ChatGPT robotic process automation and AI chatbots for education
This study delved into the development of a novel educational chatbot application known as Edu-AI. The overarching goal was twofold: first, to optimize prompt engineering by selectively utilizing state-of-the-art Language Models (LLMs) to generate high-quality prompts for users; and second, to pro...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2023
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/171966 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | This study delved into the development of a novel educational chatbot application known as Edu-AI.
The overarching goal was twofold: first, to optimize prompt engineering by selectively utilizing state-of-the-art Language Models (LLMs) to generate high-quality prompts for users; and second, to provide
accurate responses to user inquiries. Multiple techniques, including introspection and fact-checking,
were implemented to ensure response precision.
During development, it became evident that inherent language comprehension differences existed
between LLMs and humans. However, meticulously crafted prompts could mitigate this limitation and
empowered LLMs to produce more refined responses. The Edu-AI application was founded on an
innovative recursive GPT prompt engineering approach, which involved the LLM iteratively evaluating
and ranking the prompts it generated. This method was seamlessly incorporated into Edu-AI to amplify
prompt generation, allowing users to select superior options and obtain correspondingly improved
responses.
Additionally, Edu-AI employed fact-checking techniques inspired by prompt pattern concepts. This
directed the LLM to furnish corroborating information for its generated responses, thereby upholding
accuracy and credibility.
The developmental process faced substantial challenges, primarily stemming from the relative nascency
of LLMs and the limited exploration of recursive prompting techniques up to that point. The consequent
scarcity of comprehensive resources introduced further complexity. However, the endeavor provided
valuable insights into prompt optimization and fact-checking methods for advancing chatbot quality. |
|---|