<< To the FIN-CLARIAH Meeting event page
9.–10.11.2026
This FIN-CLARIAH network event Small Languages, Large Models explores current challenges in developing AI for under-resourced languages. During the afternoon, we will have expert talks that focus on how data can be collected for smaller languages and how hallucinations can be detected in language models trained on limited resources. Following the talks and discussion, the event continues with a gathering celebrating the 30-year journey of the Language Bank of Finland.
On Tuesday, the event continues as part of the ASTIN Conference: Language Technologies and AI in the Nordic Countries 2026, where representatives from across the Nordic countries share their insights. Join us for a discussion on the future of language technology and AI beyond the world’s largest languages.
Monday 9.11.2026 | University of Helsinki, Athena Building (Siltavuorenpenger 3 A), room 302
Krister Lindén / Vice Dean for Research / Faculty of Humanities, University of Helsinki
Abstract: As the Language Bank of Finland (Kielipankki) celebrates its 30th anniversary (1996–2026), the digital humanities landscape faces a profound paradigm shift driven by the rise of generative AI and Large Language Models (LLMs). While these commercial models offer unprecedented computational scale, they present an existential risk of cultural flattening and domain loss for statistically lower-resource languages like Finnish and Finland-Swedish.
This presentation traces the evolutionary journey of Kielipankki to its current position as a high-performance cornerstone of the FIN-CLARIAH infrastructure. We argue that in the modern AI paradigm, three decades of rigorous academic curation are no longer just historical assets; they are strategic necessities. High-quality, legally cleared, and deeply annotated national corpora serve as the vital baseline required to train, fine-tune, and benchmark localized AI models.
Pirjo Kristiina Virtanen, University of Helsinki
Abstract: Research materials can also be considered subjects that act within relations. This paper discusses Indigenous Data Sovereignty, which emphasizes the rights of Indigenous peoples to govern data about their communities, lands, cultures, and knowledge systems as an integral dimension of their broader political and cultural sovereignty. Who owns the data and how can both ethical and critical research combined? Recently established CARE Principles for Indigenous Data Governance (Collective Benefit, Authority to Control, Responsibility, and Ethics) sets new implications for research practices, data management, and institutional policies. Who has responsibilities in relation to planning, collection, classification, access, use, and archiving of research materials? Indigenous knowledge production remains essential for the continuity of Indigenous knowledge systems yet frequently remain unrecognized within academic data infrastructures, especially what comes to its relational and immaterial dimensions. The paper invites also to rethink restrictions with the guidance of Traditional Knowledge Labels (TK Labels) that can be used in archiving data and thinking collective Indigenous rights over knowledge and data.
Jörg Tiedemann, University of Helsinki
Abstract: AI-tools based on generative language models have taken the World by storm. However, this development comes with an old challenge in NLP that does not go away: Data distributions are heavily skewed and many languages are left behind. The need for data collection in under-resourced languages (i.e. non-English languages) has been there for decades but finding solutions is even more crucial now with the heavy societal impact of AI-applications entering the market. Reducing linguistic discrimination requires re-thinking data collection and distribution. In this talk I will discuss on-going efforts in closing the data gap and discuss shortcomings and problems we face. I will argue for a new strategy in opening archival data to open and transparent AI development as the only chance to make a significant impact. Academic research should be on the forefront of that development and, especially Europe could and should play a major role on this.
Timothee Mickus, University of Helsinki
Abstract: With the ever-growing presence of LLMs in our field and our society, we have also all grown accustomed to their error patterns, and in particular ”hallucinations”: outputs that are completely fluent, sound plausible, but are ultimately incorrect. Despite its prominence, the concept of a hallucination remains rather ill-defined: not only does it assume a clear, objective way to tease apart incorrect text from correct text, it also conflates distinct causes that can lead to erroneous productions. To illustrate this point, my talk will focus on results from the SHROOM series of shared tasks organized by Helsinki. SHROOM has garnered the interest of more than 100 distinct research teams across the years, all trying to develop better means of detecting when these errors occur across languages ranging from English to Malayalam.
Under the theme Infrastructures, the FIN-CLARIAH network and collaborators from the University of Helsinki will showcase research infrastructures, services, laboratories, and projects that support research in the field of social sciences and humanities.
Date and time: 9.11.2026 at 16:30-17:30
Location of the posters: Foyer (near the room Christina) in the University of Helsinki Main Building (the ”old side”, entrance from Unioninkatu 34, 2nd floor)
Services of the University of Helsinki
FIN-CLARIAH projects and resources
After the poster session, we will have a reception organized by the University of Helsinki at 17:30-19:00 in the room Christina. We warmly welcome the participants of the ASTIN Conference and FIN-CLARIAH Meeting to join the event!
On Tuesday, November 10, the FIN-CLARIAH Meeting continues as part of the ASTIN Conference at Clarion Hotel Helsinki.
The FIN-CLARIAH Meeting: Small Languages, Large Models is organized by the University of Helsinki, Department of Digital Humanities.
The ASTIN Conference: Language Technologies and AI in the Nordic Countries 2026 is organized by the University of Helsinki and ASTIN (the Working Group for Language Technology in the Nordic Region). It aims to include all Nordic languages and language communities in the ongoing efforts in language technology in the Nordic countries. The presentations will be held in Scandinavian languages and English.
Contact the organizers for further details:
University of Helsinki
Krister Lindén (krister.linden(at)helsinki.fi)
Wilhelmina Dyster (wilhelmina.dyster(at)helsinki.fi)
Back to top
<< FIN-CLARIAH Overview
Last modified on 2026-11-05
