Leveraging Large Language Models for KG Construction and Reasoning

Overview

Large language models (LLMs) have recently demonstrated impressive performance across various natural language processing (NLP) tasks, showcasing their ability to understand and generate human-like text. However, their potential in constructing and reasoning over Knowledge Graphs (KGs) remains underexplored. KGs are structured representations of knowledge that connect entities and their relationships in a graph form. These graphs can be utilized in numerous applications, including question-answering systems and recommendation engines.

Objectives

This thesis aims to investigate the use of LLMs in constructing KGs from text and performing reasoning tasks. KG construction includes tasks such as named entity recognition (NER), relation extraction (RE), event extraction (EE), and entity linking (EL). Additionally, KG reasoning tasks include link prediction, which involves predicting missing relationships between entities, and other tasks that enrich the KG by uncovering hidden knowledge and providing deeper insights. The goal is to leverage the advanced capabilities of LLMs to enhance these processes, automating and improving KG construction and reasoning tasks.

Prerequisites

• Strong background in natural language processing methods (e.g., named entity recognition, text embedding, large language models)
• Proficiency with Python and deep learning frameworks (e.g., PyTorch or TensorFlow)
• Knowledge of knowledge graphs and semantic web data

Tasks

• Develop methods for constructing KGs from textual data using LLMs
• Benchmark the performance of these methods against existing approaches
• Benchmark the performance of LLMs in KG reasoning tasks
• Summarize the impact of LLMs on the accuracy and efficiency of KG construction and reasoning tasks

Related material

1. LLMs for Knowledge Graph Construction and Reasoning: Recent Capabilities and Future Opportunities (https://arxiv.org/abs/2305.13168)
2. Iterative Zero-Shot LLM Prompting for Knowledge Graph Construction (https://arxiv.org/abs/2307.01128)
3. Complex Logical Reasoning over Knowledge Graphs using Large Language Models (https://arxiv.org/abs/2305.01157)