Effortlessly Find and Understand Exam Regulations in your University with GPT-4

1 minute read

Introduction

I always hated reading through the exam regulations of my university. It’s a tedious process that involves reading through hundreds of pages of text to find the information you need. It’s also difficult to understand the regulations because they are written in a very formal and legal language. This makes it difficult to understand the regulations and find the information you need.

So i decided to build a program that can automatically find and understand the exam regulations of my university. I used GPT-4 to generate answers to questions about the exam regulations and Semantic Search to find broader information context. The program is written in Python and can be easily extended to find information in other universities.

AI Student

Yesterday i got an invite from OpenAI for the GPT-4 API. It’s a state-of-the-art language model that can generate text in a variety of domains. It’s been trained on a large dataset of text from the internet, which makes it capable of generating text that is indistinguishable from human-written text. It’s also capable of generating text in a variety of domains, including news articles, scientific papers, and even legal documents.

Web UI for Question Answering

This was one of my first real frontend development experience and i really enjoyed it. I’ve build it with Flask and Tailwindcss. You can choose your prefered exam regulations and start asking the AI, or you can decide to use the semantic search.

Technical Details

First i’ve created embeddings from the exam regulations documents, with OpenAI’s text-embedding-ada-002 model. I used an open-source embedding database called Chroma to store the embeddings offline. I opted against a cloud-based solution like Pinecone, as I preferred having complete control over my database and avoiding long-term expense. Additionally i utilized LangChain, a framwork for developing applications powered by language models, which made the application development process much more straightforward.