Computational Linguistics | Frame Semantics | Events & Perspectives
I was kindly invited by TW!ST, the Groningen linguistics student club, to speak at their career event. I gave a presentation about becoming a Computational Linguistics PhD student, as well as a workshop about "CheckList: do computers *really* understand human language?". [If anyone's interested, my slides are here and here, and my workshop's Colab notebook is here.]
My paper for the 14th International Conference on Computational Semantics was accepted!
Gosse Minnema & Malvina Nissim (2021): "Breeding Fillmore's Chickens and Hatching the Eggs: Recombining Frames and Roles in Frame-Semantic Parsing"
Abstract: Frame-semantic parsers traditionally predict predicates, frames, and semantic roles in a fixed order. This paper explores the ‘chicken-or- egg’ problem of interdependencies between these components theoretically and practically. We introduce a flexible BERT-based sequence labeling architecture that allows for predicting frames and roles independently from each other or combining them in several ways. Our results show that our setups can approximate more complex traditional models’ performance, while allowing for a clearer view of the interdependencies between the pipeline’s components, and of how frame and role prediction models make different use of BERT’s layers.
You can now find all of my research output and other academic activities (such as presentations and conference posters) in the University of Groningen's Pure database, as well as visualize my collaboration graph. My Pure profile is here.
My fellow PhD student on the Dutch FrameNet Project, Levi Remijnse, and me were invited to write a short article for the magazine of NOTaS, the Netherlands Association for Language and Speech Technology. It was a nice way of bringing our work to a wider audience (the magazine is aimed at people working in the language technology industry as well as academia) and a fun experience writing in Dutch for a change!
Minnema, G. & Remijnse, L. (2020). "Framing, betekenis en pragmatiek" (trans.: "Framing, meaning, and pragmatics"). DIXIT 17, p. 14-16
Contribution to the Groningen entry in the WMT Shared Task on Inuktitut Machine Translation. Inuktitut is an Inuit language spoken in the Canadian arctic; it is very low in NLP resources and it's also a very cool (pun intended) language with a very different grammatical structure from European languages. My main contribution was to help with discovering and scraping additional training data for our system.
Roest, C., Edman, L., Minnema, G.F., Kelly, K., Spenader, J. & Toral, A. (2020). "Machine Translation for English–Inuktitut with Segmentation, Data Acquisition and Pre-Training". Proceedings of WMT 2020, p. 274-281
This was the first time I had the chance to meet the international FrameNet community (unfortunately, through Zoom instead of in Marseille!), and to present the first major paper of my PhD. In this paper, Levi Remijnse and discuss the problem of FrameNet annotation from the perspective of the Dutch FrameNet project: how can we make sure our annotations cover all mentions of a given real-world event, and shed a light on how the event is conceptualized and framed?
Remijnse, L. & Minnema, G.F. (2020). "Towards Reference-Aware FrameNet Annotation". Proceedings of IFNW 2020, p. 13-22