Pipeline for Force Dynamics Annotation in Dream Narratives: Complete Prompt and Implementation Code
Open MIND January 1, 2026 DOI: 10.17605/osf.io/6nmuv (opens in new tab) via OpenAlex
Summary
AI-generated from the abstractThis project provides supplementary materials for a forthcoming article on force dynamics in dream narratives. It includes a system prompt for automated annotation of Force Dynamics patterns based on Talmy's framework, eight few-shot examples from a manually annotated corpus of 25 pandemic-era dream reports, and Python code for the annotation pipeline using the Anthropic API. The annotation scheme identifies five semantic fields, three normativity types, force balance outcomes, and emotional valence per minimal scene. Applied to 524 dreams from DreamBank across 7 series, the pipeline achieved 79% coverage of dreams with identifiable FD patterns and 79.2% prevalence of AGO<ANT configurations. Inter-rater reliability on the primary corpus was κ = 0.82.
Study at a glance
| Characteristics | Peer reviewed |
|---|---|
| Keywords | Annotation Counterfactual thinking Pipeline software Dream Python programming language |
| Key finding | The annotation pipeline achieved 79% coverage of dreams with identifiable Force Dynamics patterns and a 79.2% prevalence of AGO<ANT configurations across 524 dreams from 7 series. |
Abstract
This project contains the supplementary materials for the article "Force dynamics in dream worlds: Adversarial structure and counterfactual simulation in dream narratives" (Fuentes Bravo, 2026, manuscript submitted for publication). The materials include: (1) a complete system prompt for automated annotation of Force Dynamics (FD) patterns in dream narrative texts, based on Leonard Talmy's (2000) theoretical framework; (2) eight few-shot examples calibrated on a manually annotated corpus of 25 dream reports collected during the COVID-19 pandemic (CorpusS21); and (3) Python implementation code for running the annotation pipeline using the Anthropic API. The annotation scheme identifies five semantic fields (Physical, Physical-Psychological, Intrapsychic, Socio-psychological, Ontological), three normativity types (Interior, Exterior, Ontological), force balance outcomes (AGOANT), and emotional valence for each minimal scene extracted from a dream report. The pipeline was used to annotate 524 dreams from the DreamBank corpus (Domhoff & Schneider, 2008) across 7 series, achieving a coverage of 79% of dreams with identifiable FD patterns and a 79.2% prevalence of AGO