With the growth and prevalence of social media, many companies have launched business pages on Facebook to engage customers and encourage positive word-of-mouth about their products and services. Meanwhile, there is little research that investigates what users actually post on these pages and how characteristics of this new form of electronic word-of-mouth drive customer engagement. The researchers used Facebook GraphAPI to collect 0.53 million posts from 41 Facebook business pages. To study the contents of these posts, they manually analyzed a sample of 12,000 posts and established seven content categories. They then built large-scale SVM classifiers to predict the content category of unseen posts. To increase the accuracy of prediction, they extracted specialized attributes besides bag-of-words. This classification model achieved over 0.7 f-measures for the majority of categories. This research advances the understanding of consumer-generated word-of-mouth on new social media platforms and testifies to the effectiveness of using large-scale text mining methods in IS research.
Using Large Scale Text Mining Method to Understand User-Generated Posts on Facebook Business Pages