Sequence-to-Sequence Language Models for Character and Emotion Detection in Dream Narratives

Language models can automatically detect characters and emotions in dream narratives, reducing the need for labor-intensive manual annotation. A sequence-to-sequence model trained on the English portion of the DreamBank corpus outperforms a much larger language model using in-context learning, despite having 28 times fewer parameters. The study evaluates how model size, prediction order of characters, and inclusion of proper names and character traits affect performance. The supervised approach effectively handles this complex task, and the model and its annotations are publicly released.

Study at a glance

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.