Table 2.

Application of text mining in online consultation platform.

Dimensionality	Author	Interaction Mode/Scenario	Research Content
Information	Peng et al. [40]	doctor-to-patient information disclosure	Aims: identify potential topics in doctors’ self-disclosure information, and explore the impact of topic diversity in doctor self-disclosure on patient choice. Methods: LDA topic model; hierarchical clustering method Results: excessive quantity of information and semantic topic diversity can raise barriers for patient’s decision.
	Park et al. [42]	patient-to-patient communication	Aims: examine how different types of supportive messages posted on OHCs encourage users to increase their health resilience. Method: directed content analysis Results: self-efficacy-oriented messages affect helpfulness, while response-efficacy-oriented messages influence the relationships among helpfulness, goal-setting, and health resilience.
Emotion	Lu et al. [43]	patient-to-patient communication	Aims: calculate the emotional representation of depressed patients in texts from an online consultation platform, and further investigate whether the use of online communities helps improve depression. Methods: Baidu AI’s natural language processing method Results: Emotional support positively affect the treatment of depression.
	Liu et al. [37]	patient-to-patient communication	Aims: explore various patterns of information exchange and social support in web-based health care communities and identify factors that affect such patterns. Methods: social network analysis; text mining techniques Results: polarized sentiment increases the chances of users to receive replies, and optimistic users play an important role in providing social support to the entire community.
Information and emotion	Chen et al. [44]	patient-to-patient communication	Aims: consider whether or not linguistic signals in posts (including sentiment valence, linguistic style matching, readability, post length, and spelling) impact the amount of support received. Methods: social support classification using SVM; structured information extraction Results: affective linguistic signals, including negative sentiment and linguistic style matching, are effective in invoking both informational and emotional support from the community.
	Jiang et al. [17]	patient-to-doctor communication	Aims: how various linguistic characteristics of patients’ communication in these communities affect their social support outcomes. Methods: linguistic analysis; exponential random graph models Results: lexical richness in health-related vocabulary negatively correlates with receiving informational support. The readability and brevity of written texts have positive relationships with incoming social support.