Abstract:Large Language Models (LLMs) have recently displayed their extraordinary capabilities in language understanding. However, how to comprehensively assess the sentiment capabilities of LLMs continues to be a challenge. This paper investigates the ability of LLMs to detect and react to sentiment in text modal. As the integration of LLMs into diverse applications is on the rise, it becomes highly critical to comprehend their sensitivity to emotional tone, as it can influence the user experience and the efficacy of sentiment-driven tasks. We conduct a series of experiments to evaluate the performance of several prominent LLMs in identifying and responding appropriately to sentiments like positive, negative, and neutral emotions. The models' outputs are analyzed across various sentiment benchmarks, and their responses are compared with human evaluations. Our discoveries indicate that although LLMs show a basic sensitivity to sentiment, there are substantial variations in their accuracy and consistency, emphasizing the requirement for further enhancements in their training processes to better capture subtle emotional cues. Take an example in our findings, in some cases, the models might wrongly classify a strongly positive sentiment as neutral, or fail to recognize sarcasm or irony in the text. Such misclassifications highlight the complexity of sentiment analysis and the areas where the models need to be refined. Another aspect is that different LLMs might perform differently on the same set of data, depending on their architecture and training datasets. This variance calls for a more in-depth study of the factors that contribute to the performance differences and how they can be optimized.

Identifying Sentiment Words Using an Optimization Model with L1 Regularization

Identifying Sentiment Words Using an Optimization-based Model Without Seed Words

Automatic Identification and Recognition of Sentiment Words Using an Optimization-Based Model with Propagation

A Mixture Language Model for the Classification of Chinese Online Reviews.

SEntiMoji: An Emoji-Powered Learning Approach for Sentiment Analysis in Software Engineering

Jointly Learning Sentiment, Keyword and Opinion Leader in Social Reviews

Emotion space model for classifying opinions in stock message board

Sentiment Analysis through LLM Negotiations

Microblog Sentiment Analysis with Emoticon Space Model

Do Large Language Models Possess Sensitive to Sentiment?

Emoticon Smoothed Language Models for Twitter Sentiment Analysis

Exploiting Co-occurrence Opinion Words for Semi-supervised Sentiment Classification.

A Training-Optimization-Based Method for Constructing Domain-Specific Sentiment Lexicon

Identification of Sentiment Labels Based on Self-training

Word Embedding Composition for Data Imbalances in Sentiment and Emotion Classification

MORSE: MultimOdal sentiment analysis for Real-life SEttings

Learning Speaker-Independent Multimodal Representation for Sentiment Analysis

Meta Semantic Template for Evaluation of Large Language Models

Sentiment-Bearing New Words Mining: Exploiting Emoticons and Latent Polarities

Dynamic Adaptive Optimization for Effective Sentiment Analysis Fine-Tuning on Large Language Models

Learning Word Ratings for Empathy and Distress from Document-Level User Responses