A Dataset and Benchmark for Consumer Healthcare Question Summarization

Area of Science:

• Natural Language Processing
• Health Informatics
• Information Retrieval

Background:

• Consumer health questions online are often lengthy and lack key details, complicating information retrieval.
• Existing summarization datasets do not adequately address the nuances of consumer health queries.
• Effective summarization is crucial for distilling essential information from vast amounts of online health data.

Purpose of the Study:

• To introduce a novel dataset specifically designed for consumer health question summarization.
• To facilitate the development of more accurate and efficient health information summarization systems.
• To provide a benchmark for evaluating summarization models in the consumer health domain.

Main Methods:

• A new dataset, CHQ-Sum, was created, containing 1,507 consumer health questions.
• Each question in the dataset is annotated by domain experts with corresponding summaries.
• The dataset was derived from community question-answering forums to reflect real-world user queries.

Main Results:

• The CHQ-Sum dataset comprises 1,507 expert-annotated consumer health questions and summaries.
• The dataset was benchmarked using several state-of-the-art summarization models.
• Initial benchmarking demonstrated the dataset's utility for evaluating summarization performance.

Conclusions:

• The CHQ-Sum dataset addresses the need for specialized resources in consumer health question summarization.
• This resource is valuable for advancing research in natural language understanding for healthcare.
• The dataset will aid in developing systems that can better understand and summarize patient-related queries.