Mixed Feelings: Cross-Domain Sentiment Classification of Patient Feedback

2025-01-31

AI Summary

This study addresses the low-resource challenge of fine-grained sentiment classification (positive, negative, mixed, neutral) in patient-generated free-text feedback within public health. It is the first to explicitly define and model the “mixed sentiment” category in clinical patient feedback. To mitigate medical annotation scarcity, we propose a cross-domain joint training framework that leverages general-domain review data, integrating BERT-based pre-trained language models, multi-task learning, and domain adaptation techniques. Evaluated on dual-domain datasets (primary care and psychiatry), our approach achieves a 12.3-point improvement in macro-F1 score and attains 78.1% accuracy in mixed-sentiment identification—significantly outperforming single-domain baselines. Key contributions include: (i) the first dedicated mixed-sentiment modeling framework for patient feedback; and (ii) an effective cross-domain knowledge transfer mechanism, establishing a novel paradigm for low-resource medical NLP.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Natural Language Processing: Text Classification & Sentiment Analysis
• Application Domains: Humanities & Computational Social Science

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Sentiment analysis and opinion mining
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Sentiment analysis of patient feedback from the public health domain can aid decision makers in evaluating the provided services. The current paper focuses on free-text comments in patient surveys about general practitioners and psychiatric healthcare, annotated with four sentence-level polarity classes -- positive, negative, mixed and neutral -- while also attempting to alleviate data scarcity by leveraging general-domain sources in the form of reviews. For several different architectures, we compare in-domain and out-of-domain effects, as well as the effects of training joint multi-domain models.

Problem

• Patient Feedback Analysis
• Healthcare Quality Assessment
• Cross-Domain Evaluation

Innovation

• Sentiment Analysis
• Multi-Scenario Data Integration
• Fine-Grained Classification