A bouzouki outline whose strings continue into a spectral staff, beside Greek folk ornament

AI for Greek Music and Cultural Heritage

Music models are trained, overwhelmingly, on Western popular music. Applied to modal and microtonal idioms, to unusual instruments and to performance practices that do not fit a twelve-tone grid, they fail quietly: they return an answer, and the answer is wrong in ways that only someone who knows the music can see.

Greek traditional, folk and laiko music is exactly such a case, and it is on our doorstep. The work here is mostly about the part nobody wants to do: building the datasets that do not exist.

Datasets first

Lyra is a dataset for Greek traditional and folk music, released at ISMIR. Alongside it we have published a multimodal dataset for electric guitar playing-technique recognition, and studied instrument playing-technique recognition with Greek instruments as the use case — a task where the technique, not the instrument, carries the stylistic information.

A short demo shows playing-technique recognition on the bouzouki.

Beyond classification

The same material supports work on musical emotion, on the relationship between a film and its soundtrack, and on structure in improvised music — our Jazz Mapping framework analyses and indexes solos. Greek-language technology is not limited to music either: Greek2MathTex is a Greek speech-to-text framework that produces LaTeX equations.

Generation is the natural continuation of this line, and it is treated separately under generative models.

Where this runs

This is the core of our contribution to the PHAROS AI Factory, Greece’s national AI Factory under the EuroHPC Joint Undertaking, where we are building a platform for the analysis of Greek traditional music with genre, instrument and regional-style recognition, semantic indexing and natural-language search.

Selected publications

  • Papaioannou, C., Valiantzas, I., Giannakopoulos, T., Kaliakatsos Papakostas, M., & Potamianos, A. A Dataset for Greek Traditional and Folk Music: Lyra 23rd International Society for Music Information Retrieval Conference (ISMIR 2022) doi
  • Mitsou, A., Petrogianni, A., Vakalaki, E. A., Nikou, C., Psallidas, T., & Giannakopoulos, T. A multimodal dataset for electric guitar playing technique recognition Data in Brief, Vol 52 (2024) doi
  • Paraskevoudis, K., & Giannakopoulos, T. Instrument Playing Technique Recognition: A Greek Music Use Case In Worldwide Music Conference (pp. 124-136). Springer (2021) doi
  • Petrogianni, A., Vassilakis, D., Klampanos, I. A., Giannakopoulos, T., & Andreopoulou, A. Jazz Mapping: An Advanced Framework for Solo Analysis and Discovery in Jazz Music Audio Engineering Society Convention 155 (2023)
  • Geroulanos, A., & Giannakopoulos, T. Emotion Recognition in Music Using Deep Neural Networks Advances in Speech and Music Technology: Computational Aspects and Applications. Cham: Springer International Publishing, 193-213 (2022) doi
  • Touros, G., & Giannakopoulos, T. Video soundtrack evaluation with machine learning: Data availability, feature extraction, and classification Advances in Speech and Music Technology (2022) doi
  • Gkritzali, E., Kaliosis, P., Galanaki, S., Palogiannidi, E., & Giannakopoulos, T. Greek2MathTex: A Greek Speech-to-Text Framework for LaTeX Equations Generation 13th Helenic Conference on AI, SETN 2024, Piraeus, Greece, September 11-13, 2024. Proceedings 4 doi
  • Giannakopoulos, T., & Petridis, S. Detection and clustering of musical audio parts using Fisher linear semi-discriminant analysis 2012 EUSIPCO. IEEE