The Digital Humanities Virtual Discussion Group for Early Career Researchers in Belgium is a space for students and researchers to come together to discuss current research in DH as part of an online discussion group. The series is geared toward early career researchers—students and researchers working in Belgian institutions who are interested in DH. For more information about the initiative, you can visit our website:
https://bib.kuleuven.be/english/artes/artesresearch/dh-virtual-discussion-group The spring schedule will be updated as details about upcoming talks are confirmed. Please check back here or on the website (linked above) for full details. Information about each session will also be circulated via the mailing list. Session 1:Date: Monday 25 March, 15h00-16h30 CET via TeamsPresenter(s): Nooshin Shahidzadeh Asadi, University of Antwerp
Title: “Developing Interoperable Open-Source Tools in the Digital Humanities: Merits and Complexities”
Abstract: For this session, Nooshin Asadi (PhD student, the University of Antwerp) will discuss her pivot from software engineering to pursuing a PhD in digital humanities and the changes it has brought about in her professional and academic life. She will talk about her experience with the occasionally harsh but ultimately rewarding reality of producing prototypical, open-source software in the world of DH, while examining in detail the current project she is working on (Axolotl, a real-time collaborative XML editor) and the challenges and opportunities she has faced throughout its development.
Session 2:Date: Monday 29 April, 15h00-16h30 CEST via TeamsPresenter(s): Kushal Jayesh Tatariya, KU Leuven
Title: "Sociolinguistically Informed Interpretability: A Case Study in Hinglish Emotion Classification"
Abstract: In this presentation, I will talk about my project that approaches the interpretability of language models in Natural Language Processing (NLP) from a sociolinguistic perspective. We apply this perspective to the task of emotion classification for code-mixed data. Studies have shown that Hinglish speakers switch to Hindi when expressing negative emotions and to English when expressing positive emotions. To understand if language models can learn these associations, we study the effect of language on emotion prediction across 3 pre-trained language models (PLMs) on a Hinglish emotion classification dataset. I will go into the details of our process of annotating the dataset, the tools we used for our interpretability analysis, and the challenges of working with such subjective data in a setting that demands objectivity. Through this process, we are able to conclude that language models do learn associations between language choice and emotional expression, and this learning can have an impact on model prediction. In conclusion, through a discussion about the results of this project, I would like to explore the general implications of leveraging linguistic theories to understand language models in NLP.
Session 3: Date: Monday 27 May 13h30-16h00 Special DH Benelux Session at Hoek 38 in BrusselsPresenter(s): multiple; tbc
**Are you participating at DH Benelux 2024? Get in touch with Leah Budke (email below) to arrange an elevator pitch at this special session!**
Location: MS Teams and Hoek 38 on Monday 27 May
The MS Teams links will be emailed on the day of each session.
This series is jointly organized by Prof. Margherita Fantoli (KU Leuven), Paavo Van der Eecken (University of Antwerp) and Dr. Leah Budke (KU Leuven). Questions can be addressed to: Leah Budke
leah.budke@kuleuven.be, Margherita Fantoli
margherita.fantoli@kuleuven.be, or Paavo Van der Eecken
paavo.vandereecken@uantwerpen.be.
*If you previously registered for this series, you are on our mailing list and will automatically receive the emails regarding our spring 2024 series, so there is no need to register again. Please contact us directly if you would like to be removed from our mailing list.
