SCRIBE Code (Semantic Annotation Pipeline)
Year: 2026
Type: software
Abstract: ARENA Deliverable P5_WP1_03: Approach for fine-grained semantic annotation via LLMs Codebase for SCRIBE (Scalable Contextual Representations of Interpretable Brain-grounded Semantics). The repository contains scripts for: (1) generating LLM-based semantic annotations for the Binder et al. (2016) semantic feature dataset, enabling validation against human semantic ratings; (2) constructing and annotating a synthetic contextual dataset designed to manipulate concept–property associations and evaluate context-dependent semantic changes; and (3) performing brain encoding analyses on the Moth Radio Hour fMRI dataset, including the generation of contextual semantic annotations and evaluation of their ability to predict neural responses during natural language comprehension.
Authors: Kolling, Camila, Çelik, Emin
Institution: Max Planck Institute for Software Systems
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7165367201