Leveraging large language models for accurate Cypher query generation : Natural language query to Cypher statements

Sammanfattning

The rise of Large Language Models (LLMs) has transformed various fields, including education, health, natural language processing, code generation, content creation, and more. The study seeks to use large language models to generate Cypher Queries based on natural language questions. The main objective of the study is to leverage and evaluate large language models and measure their Cypher Query Generation capabilities. The study utilizes GPT-3.5 turbo and Code Llama 2 for cypher generation in datasets collected and annotated across three categories: movies, network management, and companies. The study uses In-Context learning and QLoRA for fine-tuning the large language models. The BLEU and ROUGE evaluations indicate that GPT-3.5 turbo, utilizing the InContext learning method, outperforms the Code Llama 2, a fine-tuned model with QLoRA. The main challenges faced in this study are the unavailability of datasets and limited computational resources, such as GPU.