Greek Lexicon
Enhances Large Language Models with a real-time, graph-based lexical tool for high-precision morphosyntactic and functional grammar analysis of Ancient Greek texts.
About
This project introduces a novel methodology to overcome traditional limitations in computational analysis of Ancient Greek texts, which often struggle with dialectical complexity and ambiguous word forms in rule-based systems or the imprecision of similarity-based Retrieval-Augmented Generation (RAG) methods. It bridges this gap by integrating Large Language Models (LLMs) with a dedicated, real-time lexical tool, enabling exact lemma identification and subsequent linguistic analysis. This tool-augmented approach allows LLMs to dynamically query a structured lexical database when precise information is needed, moving beyond simple context-stuffing.
Key Features
- Integrates Large Language Models with a real-time, structured lexical database
- Leverages a Neo4j Knowledge Graph built from a scholarly Greek-Dutch lexicon with over 40,000 lemma entries
- Enables dynamic lexicon queries by LLMs via the Model Context Protocol (MCP) server
- Supports diacritic-insensitive indexing for robust matching against unaccented Greek input
- Facilitates precise lexical data retrieval using Cypher queries for single and batch lookups
- 0 GitHub stars
Use Cases
- Perform high-precision morphosyntactic parsing of Ancient Greek texts
- Analyze functional grammar of Ancient Greek with enhanced accuracy
- Evaluate Large Language Models' performance on 'unseen' or less common Greek texts
