Co-located with LREC-COLING 2024
- When: Saturday, May 25th, 2024
- Where: Torino, Italy (co-located with LREC-COLING 2024)
- Deadline for submissions: February 29th, 2024
- Paper types: regular papers (long and short) and extended abstracts
- Submission link
- Shared data
- Deadline for camera ready papers: March 29th, 2024 (extension of original deadline on March 20th)
- Contact: pia.sommerauer@vu.nl
9:00 - 9:05 walk-in
9:05 - 09:10 opening remarks
9:10 - 10:10 keynote 1 (Vered Shwartz)
10:10-10:30: Tracking Perspectives on Event Participants: a Structural Analysis of the Framing of Real-World Events in Co-Referential Corpora Levi Remijnse, Pia Sommerauer, Antske Fokkens and Piek T.J.M. Vossen
10:30 - 11:00: coffee break
11:00-11:20: TimeFrame: Querying and Visualizing Event Semantic Frames in Time Davide Lamorte, Marco Rovera, Alfio Ferrara and Sara Tonelli
11:20-11:50: Comparing News Framing of Migration Crises using Zero-Shot Classification Nikola Ivacˇicˇ, Matthew Purver, Fabienne Lind, Senja Pollak, Hajo Boomgaarden and Veronika Bajt
11:50-12:20: Manosphrames: exploring an Italian incel community through the lens of NLP and Frame Semantics Sara Gemelli and Gosse Minnema
12:20-12:40: Broadening the coverage of computational representations of metaphor through Dynamic Metaphor Theory Xiaojuan Tan and Jelke Bloem
12:40-14:00: lunch
14:15-15:15: keynote 2 (Maria Antoniak)
15:15-16:30: discussion/data session
Title: tba
Title: It's not what you said, it's how you said it: Reference, Framing, and Perspective
Abstract:
Lexical variability, i.e. the ability to express the same meaning in various surface forms, poses multiple challenges for NLP applications, even in the age of LLMs. First, models need to respond consistently to queries regardless of their phrasing, which they still struggle to do. Second, resolving co-references to the same real-world events and entities is not trivial when it requires reading between the lines and making inferences. Third, people from diverse cultural backgrounds may differ in their lexical choices; for example, describing events with a focus on either the agent or the contextual factors; and their interpretation of references, such as which hours count as “morning” and which shades count as “blue”. LLMs, which are predominantly trained on English text from US-based users, “understand” the world through a narrow Western or North American lens. Finally, speakers often choose wording that serves their agenda or signals their belonging to a certain group. People’s lexical choice and differences in framing can be used to identify their opinions even when they are not explicit. LLMs’ lexical choice is increasingly controlled by the developers to avoid generating harmful content, which may eventually change the way we speak.
Vered Shwartz is an Assistant Professor of Computer Science at the University of British Columbia, and a CIFAR AI Chair at the Vector Institute. Her research interests include commonsense reasoning, multimodal models, computational semantics and pragmatics, and multiword expressions. Previously, Vered was a postdoctoral researcher at the Allen Institute for AI (AI2) and the University of Washington, and received her PhD in Computer Science from Bar-Ilan University.
When something happens in the world, we have access to an unlimited range of ways (from lexical choices to specific syntactic structures) to refer to the same real-world event. We can chose to express information explicitly or imply it. Variations in reference may convey radically different perspectives. This process of making reference to something by adopting a specific perspective is also known as framing. Although previous work in this area is present (see Ali and Hassan (2022)’s survey for an overview), there is a lack of a unitary framework and only few targeted datasets (Chen et al., 2019) and tools based on Large Language Models exist (Minnema et al., 2022). In this workshop, we propose to adopt Frame Semantics (Fillmore, 1968, 1985, 2006) as a unifying theoretical framework and analysis method to understand the choices made in linguistic references to events. The semantic frames (expressed by predicates and roles) we choose give rise to our understanding, or framing, of an event. We aim to bring together different research communities interested in lexical and syntactic variation, referential grounding, frame semantics, and perspectives. We believe that there is significant overlap within the goals and interests of these communities, but not necessarily the common ground to enable collaborative work.
One way to study variation in framing is to conduct contrastive analyses of texts reporting on the same real-world event. Such an analysis can help to reveal the extent of variation in framing and possibly give rise to the underlying factors that lead to different choices in framing the same event. We collected such a corpus about the Eurovision Song Festival and make it available as a Shared Dataset for the Workshop. The purpose of this corpus is to enable exploratory analyses, facilitate discussion among participants, and, last but not least, make our workshop a real working workshop.
