No perspective, no perception!! Perspective-aware Healthcare Answer Summarization

TL;DR

This work introduces perspective-aware healthcare answer summarization by proposing a new dataset (PUMA) and a controllable model (PLASMA) that generate perspective-specific summaries from CQA threads. PLASMA combines prompt-based control, prefix tuning on Flan-T5, and an energy-based loss to enforce perspective attributes, tone, and anchor-text constraints. On PUMA's 3167 threads, PLASMA outperforms five strong baselines across standard metrics, with ablations confirming the importance of each energy component and prompt element. The dataset and model enable targeted, faithful summaries that reflect diverse user viewpoints, offering practical benefits for users seeking tailored medical information while respecting ethical considerations in medical content.

Abstract

Healthcare Community Question Answering (CQA) forums offer an accessible platform for individuals seeking information on various healthcare-related topics. People find such platforms suitable for self-disclosure, seeking medical opinions, finding simplified explanations for their medical conditions, and answering others' questions. However, answers on these forums are typically diverse and prone to off-topic discussions. It can be challenging for readers to sift through numerous answers and extract meaningful insights, making answer summarization a crucial task for CQA forums. While several efforts have been made to summarize the community answers, most of them are limited to the open domain and overlook the different perspectives offered by these answers. To address this problem, this paper proposes a novel task of perspective-specific answer summarization. We identify various perspectives, within healthcare-related responses and frame a perspective-driven abstractive summary covering all responses. To achieve this, we annotate 3167 CQA threads with 6193 perspective-aware summaries in our PUMA dataset. Further, we propose PLASMA, a prompt-driven controllable summarization model. To encapsulate the perspective-specific conditions, we design an energy-controlled loss function for the optimization. We also leverage the prefix tuner to learn the intricacies of the health-care perspective summarization. Our evaluation against five baselines suggests the superior performance of PLASMA by a margin of 1.5-21% improvement. We supplement our experiments with ablation and qualitative analysis.

Figures (5)

• Figure 1: Class-wise distribution of spans (left) and summaries(right) across health categories.
• Figure 2: The proposed PLASMA model. Given an input, a perspective-conditioned prompt is generated following the prompt template. Subsequently, the prompt is fed to Flan-T5 with prefix tuner to generate the summary. An energy-driven loss function ($\ell_{\text{\em Perspective}}$) is incorporated along with the standard cross-entropy (CE) loss to enforce the perspective attributes in the generated summary.
• Figure 3: Labelling spans and Writing perspective oriented summaries
• Figure 4: Perspective co-occurence across answers
• Figure 5: Perspective co-occurence within answer span