The project aims to create a personal AI tutor tailored to specific university courses, such as CSC311 (Introduction to Machine Learning), CSC207 (Software Design), CSC258 (Computer Organization), etc. The backend system will streamline the addition of contents to the AI tutor’s knowledge base by automatically processing collected course resources - categorizing, summarizing, and tagging them into a Directed Acyclic Graph (DAG) structure, stored in a Neo4j database. A GraphRAG (Graph Retrieval Augmented Generation)-powered chatbot will enable learners to navigate through the course material, get answers to course-specific questions, and obtain guidance on prerequisite and post-requisite topics through the curation of resources and recommendations.
The first iteration of this project will be a system that is capable of ingesting various types of resources. The system will then generate a DAG that reflects the underlying relationships between all of the different topics. Each topic in the knowledge tree will be propagated with useful snippets of information as the result of the ingested resources. The second iteration of this project aims to introduce a GraphRAG model capable of traversing through the generated DAG. This GraphRAG model will leverage the relationships created by the DAG in order to compile relevant resources for the user. Lastly, we will aim to increase the chatbot’s capabilities by integrating contextual information about the user into the chatbot’s responses (e.g. quizzing the user to see what concepts they might still need to review). This will allow the system to narrow down what the user understands well, and what they still need to work on, and minimize repeating mentions of the same concepts.