Nowadays, social media is one of the essential sharing of information and proliferation tools because it spreads text messages, news, pictures, or videos in real-time. During the disaster, Japanese people use social media to exchange real-time information for their social interaction. Twitter is the most popular tool that has been used for disaster response in Japan. Even though many disaster systems have been created and used for disaster mitigation in Japan, most of them are assumed to be used by the Japanese in the Japanese language. From this problem, this study focuses on the way to create a disaster response system and community service to help, collect, and extract information on social media to help disaster mitigation becomes more important. This paper aims to investigate the tweets by focusing on noun keywords during the Osaka North Earthquake on 18 June 2018 with a data set of more than 9,000,000 tweets. The process presented classify social media messages by using ontology, word similarity, frequency of keyword, and evaluate results of natural language processing. We organize the messages into 15 categories and used as the classification algorithms with machine learning features of the count of each category word in the sentences. The result tweets were statistically compared with the keyword in each category to classify the content and collecting disaster information and using the result to build the analysis system.