Current work on image-based story generation suffers from the fact that the existing image sequence collections do not have coherent plots behind them. We improve visual story generation by producing a new image-grounded dataset, Visual Writing Prompts (VWP). VWP contains almost 2K selected sequences of movie shots, each including 5-10 images. The image sequences are aligned with a total of 12K stories which were collected via crowdsourcing given the image sequences and a set of grounded characters from the corresponding image sequence. Our new image sequence collection and filtering process has allowed us to obtain stories that are more coherent, diverse, and visually grounded compared to previous work. We also propose a character-based story generation model driven by coherence as a strong baseline. Evaluations show that our generated stories are more coherent, visually grounded, and diverse than stories generated with the current state-of-the-art model. Our code, image features, annotations and collected stories are available at https://vwprompt.github.io/.
14 February 2024 in Journal of Language and Politics
The term dogwhistle refers to an expression conveying a message to a subset of an audience which is not perceived by the rest of the group, in addition to a primary meaning directed at the group at large. We follow up on previous work in linguistics and political communication on defining dogwhistles, taking into account how they likely function in real-life political contexts. We consider the utility of dogwhistles in terms of their sensitivity and their specificity, which allows us to consider dogwhistles in terms of an idealized “life cycle”, whose phases we describe in terms of a multi-dimensional utility tradeoff, described in terms of dogwhistle users, the benefit they expect to receive from dogwhistling, and the deniability of controversial dogwhistle meanings. We propose an approach for the longitudinal study of dogwhistles, and describe the first stages of an experiment to characterize dogwhistles in terms of their lexical properties.
Max Boholm, Björn Rönnerstrand, Ellen Breitholtz, Robin Cooper, Elina Lindgren, Gregor Rettenegger, Asad Sayeed. Proceedings of the 5th Workshop on Computational Approaches to Historical Language Change. 2024.
Xudong Hong, Sharid Loáiciga, Asad Sayeed. Proceedings of the BabyLM Challenge at the 27th Conference on Computational Natural Language Learning. 2023.