Analyzing Word-Of-Mouth In Social Media: The Impact And Generation Patterns

Online word-of-mouth (WOM) has been recognized as an important information source in influencing consumer purchase decisions in literature. WOM can influence product sales through both awareness effect and persuasive effect. The explosive growth of social media which allows the creation and exchange of user-generated content (including WOM) is changing the way consumers search for information and communicate with each other. Analyzing the impact of WOM and the generation patterns in the context of social media becomes important. Social broadcasting network is an important type of social media, and user generated contents such as tweets in social broadcasting system have become an important source of WOM. In a social broadcasting system, the awareness effect of WOM is strengthened by the rapid information diffusion, and the persuasive effect of WOM is influenced by social interactions. In my dissertation, I try to investigate the impact of WOM from tweets on product sales while controlling for other WOM sources such as product reviews and blogs. My second research question is the generation patterns of WOM. Generation pattern in my study refers to some specific characters of the rating sequence for a given product. Extant research typically examined the relationship between WOM and product sales by using aggregated WOM data, which ignored the dynamics of WOM generation. In fact, WOM of a product changes from time to time. For two products which have the same aggregated ratings and the same aggregated volumes of ratings, information included in the rating sequence might be different. In my dissertation, I want to identify different patterns of WOM generation, and further to investigate the impact of WOM generation pattern on product sales. To achieve these objectives, I collect data of sales, tweets, reviews, and blogs for a book list from the on-sale calendar of Publishers Weekly. This book list includes all adults’ books which published during the period from October 2011 to December 2011. For the methodology, I propose to use data mining tools such as text mining and clustering to extract information in tweets and reviews. Panel data are analysed for the impact of tweets, and time series are used to identify generation patterns.