Life event classification of Facebook posts augmented by their relationships

• Main Author: Te, Robee Khyra Mae J.
• Format: text
• Language: English
• Published: Animo Repository 2019
• Subjects: Facebook (Electronic resource); Ensemble learning (Machine learning); Computer Sciences
• Online Access: https://animorepository.dlsu.edu.ph/etd_masteral/6526; https://animorepository.dlsu.edu.ph/cgi/viewcontent.cgi?article=13522&context=etd_masteral
• Institution: De La Salle University

The use of social media, in particular, Facebook, to share information about ourselves is very common nowadays. Facebook users can easily adapt on how they record or share important happenings in their lives. With Facebook, they post updates to share their daily activities and life events with their friends. In some cases, Facebook users tend to share related events through separate posts producing a dependency between these posts. These posts may contain relationships that could help us in the classification task. Previous work on text-based life event classification focused only on topic and life event classification of independent posts or tweets of social media content. The use of graph-based classification remains unexplored in this particular domain. In this study, graph-based classification technique is used to build a classifier model to classify a Facebook post based on its relationships to previous posts. Results for the graph-based classifier are compared to the test results of the traditional classifier. This shows that the traditional machine learning technique still performed better.