Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

nicolay-r at SemEval-2024 Task 3: Using Flan-T5 for Reasoning Emotion Cause in Conversations with Chain-of-Thought on Emotion States

Nicolay Rusnachenko, Huizhi Liang

TL;DR

This work tackles emotion-cause extraction in conversations by formulating Subtask 1 of ECAC on the Friends dataset and introducing a two-stage instruction-tuning approach based on Three-hop Reasoning (THoR). Stage 1 (THoRstate) predicts the emotion state of the target utterance, while Stage 2 (THORcause) predicts the emotion caused by a source utterance, with an optional THOR-rr revision that injects the inferred source state into reasoning. On Flan-T5-base (250M), THoR-based tuning yields consistent improvements over prompt baselines, and the final THORcause-rr setup with algorithmic spans correction achieves competitive dev/test F1 metrics, demonstrating effective chain-of-thought reasoning for emotion analysis in dialogue. The work also provides an open-source THOR fork to support future research on reasoning-aware emotion understanding in conversations.

Abstract

Emotion expression is one of the essential traits of conversations. It may be self-related or caused by another speaker. The variety of reasons may serve as a source of the further emotion causes: conversation history, speaker's emotional state, etc. Inspired by the most recent advances in Chain-of-Thought, in this work, we exploit the existing three-hop reasoning approach (THOR) to perform large language model instruction-tuning for answering: emotion states (THOR-state), and emotion caused by one speaker to the other (THOR-cause). We equip THOR-cause with the reasoning revision (rr) for devising a reasoning path in fine-tuning. In particular, we rely on the annotated speaker emotion states to revise reasoning path. Our final submission, based on Flan-T5-base (250M) and the rule-based span correction technique, preliminary tuned with THOR-state and fine-tuned with THOR-cause-rr on competition training data, results in 3rd and 4th places (F1-proportional) and 5th place (F1-strict) among 15 participating teams. Our THOR implementation fork is publicly available: https://github.com/nicolay-r/THOR-ECAC

nicolay-r at SemEval-2024 Task 3: Using Flan-T5 for Reasoning Emotion Cause in Conversations with Chain-of-Thought on Emotion States

TL;DR

This work tackles emotion-cause extraction in conversations by formulating Subtask 1 of ECAC on the Friends dataset and introducing a two-stage instruction-tuning approach based on Three-hop Reasoning (THoR). Stage 1 (THoRstate) predicts the emotion state of the target utterance, while Stage 2 (THORcause) predicts the emotion caused by a source utterance, with an optional THOR-rr revision that injects the inferred source state into reasoning. On Flan-T5-base (250M), THoR-based tuning yields consistent improvements over prompt baselines, and the final THORcause-rr setup with algorithmic spans correction achieves competitive dev/test F1 metrics, demonstrating effective chain-of-thought reasoning for emotion analysis in dialogue. The work also provides an open-source THOR fork to support future research on reasoning-aware emotion understanding in conversations.

Abstract

Emotion expression is one of the essential traits of conversations. It may be self-related or caused by another speaker. The variety of reasons may serve as a source of the further emotion causes: conversation history, speaker's emotional state, etc. Inspired by the most recent advances in Chain-of-Thought, in this work, we exploit the existing three-hop reasoning approach (THOR) to perform large language model instruction-tuning for answering: emotion states (THOR-state), and emotion caused by one speaker to the other (THOR-cause). We equip THOR-cause with the reasoning revision (rr) for devising a reasoning path in fine-tuning. In particular, we rely on the annotated speaker emotion states to revise reasoning path. Our final submission, based on Flan-T5-base (250M) and the rule-based span correction technique, preliminary tuned with THOR-state and fine-tuned with THOR-cause-rr on competition training data, results in 3rd and 4th places (F1-proportional) and 5th place (F1-strict) among 15 participating teams. Our THOR implementation fork is publicly available: https://github.com/nicolay-r/THOR-ECAC
Paper Structure (9 sections, 2 figures, 4 tables, 1 algorithm)

This paper contains 9 sections, 2 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Task Overview
  2. Methodology
  3. Chain-of-Thought Prompting
  4. Reasoning Revision with Supervision
  5. Datasets and Experiential Setup
  6. Experiments
  7. Algorithm-based Spans Correction
  8. Final Submission Analysis
  9. Conclusion

Figures (2)

  • Figure 1: Two-stage LLM tuning methodology for inferring emotion caused by $u^{src}$ towards $u^{tgt}$ in context $X$ by adapting THoRFeiAcl23THOR to reason and answer: (i) $u^{tgt}_{state}$ (THoRstate), and (ii) emotion caused by $u^{src}$ towards $u^{tgt}$ (THoRcause), optionally enhanced by Reasoning-Revision and by predicting $u^{src}_{state}$ (THoRcause-rr).
  • Figure 2: Result analysis of the preliminary fine-tuning of Flan-T5base on $D\text{state}$dev using THoRstate technique per epoch by $F1({E'{}})$