
Sahar Ghannay
Présentation
Associate professort at LIMSI, CNRS, Université Paris-Saclay
Team: ILES
Email: sahar.ghannay@limsi.fr
Website: https://saharghannay.github.io
Short Bio
Sahar Ghannay is an associate professor at Université Paris-Saclay, in the CNRS, LISN research center, since September 2018.
She received a PhD in Computer Science from Le Mans University on Septembre 2017. Her thesis work is part of the ANR VERA (AdVanced ERror Analysis for speech recognition) project. During her PhD, she spent a few months as @ visiting researcher at Apple within the Siri Speech team.
As a postdoctoral researcher at LIUM, she worked on neural end-to-end systems for the detection of named entities, speech understanding, as part of the Chist-Era M2CR (Multimodal Multilingual Continuous Representation for Human Language Understanding) project.
Her main research interests are continuous representations learning and their application to natural language processing and speech recognition tasks, semantic information extraction form spoken and writen language and dialog system.
CV
Education
- PHD in computer science, at LIUM, Le Mans université, 2017
- MS in computer science, Le Mans université, 2013
- BS in computer science, Le Mans université and université de sfax, 2011
Work Experience
- 2018 - now: Associate Professort at LISN, CNRS, Université Paris-Saclay
- 2017-2018: Post-doc at LIUM, Le Mans université
- 2017 (4 moths): Internship at Apple within the Siri Speech team at Cupertino
- April 2013-Sept. 2014: research engineer
Publications
Publications
|
Evaluating the Confidentiality of Synthetic Clinical Texts Generated by Language Models23rd International Conference on Artificial Intelligence in Medicine (AIME), Jun 2025, Pavie, Italy
Communication dans un congrès
hal-05046326
v1
|
|
Small Language Models are Good Too: An Empirical Study of Zero-Shot ClassificationThe 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), May 2024, TURIN, Italy. pp.14923-14936
Communication dans un congrès
hal-04519930
v2
|
|
New Semantic Task for the French Spoken Language Understanding MEDIA BenchmarkThe 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), May 2024, Torino, Italy. pp.12227-12246
Communication dans un congrès
hal-04523286
v1
|
|
mALBERT: Is a Compact Multilingual BERT Model Still Worth It?The 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), May 2024, Torino, Italy. pp.11023-11029
Communication dans un congrès
hal-04520797
v1
|
|
Multi-Level Information Retrieval Augmented Generation for Knowledge-based Visual Question AnsweringThe 2024 Conference on Empirical Methods in Natural Language Processing (EMNLP 2024), Nov 2024, Miami, United States. pp.16499-16513, ⟨10.18653/v1/2024.emnlp-main.922⟩
Communication dans un congrès
hal-04852275
v1
|
Specialized Semantic Enrichment of Speech Representations2023 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW), Jun 2023, Rhodes Island, Greece. pp.1-5, ⟨10.1109/ICASSPW59220.2023.10193452⟩
Communication dans un congrès
hal-04425710
v1
|
|
|
Projet Gender Equality Monitor (GEM)CORIA-TALN 2023 30e Conférence sur le Traitement Automatique des Langues Naturelles (TALN), Jun 2023, Paris, France. pp.21-21
Communication dans un congrès
hal-04208588
v1
|
|
Continual self-supervised domain adaptation for end-to-end speaker diarizationIEEE Spoken Language Technology Workshop (SLT 2022), IEEE Speech and Language Processing Technical Committee, Jan 2023, Doha, Qatar. pp.626 - 632, ⟨10.1109/SLT54892.2023.10023195⟩
Communication dans un congrès
hal-03824546
v1
|
|
Benchmarking Transformers-based models on French Spoken Language Understanding tasksINTERSPEECH 2022, Sep 2022, Incheon, South Korea
Communication dans un congrès
hal-03715340
v2
|
|
Analyzing BERT Cross-lingual Transfer Capabilities in Continual Sequence LabelingFirst Workshop on Performance and Interpretability Evaluations of Multimodal, Multipurpose, Massive-Scale Models, Oct 2022, virtual, South Korea
Communication dans un congrès
hal-03824597
v1
|
|
Evaluating the carbon footprint of NLP methods: a survey and analysis of existing toolsEMNLP, Workshop SustaiNLP, Nov 2021, Punta Cana, Dominican Republic
Communication dans un congrès
hal-03435068
v1
|
|
Overlap-Aware Low-Latency Online Speaker Diarization Based on End-To-End Local SegmentationIEEE Automatic Speech Recognition and Unserstanding Workshop, Dec 2021, Cartagena, Colombia
Communication dans un congrès
hal-03375330
v1
|
Error analysis applied to end-to end spoken language understanding45th International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2020), May 2020, Barcelona, Spain. pp.8514-8518, ⟨10.1109/ICASSP40776.2020.9054455⟩
Communication dans un congrès
hal-02465899
v1
|
|
|
What is best for Spoken Language Understanding: Small but Task-dependant Embeddings or Huge but Out-of-domain Embeddings?45th International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2020), May 2020, Barcelona, Spain. pp.8114-8118, ⟨10.1109/ICASSP40776.2020.9053278⟩
Communication dans un congrès
hal-02503694
v1
|
|
A Metric Learning Approach to Misogyny CategorizationWorkshop on Representation Learning for NLP, Jul 2020, Online, France. pp.89-94, ⟨10.18653/v1/2020.repl4nlp-1.12⟩
