Sequence-to-Sequence Language Models for Character and Emotion Detection in Dream Narratives
arXiv (Cornell University) March 21, 2024 preprint DOI: 10.48550/arxiv.2403.15486 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Empirical study |
|---|---|
| Population | English dream narratives from the DreamBank corpus |
| Keywords | Narrative Dream Character mathematics Natural language processing Sequence biology Context archaeology Artificial intelligence Language model Task project management Annotation Process computing Sequence learning Consciousness Linguistics Programming language Topic modeling |
| Key points | Supervised language models with fewer parameters outperform a larger model using in-context learning for character and emotion detection in dream narratives. |
Abstract
The study of dreams has been central to understanding human (un)consciousness, cognition, and culture for centuries. Analyzing dreams quantitatively depends on labor-intensive, manual annotation of dream narratives. We automate this process through a natural language sequence-to-sequence generation framework. This paper presents the first study on character and emotion detection in the English portion of the open DreamBank corpus of dream narratives. Our results show that language models can effectively address this complex task. To get insight into prediction performance, we evaluate the impact of model size, prediction order of characters, and the consideration of proper names and character traits. We compare our approach with a large language model using in-context learning. Our supervised models perform better while having 28 times fewer parameters. Our model and its generated annotations are made publicly available.