Conference Keynote Speakers

Keynote Talks


Headshot1
Paula Carvalho
University of Aveiro | INESC-ID Lisboa

Paula Carvalho, Ph.D. in Linguistics (2008), is a researcher at INESC-ID Lisboa and currently an Assistant Professor at the University of Aveiro, Portugal. Her research interests span Natural Language Processing, Corpus Linguistics, and Computational Social Sciences, with a focus on sentiment analysis, irony recognition, and information disorder detection. Carvalho has been actively involved in interdisciplinary research projects aimed at advancing language resources and methodologies to address challenges in the digital humanities and social sciences. Currently, her research particularly focuses on the analysis and detection of disinformation on social media, including hate speech and conspiracy theories. She has recently led two research projects in this area, HATE COVID-19 (https://hate-covid.inesc-id.pt) and MAICT (https://maict.inesc-id.pt/). Additionally, she led a working group in kNOwHATE (https://knowhate.eu), focused on creating language resources to address online hate speech and counter speech in Portuguese.

Title: An Interdisciplinary and culturally sensitive approach to unpack online hate speech

9th September @ 9:00, MACI Auditorium (Room 1)

Abstract:

In the digital age, online hate speech (OHS) poses a significant threat to democratic societies by fostering discrimination, violence, and social division while normalizing harmful behavior. Automation is essential for providing real-time, effective responses to the vast amount of OHS disseminated daily. However, the lack of consensus on the definition of hate speech and its broad scope presents significant challenges in developing accurate identification strategies. Moreover, automated solutions often overlook the nuanced intricacies underlying hateful discourse, neglecting subtle or indirect manifestations. Advanced language models have the potential to capture complex hate speech manifestations, yet their efficacy depends on access to extensive and diverse data, which is often limited, especially in resource-scarce languages. In addition, understanding hate speech, particularly its subtle forms, requires a deep examination of its content within social, historical, and cultural contexts, which are frequently overlooked in current detection approaches. In this talk, I will discuss how we address these challenges within the kNOwHATE project (https://knowhate.eu/), employing a comprehensive, culturally sensitive approach grounded in social psychology and language sciences. he talk will also cover key findings from the project and discuss future directions.


Headshot2
Aurélie Névéol
LISN, Université Paris-Saclay

Aurélie Névéol is a Senior Researcher at the Centre National pour la Recherche Scientifique (CNRS). She received a PhD in Computer Science in 2005 and spent ten years at the National Library of Medicine working on information extraction from biomedical documents. She now heads the Language Science and Technologies department at LISN within the Université Paris-Saclay campus. She leads research on clinical natural language processing for languages other than English. Her research interests include information extraction, knowledge representation in specialized domains and ethics in NLP. She has co-chaired the evaluation task for the BLOOM language model. She has also contributed to the organization of the CLEF eHealth shared tasks (2015-2018) for the evaluation of NLP systems.

Title: Evaluation in the era of Large Language Models

11th September @ 9:00, MACI Auditorium (Room 1)

Abstract:

Large Language models have brought about a paradigm shift in Natural Language Processing by creating tools that generate natural language texts with unprecedented fluency in well resourced languages. Furthermore, language models are becoming ubiquitous and easily accessible to the general public. In this talk I will discuss how this impacts the way evaluations are conducted in theory and in practice. I will explain how the nature of large language models can make it difficult to efficiently deploy classic evaluation practices. I will also outline critical evaluation dimensions beyond task performance and will point out the urgency to include social and environmental impact aspects to conduct comprehensive evaluations in the field of Natural Language Processing and Information Retrieval.