Communication dans un congrès
hal-02989293
v1
|
|
Experiments from LIMSI at the French Named Entity Recognition Coarse-grained taskConference and Labs of the Evaluation Forum, Sep 2020, Thessaloniki, Greece
Communication dans un congrès
hal-04395545
v1
|
|
A Cooking Knowledge Graph and Benchmark for Question Answering Evaluation in Lifelong Learning ScenariosInternational Conference on Applications of Natural Language to Information Systems, Elisabeth Métais and Farid Meziane and Helmut Horacek and Philipp Cimiano, Jun 2020, Saarbrücken, Germany
Communication dans un congrès
hal-03006228
v1
|
|
Neural Networks approaches focused on French Spoken Language Understanding: application to the MEDIA Evaluation TaskIn Proceedings of The 28th International Conference on Computational Linguistics (COLING’2020), 2020, Dec 2020, Barcelona (online), Spain
Communication dans un congrès
hal-03007482
v1
|
|
A Comparison of Metric Learning Loss Functions for End-To-End Speaker VerificationInternational Conference on Statistical Language and Speech Processing, Oct 2020, Cardiff, United Kingdom. pp.137-148, ⟨10.1007/978-3-030-59430-5_11⟩
Communication dans un congrès
hal-02989334
v1
|
|
Lifelong learning and task-oriented dialogue system: what does it mean?International Workshop on Spoken Dialogue Systems Technology, Apr 2019, Siracusa, Italy
Communication dans un congrès
hal-02301089
v1
|
|
Simulation d'erreurs de reconnaissance automatique dans un cadre de compréhension de la paroleXXXIIe Journées d'Etudes sur la Parole (JEP 2018), Jun 2018, Aix-en-Provence, France
Communication dans un congrès
hal-01757770
v1
|
|
Simulating ASR errors for training SLU systemsLREC 2018, May 2018, Miyazaki, Japan
Communication dans un congrès
hal-01715923
v1
|
Task Specific Sentence Embeddings for ASR Error DetectionInterspeech 2018, Sep 2018, Hyderabad, India. ⟨10.21437/Interspeech.2018-2211⟩
Communication dans un congrès
hal-01870864
v1
|
|
Représentations de phrases dans un espace continu spécifiques à la tâche de détection d'erreursXXXIIe Journées d'Etudes sur la Parole (JEP 2018), Jun 2018, Aix-en-Provence, France
Communication dans un congrès
hal-01757774
v1
|
|
|
End-to-end named entity and semantic concept extraction from speechIEEE Spoken Language Technology Workshop, Dec 2018, Athens, Greece. ⟨10.1109/slt.2018.8639513⟩
Communication dans un congrès
hal-01987740
v2
|
|
Enriching confusion networks for post-processingStatistical Language and Speech Processing 2017, Oct 2017, Le Mans, France
Communication dans un congrès
hal-01585768
v1
|
|
ASR error management for improving spoken language understandingInterspeech 2017, Aug 2017, Stockholm, Sweden
Communication dans un congrès
hal-01526298
v1
|
Recent improvements on error detection for automatic speech recognition1st International Workshop on Multimodal Media Data Analytics (MMDA 2016), in Conjunction with the 22nd European Conference on Artificial Intelligence, 2016, The Hague The, Netherlands
Communication dans un congrès
hal-01433168
v1
|
|
Word embedding evaluation and combination10th edition of the Language Resources and Evaluation Conference (LREC 2016), 2016, Portorož, Slovenia
Communication dans un congrès
hal-01433185
v1
|
|
Acoustic word embeddings for ASR error detectionInterspeech 2016, 2016, San Francisco (CA, USA), Unknown Region
Communication dans un congrès
hal-01433176
v1
|
|
Evaluation of acoustic word embeddingsRepEval@ACL 2016: The 1st Workshop on Evaluating Vector-Space Representations for NLP, 2016, Berlin, Germany
Communication dans un congrès
hal-01433181
v1
|
|
Combining continous word representation and prosodic features for ASR error prediction3rd International Conference on Statistical Language and Speech Processing (SLSP 2015), 2015, Budapest, Hungary
Communication dans un congrès
hal-01433203
v1
|
|
Word embeddings combination and neural networks for robustness in ASR error detection2015 European Signal Processing Conference (EUSIPCO 2015), 2015, Nice, France
Communication dans un congrès
hal-01433210
v1
|
|
Which ASR errors are hard to detect?Workshop Errors by Humans and Machines in multimedia, multimodal and multilingual data processing (ERRARE 2015), 2015, Sinaia, Romania
Communication dans un congrès
hal-01433201
v1
|
|
Using Hypothesis Selection Based Features for Confusion Network MT System CombinationThird Workshop on Hybrid Approaches to Translation (HyTra), EACL 2014, 2014, Gothenburg, Sweden
Communication dans un congrès
hal-01433229
v1
|
|
Entity-aware cross-modal pretraining for Knowledge-Based Visual Question Answering2024
Pré-publication, Document de travail
cea-04910767
v1
|
|
Has AlphaFold3 achieved success for RNAs?2025
Pré-publication, Document de travail
hal-04911522
v1
|
|
LIMSI_UPV at SemEval-2020 Task 9: Recurrent Convolutional Neural Network for Code-mixed Sentiment Analysis2021
Pré-publication, Document de travail
hal-03294371
v1
|
|
Semantic enrichment towards efficient speech representationsLISN. 2023
Rapport
hal-04425932
v1
|
|
Étude sur les représentations continues de mots appliquées à la détection automatique des erreurs de reconnaissance de la paroleInformatique et langage [cs.CL]. Université du Maine, 2017. Français. ⟨NNT : 2017LEMA1019⟩
Thèse
tel-01661491
v1
|