Discrete vs Continuous Embeddings in Developing Intelligent-Solution Suggestive Plane-Crash RAG System

The goal of the paper is to demonstrate and evaluate continuous versus discrete embeddings in developing an intelligent-solution suggestive RAG recommender system for plane accident/crash analysis. The paper incorporates theoretical and practical knowledge from the areas of Accident Analysis, LLMs, NLP, Recommender Systems, and modern AI technologies. An indepth analysis of the current era of LLMs and the tools available is presented, along with the results of applying discrete vs. continuous embeddings in the development of the plane crash RAG analyzer and solution-suggestive recommender using the National Transportation Safety Board (NTSB) dataset. This research seeks to facilitate Crash/Accident Analysts and Investigators in conducting faster causal analysis and large-scale comparisons using the historical context of crash occurrences, as well as providing practical solutions for accident remediation and preventive policy development.