The corpus is composed of news articles reporting on the Eurovision Song Contest that took place in Rotterdam, the Netherlands (canceled in 2020 and held in 2021). The news articles have been collected using the structured data-to-text approach (Vossen et al., 2018). The corpus contains news articles in multiple languages. We invite participants to submit short and targeted analyses using the data (extended abstracts to be discussed in a hands-on data session). Participants are also free to use the data in regular contributions.
Read more about the shared data.
Regular contributions: We aim to lay the groundwork for such efforts. We invite contributions (regular long papers of 8 pages or short papers of 4 pages) targeting any of the following - non-exhaustive - list of topics:
- Theoretical models of framing and perspective
- Annotation frameworks for framing and perspectives
- Computational models of framing and perspective
- Approaches for creating and analyzing referentially grounded datasets (containing different perspectives, written at different points in time, written in different languages)
- Approaches for and analyses of texts about contested and divisive events triggering different opinions and perspectives
- Analyses of and methods for analyzing (diachronic) lexical variation and framing
- Language resources for reference, frames, and perspectives
- Approaches and tools to compare claims of sources
- Frames as expressions of bias in the representation of social groups
- User interface for the visualization of multiple perspectives
COLING-LREC policy for submissions:
When submitting a paper from the START page, authors will be asked to provide essential information about resources (in a broad sense, i.e. also technologies, standards, evaluation kits, etc.) that have been used for the work described in the paper or are a new result of your research. Moreover, ELRA encourages all LREC-COLING authors to share the described LRs (data, tools, services, etc.) to enable their reuse and replicability of experiments (including evaluation ones).
We invite extended abstracts (1,500 words maximum) about small analyses or experiments conducted on our Shared Data. The abstracts will be non-archival and discussed in a dedicated data session.
Please submit your contribution via this link: Submission link
You can select the appropriate paper category (long paper, short paper, extended abstract)
Pia Sommerauer, Tommaso Caselli, Malvina Nissim, Levi Remijnse, Piek Vossen
Özge Alacam Jelke Bloem Oliver Czulo Loic De Langhe Mark Finlayson Antske Fokkens Fritz Günther Sanne Hoecken Eva Huber Allison Lahnala Lucy Li Gosse Minnema Roser Morante Seyed Mahed Mousavi Sebastian Padó Massimo Poesio Marco Rovera Thomas Schmidt Sara Tonelli Tiago Torrent Stella Verkijk Leonie Weissweiler Philipp Wicke Alexander Willich Alexander Ziem
Mohammad Ali and Naeemul Hassan. 2022. A survey of computational framing analysis approaches. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 9335–9348, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics.
Sihao Chen, Daniel Khashabi, Wenpeng Yin, Chris Callison-Burch, and Dan Roth. 2019. Seeing things from a different angle: discovering diverse perspectives about claims. In Proceedings of the 2019 Con- ference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Pa- pers), pages 542–557, Minneapolis, Minnesota. As- sociation for Computational Linguistics.
Gosse Minnema, Sara Gemelli, Chiara Zanchi, Tom- maso Caselli, and Malvina Nissim. 2022. SocioFillmore: A tool for discovering perspectives. In Proceedings of the 60th Annual Meeting of the Associa- tion for Computational Linguistics: System Demon- strations, pages 240–250, Dublin, Ireland. Associa- tion for Computational Linguistics.
Charles J. Fillmore. 1968. The case for case. In E. Bach and R. T. Harms, editors, Universals in linguistic theory, pages 1–88. Holt, Rhinehart, and Winston, New York.
Charles J. Fillmore. 1985. Frames and the semantics of understanding. Quaderni di semantica, 6(2):222– 254.
Charles J. Fillmore. 2006. Frame semantics. In D. Geeraerts, editor, Cognitive Linguistics: Basic Readings, pages 373–400. De Gruyter Mouton, Berlin, Boston. Originally published in 1982.
Piek Vossen, Filip Ilievski, Marten Postma, and Roxane Segers. 2018. Don’t annotate, but validate: a data-to-text method for capturing event data. In Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018), Miyazaki, Japan. European Language Resources Association (ELRA).