Span-level Emotion-Cause-Category Triplet Extraction with Instruction Tuning LLMs and Data Augmentation

TL;DR

This paper addresses the fine-grained task of span-level emotion-cause-category triplet extraction, proposing an end-to-end framework that combines instruction tuning of large language models with LoRA and prompt-based data augmentation. By transforming the task into structured instructions and generating high-quality synthetic data, the approach mitigates data scarcity and avoids error propagation common in multi-stage pipelines. Empirical results on the Chinese GuiWXLZ16 dataset show substantial gains over strong baselines in both span-level and word-level metrics, with notable improvements when using InstruDAGLM and data augmentation. The work demonstrates the effectiveness and robustness of instruction-guided LLMs for structured emotion analysis and suggests promising directions for scalable, cross-task structured prediction with automated data generation.

Abstract

Span-level emotion-cause-category triplet extraction represents a novel and complex challenge within emotion cause analysis. This task involves identifying emotion spans, cause spans, and their associated emotion categories within the text to form structured triplets. While prior research has predominantly concentrated on clause-level emotion-cause pair extraction and span-level emotion-cause detection, these methods often confront challenges originating from redundant information retrieval and difficulty in accurately determining emotion categories, particularly when emotions are expressed implicitly or ambiguously. To overcome these challenges, this study explores a fine-grained approach to span-level emotion-cause-category triplet extraction and introduces an innovative framework that leverages instruction tuning and data augmentation techniques based on large language models. The proposed method employs task-specific triplet extraction instructions and utilizes low-rank adaptation to fine-tune large language models, eliminating the necessity for intricate task-specific architectures. Furthermore, a prompt-based data augmentation strategy is developed to address data scarcity by guiding large language models in generating high-quality synthetic training data. Extensive experimental evaluations demonstrate that the proposed approach significantly outperforms existing baseline methods, achieving at least a 12.8% improvement in span-level emotion-cause-category triplet extraction metrics. The results demonstrate the method's effectiveness and robustness, offering a promising avenue for advancing research in emotion cause analysis. The source code is available at https://github.com/zxgnlp/InstruDa-LLM.

Figures (4)

• Figure 1: Three methods for extracting emotion-cause-category triplets.
• Figure 2: Overview of the proposed model, which mainly consists of two parts: (a) data augmentation and (b) instruction tuning. The data augmentation module generates an augmented dataset, which is then used for instruction tuning to predict emotion-cause-category triplets.
• Figure 3: Data augmentation prompt.
• Figure 4: Triplet extraction instruction.

Theorems & Definitions (1)

• Example 1