KEYNOTE SPEAKERS:
|
Professor Tania Cerquitelli, Department of Control and Computer Engineering, Polytechnic of Turin, Italy
website: https://www.polito.it/en/staff?p=tania.cerquitelli/ |
SPEECH TITLE: "Leveraging AI for Inclusive Writing: A Bias-Free Methodology in NLP"
SUMMARY:
Inclusive language is essential for expressing ideas, drafting laws and regulations, and delivering news in a way that fosters effective communication and ensures everyone feels represented. Unfortunately, the negative impact of non-inclusive writing is often underestimated, as it can unintentionally convey biased, exclusionary, or even offensive messages. To promote inclusive communication, this talk presents an AI-powered tool, Inclusively, which eliminates stereotypes related to gender, ethnicity, disability, and age in written documents by suggesting alternative, inclusive forms. For example, when a non-expert writer drafts formal documents, Inclusively automatically reviews the initial draft and suggests revisions to ensure it aligns with inclusive language standards. This could include replacing gender-specific pronouns with gender-neutral ones or suggesting alternative terms for certain disabilities that are less stigmatizing.
Inclusively functions as both a proofreader and a learning tool, helping writers easily create bias-free, inclusive content. It is designed for use by both general users and experts, encouraging feedback and continuous improvement through a human-in-the-loop approach. In collaboration with linguistic experts, we first established a comprehensive set of linguistic criteria to model inclusive writing in Italian. Based on these guidelines, we curated and annotated a dataset of Italian administrative documents, enriched with detailed inclusivity annotations. We then trained deep learning models on this dataset, implementing a two-step pipeline: (1) a classifier that identifies non-inclusive language and (2) a generative model that reformulates the detected text to be more inclusive. Inclusively also provides explanations of the models' outputs to increase system transparency. Furthermore, it allows expert end-users to provide further annotations for system fine-tuning.
This study was carried out within the project ''E-MIMIC: Empowering Multilingual Inclusive Communication'', funded by the Ministero dell'Università e della Ricerca - with the PRIN 2022 (D.D. 104 - 02/02/2022) program.
ABOUT THE SPEAKER:
Tania Cerquitelli is a Full Professor at the Department of Control and Computer Engineering at the Polytechnic of Turin, Italy. In addition to her academic role, Tania holds responsibilities for aggregate functions under the Deputy Rector for Society, Community, and Program Delivery, where she supports various initiatives for the Polytechnic University Community. She also serves as a member of the university's trade union relations delegation and actively participates in the Gender, Equality, Diversity, and Inclusion (GEDI) Observatory.
Tania's research focuses on several cutting-edge topics within the areas of data science and machine learning, including techniques for explaining black-box models, algorithms designed to democratize data science, and early detection of concept drift. She is a member of the editorial boards of several international journals, including Computer Networks, Future Generation Computer Systems, Expert Systems with Applications, Engineering Applications of Artificial Intelligence (all published by Elsevier), Knowledge and Information Systems (Springer), and IEEE Data Descriptions.
Her involvement in the academic community extends to her role on the steering committee of ECML-PKDD and ADBIS. She has served as Program Co-Chair for ADBIS 2022, Co-Chair for the Journal Track of ECML-PKDD 2023, and Co-Chair for the EDI Special Day at ACM KDD 2024. Additionally, Tania has organized over 15 international workshops on data science and machine learning topics in recent years.
Her research has been supported by funding from various sources, including the European Union, the Piedmont Region, the Ministry of University and Research, and private companies.
|
Professor Rebecca Hwa, Chair Computer Science Department, George Washington University, US
website: https://cs.engineering.gwu.edu/rebecca-hwa |
SPEECH TITLE: "The Impact of Language Models in Persuasive Media"
SUMMARY:
Large-scale pre-trained language models have had significant impact on a wide range of Natural Language Processing applications. However, their properties and characteristics are not well understood. In this talk, I explore different uses of language models for processing persuasive media such as social media posts and advertisement images. I will first examine the relationship between implicit political ideology biases in social media posts and language models and discuss methods to mitigate the impact of the biases. Next, I will examine the applicability of language models as a knowledge source for recognizing symbolism in images and text and discuss its implications.
ABOUT THE SPEAKER:
Professor Rebecca Hwa's research sits at the intersection of natural language process, machine learning, and human computer interaction. Her work focuses on developing machine learning methods that reveal the hidden syntactic and semantic structures within languages. These methods have applications in diverse domains, including health, education, and the social sciences. Some of her recent projects include: modeling student behaviors in revising argumentative essays, identifying symbolisms in multimodal rhetorics, and recognizing group biases in social media.
|
Professor Jamal Hasanov, School of IT and Engineering, ADA University, Azerbaijan
website: https://ada.edu.az/en/schools/site/members/faculty/63-jamaladdin-hasanov |
SPEECH TITLE: "The Importance of Features in Specific Computer Vision Problems"
SUMMARY:
Extracting informative data—features is critical to building robust and accurate computer vision models. This presentation delves into the complexities of selecting meaningful features and examines their impact on model performance. The importance of a domain-specific approach to feature extraction and the evaluation of feature efficiency will be discussed. Special attention will be given to the potential pitfalls of transfer learning, as pre-trained models may transfer irrelevant or misleading features when applied to domain-specific tasks. Case studies from different fields will be analyzed, and key conclusions will be shared. This presentation aims to provide a practical understanding of optimizing feature extraction and selection in computer vision tasks, thereby enhancing model precision and adaptability.
ABOUT THE SPEAKER:
Dr. Jamal Hasanov is an Associate Professor of Computer and Information Sciences at ADA University's School of IT and Engineering. With over 20 years of industry experience, he specializes in teaching courses that bridge academic knowledge with real-world industrial applications in Information Technology. His research focuses on Computer Vision, particularly in video captioning and medical image classification, where he actively contributes to advancements in these fields.