Health-Related Hot Topic Detection in Online Communities Using Text Clustering

Abstract

Recently, health-related social media services, especially online health communities, have rapidly emerged. Patients with various health conditions participate in online health communities to share their experiences and exchange healthcare knowledge. Exploring hot topics in online health communities helps us better understand patients’ needs and interest in health-related knowledge. However, the statistical topic analysis employed in previous studies is becoming impractical for processing the rapidly increasing amount of online data. Automatic topic detection based on document clustering is an alternative approach for extracting health-related hot topics in online communities. In addition to the keyword-based features used in traditional text clustering, we integrate medical domain-specific features to represent the messages posted in online health communities. Three disease discussion boards, including boards devoted to lung cancer, breast cancer and diabetes, from an online health community are used to test the effectiveness of topic detection. Experiment results demonstrate that health-related hot topics primarily include symptoms, examinations, drugs, procedures and complications. Further analysis reveals that there also exist some significant differences among the hot topics discussed on different types of disease discussion boards.

A Qualitative Analysis of “Naturalistic” Conversations in a Peer-Led Online Support Community for Lung Cancer

“…22,51,52 On the other hand, other researchers found that more online participation by support persons occurred when the prognosis for a loved one was poor or when the disease had advanced. 39 As found by Lu et al, 53 who compared topical content volunteered in breast and lung cancer online communities, users in this study also focused on informational support and appraisal of threatening new or long-standing and unrelieved symptoms such as chest pain to boost coping strategies based on advice provided by other experienced users. Information exchange was rare in this online support community and other online communities without empathic support.…”

Web news mining in an evolving framework

“…It arose from the related fields of data mining, artificial intelligence, statistics, databases, library science, and linguistics. As it is detailed in [3], since text mining is a multidisciplinary field, this term has been used to describe different applications such as text categorization [26,27], prediction [28,29], text clustering [30,31], association discovery [32,33] and finding patterns in text databases [34].…”

Use and Taxonomy of Social Media in Cancer-Related Research: A Systematic Review

“…36,56,67,76 Other cancers mentioned were osteosarcoma, 73 rhabdomyosarcoma, 55 testicular, 11 skin, 23 cervical, 37,53,59,65 ovarian, 21, lung, 34 and colorectal. 14,24 Cancer in general was the focus of 21 articles (30.4%),9,15,19,20,31–33,38,40,44,46–48,52,54,57,62–64,70,72 and 11 (15.9%) discussed multiple types of cancer.…”