Semantic Consistency Regularization with Large Language Models for Semi-supervised Sentiment Analysis

TL;DR

Data scarcity is a major bottleneck for sentiment analysis, which the SCR framework addresses by leveraging pretrained Large Language Models to semantically augment unlabeled text. SCR uses two prompting strategies, SCR-EE and SCR-CE, to produce semantically consistent augmentations and trains with a confidence-thresholded consistency loss, where a threshold $\tau$ governs contribution via $L_{con}$. A class-reassemble mechanism inspired by class-space shrinking reduces the effective class set for uncertain samples, enabling additional supervision through $L_{sh}$. Across FSA and Amazon, SCR achieves state-of-the-art results under multiple labeling regimes, demonstrating improved data efficiency, robustness, and generalization for semi-supervised sentiment analysis.

Abstract

Accurate sentiment analysis of texts is crucial for a variety of applications, such as understanding customer feedback, monitoring market trends, and detecting public sentiment. However, manually annotating large sentiment corpora for supervised learning is labor-intensive and time-consuming. Therefore, it is essential and effective to develop a semi-supervised method for the sentiment analysis task. Although some methods have been proposed for semi-supervised text classification, they rely on the intrinsic information within the unlabeled data and the learning capability of the NLP model, which lack generalization ability to the sentiment analysis scenario and may prone to overfit. Inspired by the ability of pretrained Large Language Models (LLMs) in following instructions and generating coherent text, we propose a Semantic Consistency Regularization with Large Language Models (SCR) framework for semi-supervised sentiment analysis. We introduce two prompting strategies to semantically enhance unlabeled text using LLMs. The first is Entity-based Enhancement (SCR-EE), which involves extracting entities and numerical information, and querying the LLM to reconstruct the textual information. The second is Concept-based Enhancement (SCR-CE), which directly queries the LLM with the original sentence for semantic reconstruction. Subsequently, the LLM-augmented data is utilized for a consistency loss with confidence thresholding, which preserves high-quality agreement samples to provide additional supervision signals during training. Furthermore, to fully utilize the uncertain unlabeled data samples, we propose a class re-assembling strategy inspired by the class space shrinking theorem. Experiments show our method achieves remarkable performance over prior semi-supervised methods.

Figures (5)

• Figure 1: Illustration of the overall framework of the proposed SCR.
• Figure 2: The prompting process of SCR-EE.
• Figure 3: The prompting process of SCR-CE.
• Figure 4: The accuracy trends across different methods during training.
• Figure 5: Visualization of the word cloud with different methods.