ABSTRACT

Social media such as Social network systems, blogs, and microblogs is the most popular services. Using social media, users who are members of a social media community post and exchange information that is related to personal behavior, experimentation, and their own sentiments. These information is not written in ordinary web pages, but it is sometimes important information for the users. However, it is difficult to extract such important information from social media, because so much information exists. We propose a method to extract such important information from social media. We call such information “tip information”. Our proposed tip information is including a user’s experiment and it has common important words. We call the common important words “tip keywords”. We first classify social media based on topic by using Latent Dirichlet Allocation (LDA). Next, we extract user’s experiment sentences from social media based on experience mining method and we extract sentences which are include tip keyword from them. They become tip information. After extracting tip information, we classify it according to four categories which are “suggestion and recommendation”, “restraint”, “briefing”, and “possible and impossible”. Then we present tip information based on each category.