Mining Social Media Text: Extracting Knowledge from Facebook

Abstract

Social media websites allow users to communicate with each other through several tools like chats, discussion forums, comments etc. This results in learning and sharing of important information among the users. The nature of information on such social networking websites can be straight forward categorized as unstructured and fuzzy. In regular day-to-day discussions, spellings, grammar and sentence structure are usually neglected. This may prompt various sorts of ambiguities, for example, lexical, syntactic, and semantic, which makes it difficult to analyse and extract data patterns from such datasets. This study aims at analyzing textual data from Facebook and attempts to find interesting knowledge from such data and represent it in different forms. 33815 posts from 16 news channels pages over Facebook were extracted and analyzed. Different text mining techniques were applied on the collected data. Findings indicated that Fox news is the most news channel that share posts on Facebook, followed by CNN and ABC News respectively. Results revealed that the most frequent linked words are focused on the USA elections. Moreover, results revealed that most of the people are highly interested in sharing the news of Mohammed Ali Clay through all the news channels. Other implications and future perspectives are presented within the study.