SpråkbankenText's research meetings during Autumn 2026, will take place on Thursdays, from 10:15 to 11.30 (occasionally to 12:00), typically in room F314, unless otherwise announced. The meetings are normally open for all members of SpråkbankenText. On some occasions, we invite a broader audience, these public seminars are announced in SpråkbankenText's calendar.
Please update the list of conference deadlines. Moreover, to access previous meeting periods, which remain in this document as commented sections, you will need to log in and navigate through it. Similarly, a list of potential funders is available. Please feel free to edit, update, and expand the list as appropriate.
The schedule for the Autumn of 2026 is as follows (cells highlighted in orange indicate related events that occur outside the meeting’s usual date and/or time):
| Date | Topic | Comments | August |
|---|---|---|---|
| 20 | Abstract: Digitally collected and stored does not simply equate to digitally available. Mass digitization and the steady growth of born-digital material have made heritage collections larger than ever, but size alone does not make them usable. Copyright restrictions, data protection rules, outdated technical systems, and sparse description still limit how researchers can find and work with relevant material — much of it preserved but effectively forgotten. AI-based search does not remove these barriers, but it changes their shape — and has quietly become part of the humanities’ research infrastructure. Drawing on work at KBLab at the National Library of Sweden — including multimodal image search across historical postcards and AI-generated transcription and search across radio collections — this talk examines how AI models are being used to create new entrance points into heritage collections, and what it means to make previously hard-to-find material newly visible. What happens when source selection is guided not only by cataloguers and metadata, but also by models trained to detect patterns and similarity? How do AI-based search systems relate to more traditional forms of description, and what kinds of blind spots or distortions do they introduce? The talk also considers how such techniques can reshape the identity of a collection itself — shifting attention away from what has already been catalogued and named, toward what has previously escaped notice. More broadly, it asks where humanities expertise is most needed: in shaping such tools, in evaluating them, or in helping scholars engage critically with their outputs. |
September | |
| 3 | Abstract: Keystroke logging provides valuable insights by capturing not only the final written product, but also the pauses and revisions underlying text production, offering a glimpse into the cognitive processes behind it. In this talk, I will introduce keystroke logging and the IncluEdS research project, which is currently collecting keystroke-logged texts from learners of Norwegian. I will present some existing studies and our plans for using these process data: (a) to investigate the order in which some grammatical forms are acquired, and (b) to develop machine learning models for automatically estimating CEFR proficiency from keystroke-logged texts, with explanations designed to be pedagogically useful. Finally, I will discuss preliminary work on using keystroke data to distinguish organic typing from transcribed typing, occurring when copying text from external sources, with a focus on generative AI systems. Link to GU-calendar |
||
| 6 | |||
| 7 | Location: J415; Time: 13.15-15.00 Abstract: The field of Automatic Feedback Generation has experienced a surge of interest in the last few years, driven among other factors by the advent of generative models. While this brings exciting perspectives for students and teachers alike, feedback generation – like any blooming field – comes with its fair share of challenges, many of which have yet to be pinpointed, recognised and analysed in the existing literature. Without a clear recognition of these obstacles, perspectives for further improvements will remain limited. In my seminar, I propose to discuss three of the existing issues in feedback generation: the lack of formal definition of feedback (what is even "feedback"?) the lack of efficient and reliable evaluation method (what is even "good" feedback?), and the extreme rarity of data (how does one "make" feedback?). While it is too early to talk about 'solutions', I will share my thoughts and ideas on the matter, and will happily take in different perspectives on these brain-melting yet fascinating questions. |
||
| 10 | |||
| 11 | |
||
| 14 | |||
| 17 | Also, brief reflections and highlights from the summer (conferences, summer schools, workshops, research visits, publications, submissions, and new collaborations or ideas). Funding opportunities: small- to medium-sized projects; travel grants; other not covered by the previous, such as hiring research assistants and support staff; networking and collaboration activities, etc. Be prepared with suggestions and relevants events we should note in the calendar - for even greater convenience, feel free to edit it yourself before the meeting! |
||
| 23 | |||
| 24 | October | ||
| 1 | |||
| 5-6 | |||
| 8 | |||
| 8 | |||
| 12 | |||
| 15 | |||
| 22-23 | Project proposals with Josie Dixon | Time: 9.30-16.30 day 1 and 9.30-12.30 day 2. Location J335. Workshopen hålls på engelska. | |
| 29 | Note: This week coincides with the autumn break for those with school-age children. | The time slot can still be used, but attendance may be lower than usual. | (Note: DK is currently away) | November |
| 5 | Computational News Analysis: Methods and Applications for Slovene. By Senja Pollak. Guest talk (by Ricardo's opponent) |
Time: 10.15-12.00. Room: J442 Abstract. The talk will present natural language processing methods for automated news analysis, with a focus on Slovene. We will start with solutions for typical tasks in the media industry: keyword extraction, named entity recognition, sentiment analysis, and automatic news categorisation. We will then look at deeper questions that interest media and discourse researchers: how language and topics in the news change over time, and how computational methods can reveal framing and bias in news coverage. Bio. Senja Pollak is a researcher at the Jožef Stefan Institute in Slovenia, where she leads a team working on natural language processing, with a focus on supporting interdisciplinary research in the digital humanities and computational social sciences. She has worked extensively on news analysis: she coordinated the H2020 project EMBEDDIA on cross-lingual news analysis, co-led the ROBACOFI project on comment filtering, led the national project CANDAS on multilingual news discourse analysis, and serves as a work package leader in the LLM4DH project. She is a member of ELLIS (ELLIS Unit Slovenia) and is also a work package leader in the ELLIOT project. |
|
| 6 | Ricardo Muñoz Sanchez, defense. Opponent: Senja Pollak (Slovenia) Committee: Mathias Creutz (Finland), Moa Johansson (Chalmers, Sweden), Henrik Gyllstad (Lund university, Sweden), Christina Lindqvist (GU, Sweden; suppleant) |
J222 (public defense) & F310 (committee) 13.00-18.00 |
|
| 12 | Can be claimed | ||
| 12 | HumAI | Time: 15--17, Location: C350 | |
| 19 | NLP4CALL workshop. Invited speaker Prof. Andrea Horbach | J222 | |
| 20 | NLP4CALL workshop Invited speaker Dr. Rianne Conijn | J222 | |
| 23 | Irene Elmerot et al., Postsem. | Location: J236; Time: 13.15-15.00 | |
| 23-25 | the Swedish Language Technology Conference 2026 (SLTC 2026 | Luleå | |
| 26 | Can be claimed | December | |
| 2 | AI på GU: Hur AI används och kan användas i forskning | Time: 13:00–15:00, Torgny Segerstedtsalen, Vasaparken | |
| 3 | Dana et al.: Linking Historical Populations: Data Quality, Representativeness, and Population-Level Inference | ||
| 3 | HumAI workshop | Time: 15--17, Location: C350 | |
| 10 | Time: 10:00, Location: Staff lunchroom/Lunchrummet. The purpose is to freely share and discuss our thoughts about where we should be going as an organisation. This is particularly relevant now when a new application to VR is being prepared and will be submitted in February. | ||
| 16 | SBX Workshop - details TBA | ||
| 17 | Can be claimed | ||
| 24 | Happy Holidays! | Happy Holidays! |