Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Gustave Cortal

TL;DR

This paper presents the first study on character and emotion detection in the English portion of the open DreamBank corpus of dream narratives, and shows that language models can effectively address this complex task.

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.

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

TL;DR

This paper presents the first study on character and emotion detection in the English portion of the open DreamBank corpus of dream narratives, and shows that language models can effectively address this complex task.

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.
Paper Structure (38 sections, 3 figures, 4 tables)

This paper contains 38 sections, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Theoretical background and related works
  3. Relationship with structured emotion and sentiment analysis
  4. Existing research on automated analysis of dream narratives
  5. Dictionary and lexical database-based approach
  6. Language model-based approach
  7. Methodology
  8. DreamBank dream narrative database
  9. Hall and Van de Castle annotation scheme
  10. Character
  11. Emotion
  12. Description of our approach
  13. Natural language conversion of character and emotion codes
  14. Training
  15. Evaluation
  16. ...and 23 more sections

Figures (3)

  • Figure 1: Sequence-to-sequence approach for automating the coding of dream narratives. Codes describing characters and their emotions are converted into natural language to produce the training data. From a narrative, a language model generates the natural language description of characters and their emotions.
  • Figure 2: Distribution of emotional states.
  • Figure 3: Distribution of status (a), gender (b), identity (c) and age (d) in the dream narratives.