This project introduces a pioneering framework for a question-answering system, leveraging a knowledge graph (KG) extracted from the corpus of Securities and Exchange Commission (SEC) proceedings. It aims to enhance understanding and navigation of complex legal documents.
- implementation: This contains the full implementation of the system instance, showcasing the practical application of the framework.
- evaluation: Here, various semantic evaluation methodologies are provided, demonstrating the effectiveness and accuracy of the model.
- evaluation_helpers: Contains scripts used for generating evaluation answers from the model, crucial for validating the system's performance.
- kg_creation: Includes scripts for creating a nested knowledge graph, along with methods for uploading it to a Neo4j database, ensuring a robust and interconnected data structure.
- scripts/: Offers a collection of helper scripts for ontology. The ontology approach is heavily inspired by AnjaneyaTripathi's knowledge_graph.
- Presentation: Gain insights into the project through the presentation, available at Domain-Specific Large Language Models: A Law Case Study.
- Video Demo: View a practical demonstration of the system in action on YouTube.