Emmanouil Karystinaios
Emmanouil Karystinaios
Home
Posts
Demos
Publications
Contact
Light
Dark
Automatic
Symbolic Music
AnalysisGNN: From Prediction to Collaboration
Our ISMIR 2026 paper and the AnalysisGNN web demo show how a Roman-numeral prediction model can support a human-led annotation workflow.
Emmanouil Karystinaios
Nov 8, 2026
11 min read
Music Information Retrieval
ScorePrompts: Natural-Language Exploration of Symbolic Music Scores through Analysis
ScorePrompts aligns specialist score analyses with natural-language explanations, passage-level questions, and inspectable staff …
Emmanouil Karystinaios
,
Gerhard Widmer
PDF
Code
Project
Blog post
Conference program
From Prediction to Collaboration: Interactive Symbolic Music Analysis
RNHybrid combines contextual score representations with graph modeling to predict Roman numerals, complete partial annotations, and …
Emmanouil Karystinaios
,
Johannes Hentschel
,
Markus Neuwirth
,
Gerhard Widmer
PDF
Code
Project
Poster
Blog post
Dilemmadata: On the Interoperability of Heterogeneous Roman Numeral Datasets
Dilemmadata aligns AugmentedNet and the Distant Listening Corpus in a common note-wise format while retaining alternate analyses for …
Johannes Hentschel
,
Emmanouil Karystinaios
,
Gerhard Widmer
,
Markus Neuwirth
PDF
Dataset
ScorePrompts - Explaining Music Scores with Large Language Models
The ISMIR 2026 Late-Breaking Demo pairs structured music analysis with passage-specific questions, natural-language descriptions, and notation-linked evidence.
Emmanouil Karystinaios
Last updated on Nov 8, 2026
4 min read
Music Information Retrieval
,
Generative AI
How Far Can Pretrained LLMs Go in Symbolic Music? Controlled Comparisons of Supervised and Preference-based Adaptation
A controlled study of how instruction-tuned large language models adapt to symbolic music understanding and generation under supervised and preference-based training strategies.
Deepak Kumar
,
Emmanouil Karystinaios
,
Gerhard Widmer
,
Markus Schedl
PDF
DOI
Cite
×