Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MLD-EA: Check and Complete Narrative Coherence by Introducing Emotions and Actions

Jinming Zhang, Yunfei Long

TL;DR

This work addresses the challenge of narrative logical coherence by introducing MLD-EA, a framework that uses actions and emotions to detect and fill gaps in stories. By grounding the model in cognitive-behavioral theory and leveraging prompting and fine-tuning of LLMs, MLD-EA performs action abstraction, emotion classification, and a narrative logic check to identify missing sentence indices and generate coherent missing sentences. Empirical results on Story Commonsense show that incorporating actions and emotions improves logic checking accuracy and generation quality (BLEU, ROUGE, and BERTScore), as well as emotional alignment measured by VAD. The approach demonstrates the potential of LLMs as both logic checkers and story generators, with significant implications for more reliable and emotionally coherent AI-written narratives.

Abstract

Narrative understanding and story generation are critical challenges in natural language processing (NLP), with much of the existing research focused on summarization and question-answering tasks. While previous studies have explored predicting plot endings and generating extended narratives, they often neglect the logical coherence within stories, leaving a significant gap in the field. To address this, we introduce the Missing Logic Detector by Emotion and Action (MLD-EA) model, which leverages large language models (LLMs) to identify narrative gaps and generate coherent sentences that integrate seamlessly with the story's emotional and logical flow. The experimental results demonstrate that the MLD-EA model enhances narrative understanding and story generation, highlighting LLMs' potential as effective logic checkers in story writing with logical coherence and emotional consistency. This work fills a gap in NLP research and advances border goals of creating more sophisticated and reliable story-generation systems.

MLD-EA: Check and Complete Narrative Coherence by Introducing Emotions and Actions

TL;DR

This work addresses the challenge of narrative logical coherence by introducing MLD-EA, a framework that uses actions and emotions to detect and fill gaps in stories. By grounding the model in cognitive-behavioral theory and leveraging prompting and fine-tuning of LLMs, MLD-EA performs action abstraction, emotion classification, and a narrative logic check to identify missing sentence indices and generate coherent missing sentences. Empirical results on Story Commonsense show that incorporating actions and emotions improves logic checking accuracy and generation quality (BLEU, ROUGE, and BERTScore), as well as emotional alignment measured by VAD. The approach demonstrates the potential of LLMs as both logic checkers and story generators, with significant implications for more reliable and emotionally coherent AI-written narratives.

Abstract

Narrative understanding and story generation are critical challenges in natural language processing (NLP), with much of the existing research focused on summarization and question-answering tasks. While previous studies have explored predicting plot endings and generating extended narratives, they often neglect the logical coherence within stories, leaving a significant gap in the field. To address this, we introduce the Missing Logic Detector by Emotion and Action (MLD-EA) model, which leverages large language models (LLMs) to identify narrative gaps and generate coherent sentences that integrate seamlessly with the story's emotional and logical flow. The experimental results demonstrate that the MLD-EA model enhances narrative understanding and story generation, highlighting LLMs' potential as effective logic checkers in story writing with logical coherence and emotional consistency. This work fills a gap in NLP research and advances border goals of creating more sophisticated and reliable story-generation systems.

Paper Structure

This paper contains 25 sections, 7 equations, 3 figures, 15 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Problem Definition
  4. Methodology
  5. Action Abstraction
  6. Emotion Classification
  7. Narrative Logic Checker By Characters' Emotion and Action
  8. Action/ Emotion prediction and sentence generation
  9. Experiment
  10. Data
  11. Selected Baselines
  12. Implement Setups
  13. Evaluation Metrics
  14. Results and Analysis
  15. Action Abstraction and Emotion Classification
  16. ...and 10 more sections

Figures (3)

  • Figure 1: A task example. "Identify logical coherence with actions and emotions" is checking the logical coherence guided by the cognitive-behavioral theory.
  • Figure 2: MLD-EA model overview. Each Input Story contains $n$ sentences and $m$ characters, which have a missing sentence $s_k$ before index $k$. $e(c,s)$ and $a(c,s)$ denote the character's emotion and action in the sentence, respectively; $\hat{e}$ and $\hat{a}$ denotes the predicted emotion and action.
  • Figure 3: Emotion classification. This figure illustrates the relationship between the number of instances for each emotion class and the corresponding classification accuracy. Classes with more instances, such as 'joy,' exhibit higher classification accuracy compared to less frequent classes like 'disgust' and 'trust,' reflecting the potential influence of data imbalance on performance.