Eleonora Mancini (@helemanc___) profile picture

Research Scientist @ Deezer | prev. Ph.D. Student in Computer Science @ Language Technologies Lab, University of Bologna

Parigi, Francia
Joined April 2023
Photo by Eleonora Mancini (@helemanc___)Photo by Eleonora Mancini (@helemanc___)Photo by Eleonora Mancini (@helemanc___)
Can lyrics be understood without words? 🎧 When we hear a song, meaning often travels through rhythm, tone, and phrasing as much as through text — yet most retrieval systems still rely solely on transcriptions. 1/6
1
1
2
8
3,524
Eleonora Mancini retweeted
Accepted at #NeurIPS26. See you in Paris (hopefully).
Diffusion model ✅ Generated item ✅ Successful item ✅ ➡️ Now: Which training data was influential? ➡️ Which are the items that, without them, such generation would not have been there? This is the counterfactual question our latest paper tries to answer more accurately. 1/5
1
7
2
41
2,081
Eleonora Mancini retweeted
#ICASSP2026 Sony AI is presenting 11 papers in Barcelona this May, covering #music understanding, #generativeaudio, audio-visual alignment, and #dataquality.
1
5
7
1,149
Eleonora Mancini retweeted
LEVERAGING WHISPER EMBEDDINGS FOR AUDIO-BASED LYRICS MATCHING, cmsworkshops.com/ICASSP2026/… Led by Eleonora Mancini @helemanc___ Poster Area 25. Thu, 7 May, 16:30 - 18:30. 5/5
1
2
158
Eleonora Mancini retweeted
Next week I will be attending and presenting some stuff at #ICASSP26. Hope to meet you all again after some time! Also, if you want to just chat, feel free to approach me. In the meantime, let's list here the papers I coauthor, together with their presentation time 👇 1/5
1
5
19
1,058
Eleonora Mancini retweeted
WEALY, our lyrics matching approach was also accepted at #ICASSP26, and now you have the code and checkpoints available! Paper: arxiv.org/abs/2510.08176 Code: github.com/helemanc/audio-ba… Checkpoints: huggingface.co/audio-based-l… Kudos to @helemanc___ for putting that together!
Can lyrics be understood without words? 🎧 When we hear a song, meaning often travels through rhythm, tone, and phrasing as much as through text — yet most retrieval systems still rely solely on transcriptions. 1/6
2
8
557
Eleonora Mancini retweeted
We're also back with the lab's seminar! Today we had @helemanc___ presenting her doctoral research "Multimodal AI for Human Expression Understanding". #NLP #multimodality #speech
1
7
387
Eleonora Mancini retweeted
We just released the full training code, as well as our best pretrained model! 🎉 Feel free to use our SOTA checkpoint in your own project with 3 lines of code, or to retrain on your own data using our Lightning+Hydra+Dora codebase ⚡️🐍 🌐 github.com/sony/sampleid
Eminem sampled Aerosmith, 50 Cent sampled Nina Simone, everybody sampled Chic... Many great songs sampled existing ones! Detecting this is the topic of our latest paper with @serrjoa at @SonyAI Barcelona 😎 tl;dr: multi-track dataset + few tricks = +18% boost over SOTA 🚀 1/N
1
5
1
17
1,807
Eleonora Mancini retweeted
Could we use the lyrics of songs for doing music version matching? Is text the best way to represent these lyrics? Is Barcelona a good place to do nice research? Answer to almost all these questions in Eleonora's paper 🔽🔽🔽
Can lyrics be understood without words? 🎧 When we hear a song, meaning often travels through rhythm, tone, and phrasing as much as through text — yet most retrieval systems still rely solely on transcriptions. 1/6
3
8
771
Eleonora Mancini retweeted
📢Student Volunteer Call is now OPEN! Help make ACL 2025 a success — support the organizing team and gain valuable experience. In return, your conference registration is covered. Deadline 🗓️ June 16. Call 👉2025.aclweb.org/calls/volunt… #StudentVolunteers #ACL25
8
1
27
4,090
🚀 MM-ArgFallacy 2025 has just released the training data! Participants can now dive into multimodal fallacy detection and fallacy classification with real political debate data! Ready to tackle the challenge? 🔗 nlp-unibo.github.io/mm-argfa… @ArgminingOrg #ACL2025
1
2
185
Exciting news! Our article has been accepted at JAIR! We explore 40+ speech datasets for mental health and neurological research and introduce a checklist to guide responsible dataset creation of speech datasets 💬 Check it out! 🔎 #survey #speech #checklist #datasets
New Article: "Promoting the Responsible Development of Speech Datasets for Mental Health and Neurological Disorders Research" by Mancini, Tanevska, Galassi, Galatolo, Ruggeri, and Torroni jair.org/index.php/jair/arti…
6
183
This #NeurIPS2024 paper, co-authored with Francesco Paissan, Luca Della Libera, and Cem Subakan, is another step toward more interpretable #deeplearning for #audio classifiers. Read the paper here: arxiv.org/pdf/2405.17615 The code will be available soon in #SpeechBrain. #AI
I am happy to share that "Listenable Maps for Zero-Shot Audio Classifiers" has been accepted in #NeurIPS2024. The paper explores #explainability for #zeroshot classifiers. 📋 Paper: arxiv.org/pdf/2405.17615 The code will be released soon!
1
3
36
2,441
Eleonora Mancini retweeted
ArgMining 2024 ended with a great photo of its wonderful community. Kudos to all of your great ideas, contributions, and help in organizing.
4
1
15
1,702
Excited to present our work, "MAMKit: A Comprehensive Multimodal Argument Mining Toolkit", today at ACL 2024! Catch our oral presentation and poster session starting from 3 p.m. @ArgminingOrg #ACL2024 #ArgumentMining #Multimodal #NLP
1
8
335