From Words to Wisdom: Automatically Generating Knowledge Graphs for Interpretable Educational AI

Johaun Hatchett, Debshila Basu Mallick, Richard Baraniuk

Track: Innovations in AI for Education (Day 1)

Paper Length: short-paper (2 pages + references)

Keywords: large language models, knowledge graph construction, information extraction

TL;DR: We introduce Words to Wisdom, a framework for generating knowledge graphs from plain text, offering a cost-effective solution to integrate factual content into LLM-based tools, mitigate their risk of hallucination, and enhance their interpretability.

Abstract: Large language models (LLMs) have emerged as powerful tools with vast potential across various domains. While they have the potential to transform the educational landscape with personalized learning experiences, these models face challenges such as high training and usage costs, and susceptibility to inaccuracies. One promising solution to these challenges lies in leveraging knowledge graphs (KGs) for knowledge injection. By integrating factual content into pre-trained LLMs, KGs can reduce the costs associated with domain alignment, mitigate the risk of hallucination, and enhance the interpretability of the models' outputs. To meet the need for efficient knowledge graph creation, we introduce Words to Wisdom (W2W), a domain-independent LLM-based tool that automatically generates KGs from plain text. With W2W, we aim to provide a streamlined KG construction option that can drive advancements in grounded LLM-based educational technologies.

Submission Number: 60