Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SentiFuse: Deep Multi-model Fusion Framework for Robust Sentiment Extraction

Hieu Minh Duong, Rupa Ghosh, Cong Hoan Nguyen, Eugene Levin, Todd Gary, Long Nguyen

TL;DR

SentiFuse tackles the challenge of integrating heterogeneous sentiment models by introducing a standardization layer that unifies outputs into probabilistic class distributions and three fusion strategies: decision-level, feature-level, and adaptive fusion. The framework is evaluated on three large social-media datasets, showing systematic fusion consistently surpasses individual models and naive ensembles, with feature-level fusion delivering the strongest overall gains. Adaptive fusion enhances robustness in complex linguistic scenarios such as negation and mixed emotions, while decision fusion provides stable performance across diverse inputs. Overall, SentiFuse proves generalizable across model pools and can complement large language models, offering a practical, model-agnostic approach to robust sentiment extraction with modest additional inference cost.

Abstract

Sentiment analysis models exhibit complementary strengths, yet existing approaches lack a unified framework for effective integration. We present SentiFuse, a flexible and model-agnostic framework that integrates heterogeneous sentiment models through a standardization layer and multiple fusion strategies. Our approach supports decision-level fusion, feature-level fusion, and adaptive fusion, enabling systematic combination of diverse models. We conduct experiments on three large-scale social-media datasets: Crowdflower, GoEmotions, and Sentiment140. These experiments show that SentiFuse consistently outperforms individual models and naive ensembles. Feature-level fusion achieves the strongest overall effectiveness, yielding up to 4\% absolute improvement in F1 score over the best individual model and simple averaging, while adaptive fusion enhances robustness on challenging cases such as negation, mixed emotions, and complex sentiment expressions. These results demonstrate that systematically leveraging model complementarity yields more accurate and reliable sentiment analysis across diverse datasets and text types.

SentiFuse: Deep Multi-model Fusion Framework for Robust Sentiment Extraction

TL;DR

SentiFuse tackles the challenge of integrating heterogeneous sentiment models by introducing a standardization layer that unifies outputs into probabilistic class distributions and three fusion strategies: decision-level, feature-level, and adaptive fusion. The framework is evaluated on three large social-media datasets, showing systematic fusion consistently surpasses individual models and naive ensembles, with feature-level fusion delivering the strongest overall gains. Adaptive fusion enhances robustness in complex linguistic scenarios such as negation and mixed emotions, while decision fusion provides stable performance across diverse inputs. Overall, SentiFuse proves generalizable across model pools and can complement large language models, offering a practical, model-agnostic approach to robust sentiment extraction with modest additional inference cost.

Abstract

Sentiment analysis models exhibit complementary strengths, yet existing approaches lack a unified framework for effective integration. We present SentiFuse, a flexible and model-agnostic framework that integrates heterogeneous sentiment models through a standardization layer and multiple fusion strategies. Our approach supports decision-level fusion, feature-level fusion, and adaptive fusion, enabling systematic combination of diverse models. We conduct experiments on three large-scale social-media datasets: Crowdflower, GoEmotions, and Sentiment140. These experiments show that SentiFuse consistently outperforms individual models and naive ensembles. Feature-level fusion achieves the strongest overall effectiveness, yielding up to 4\% absolute improvement in F1 score over the best individual model and simple averaging, while adaptive fusion enhances robustness on challenging cases such as negation, mixed emotions, and complex sentiment expressions. These results demonstrate that systematically leveraging model complementarity yields more accurate and reliable sentiment analysis across diverse datasets and text types.
Paper Structure (16 sections, 5 equations, 3 figures, 3 tables)

This paper contains 16 sections, 5 equations, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Multi-model Sentiment Analysis.
  5. Standardization Layer.
  6. Fusion Strategies.
  7. Sentiment Classification.
  8. Training and Adaptation.
  9. Experiments and Evaluation
  10. Experiment Setup
  11. Datasets.
  12. Baseline models.
  13. Research Question 1
  14. Research Question 2
  15. Research Question 3
  16. ...and 1 more sections

Figures (3)

  • Figure 1: Overall architecture of the proposed SentiFuse framework.
  • Figure 2: PR curves and ROC curves on the Sentiment140 dataset.
  • Figure 3: Accuracy of proposed framework and best individual model across datasets.