Abstract
Measuring the quality of health content in online health forums has been a challenging task. The majority of the existing measures are based on nonprofessional evaluations of forum users and may not be reliable. We employed machine learning techniques, text mining methods, and Big Data platforms to construct four measures of textual quality to automatically determine the similarity of a given answer to professional answers. We then used these measures to assess the quality of 66,888 answers posted on Yahoo! Answers Health section. All four measures of textual quality revealed a higher quality for asker-selected best answers indicating that askers, to some extent, have a proper judgment to select the best answers. We also studied the presence of order effects in online health forums. Our results suggest that the textual quality of the first answer positively influences the mean textual quality of the subsequent answers and negatively influences the quantity of the subsequent answers.