Research Areas
Natural Language Processing
We develop state-of-the-art computational methods for analyzing, understanding, and generating human language. Our work spans both neural and symbolic approaches to NLP, with particular focus on low-resource languages and dialectal varieties.
Neuro-Symbolic AI
Our lab pioneers the integration of neural networks with symbolic reasoning systems. This hybrid approach combines the learning capabilities of deep learning with the interpretability and logical consistency of symbolic AI, enabling more robust and explainable language understanding systems.
Greek Language Technology
We specialize in developing NLP resources and tools for Greek and its dialectal varieties, including Cypriot, Pontic, Cretan, and Northern Greek dialects. Our work includes corpus development, benchmark creation, and model fine-tuning for Greek language varieties.
Digital Humanities
We apply computational methods to classical philology and historical linguistics, particularly focusing on ancient Greek texts. Our work includes knowledge graph extraction, automated analysis of historical texts, and tools for digital scholarship.
Current Projects
MEDEA-NEUMOUSA Platform
Advanced platform for ancient text analysis combining knowledge graph extraction with neuro-symbolic reasoning capabilities. Enables sophisticated computational analysis of classical Greek literature with automated entity recognition and relationship extraction.
Greek Regional Dialects Dataset (GRDD+)
Comprehensive dataset and benchmark for Greek dialectal varieties with fine-tuning experiments across multiple model architectures. Covers Cypriot, Pontic, Cretan, and Northern Greek dialects.
Linguistic Distance Framework
Seven-dimensional framework for analyzing linguistic change across Indo-European languages, combining computational methods with theoretical linguistics.
Thucydides goes RAGging (RAG-to-Coq)
Retrieval-augmented generation pipeline that ingests Thucydides' History, builds verified knowledge graphs, and feeds structured claims to Coq for proof obligations. RAG-to-Coq links long-context reading with symbolic verification, enabling explainable answers about classical events and actors.
Prototype combines semantic chunking, contrastive retrieval, and proof-term synthesis.
AI/NLP Applications
Our research translates into practical applications that address real-world challenges in language technology
Natural Language Understanding
Advanced systems for semantic parsing, question answering, and dialogue understanding with applications in virtual assistants and information retrieval
Text Generation & Summarization
Neural and hybrid approaches for generating coherent, contextually appropriate text and summarizing long documents
Multilingual & Low-Resource NLP
Specialized tools and models for Greek and its dialectal varieties, enabling NLP applications for under-resourced languages
Knowledge Extraction & Reasoning
Automated extraction of structured knowledge from text and integration with symbolic reasoning systems for explainable AI