Leaders: Niamh McGuigan and Emily Ferrier
Topics:
Hands-on exploration of Generative AI
Generative AI discussions in class
Ethics in Generative AI
Tools
Platforms:
BruKnow
Google Search
Claude AI
Bard
ChatGPT
Website Idea
(No detailed content provided in the slides)
Learning Outcomes
Crafting Searches and Queries: Understanding strategies depending on the platform.
Comparison: Being able to compare query results and venture explanations for differences.
Present
Quote: “Be happy in the moment that’s enough. Each moment is all we need not more.” — Mother Teresa
Research on Comparison
Types of Questions:
Factual Question (e.g., Who is XXX?)
Reading Suggestions (e.g., What should I read to understand XXX?)
Discussion of Topics/Ethical Questions (e.g., Explain the controversy about XXX)
Guidelines for Generative AI
(No detailed content provided in the slides)
Search Queries Guidelines
Content: Aggregating sources for the search queries guide.
Resources in Generative AI
(No detailed content provided in the slides)
Future
Quote: “The biggest room in the world is the room for improvement.” – Helmut Schmidt
Future Research for Generative AI
Focus Areas:
Dive deeper into each topic: Factual Questions, Reading Suggestions, and Discussion of Topics/Ethical Questions.
Develop the website idea as a verification tool instead of just tools.
Guidelines for Assignment
Tasks:
Finish contents in the documents (including instructions to register and how to use them).
Develop the interactive portion of the search queries.