Life Event Detection using Conversations from Social Media

LUIS G. MOYANO; PAULO R. CAVALIN; PEDRO P. MIRANDA

Recife

Workshop; Brazilian Workshop on Social Network Analysis and Mining; 2015

Congresso da Sociedade Brasileira de Computação

In this work, we explore the possibility to detecting life events from Social Media by means of machine learning classification algorithms. One important difficulty of this kind of detection task is that, typically, Social Media posts are quite short, and there is not much context provided. This lack of context usually implies strong ambiguity leading to poor classification performance. Here, we propose the use of conversations as a means to augment context and improve classification performance. We evaluate single-post vs. conversation classification performance and compare different models for the conversations classifier. Finally, we describe the performance of the different classifiers in a large data set with 20,000 posts.