Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster F in Poster Session F: Thursday, August 6, 1:45 – 3:30 pm, Kimmel Center, Shorin & Rosenthal Rooms

Automated recall-to-story matching for memory research

Dhruva Arekar¹, Gabriel Arthur Daniel Kressin Palacios¹, Xian Li¹, Kahlyn Eckles², Viswanath Missula¹, Janice Chen¹; ¹Johns Hopkins University, ²University of Oregon

Presenter: Dhruva Arekar

Research on memory increasingly relies on paradigms using naturalistic stimuli in which participants freely recall stories or films in their own words. For many analyses of such data it is critical to identify which story-elements are being referred to at each time point of a participant’s recall. Currently, this task is typically performed by human annotators, a process which is effortful and susceptible to human error. We present our ongoing work on an LLM-based tool, available as a Python package and CLI, that automates this recall matching. Our tool achieved strong agreement with human annotators on written narratives, though performance on movie narratives was more variable. The tool also remained highly consistent in its ratings across repeated runs. By automating story-to-recall matching, this tool aims to facilitate naturalistic memory research while improving consistency and reproducibility.

Topic Area: Methods, Tools, Theory & Neural Coding