Part 2/11:
The project's initial goal was to extract structured insights from around 22 Supreme Court opinions related to antitrust law. These decisions, originally stored as PDFs, were converted into text files. The plan was to utilize GPT-3, a state-of-the-art language model, to generate a JSON-formatted knowledge graph representing each court opinion as interconnected nodes with rich metadata.
Node Definition and Data Enrichment
Each node in this graph corresponds to a specific case citation or prior opinion. These nodes feature properties such as:
Case number and citation
Decision date and involved parties
Legal reasoning and interpretations
Precedential relationships