Two Phase Extraction Method for Multi-label Classification of Real Life Tweets

Recently, many users share their daily events and opinions on Twitter. Some are beneficial and comment on several aspects of a user's real life, i.e., eating, traffic, weather, disasters, and so on. Such posts as "The train is not coming!" are categorized in the "Traffic" aspect and will support users who want to ride the train. Such tweets as "The train is not coming due to heavy rain" are categorized in both the "Traffic" and "Weather" aspects. In this paper, we propose a multi-label method that estimates appropriate aspects against unknown tweets by extending the two phase extraction method. In it, many topics are extracted from a sea of tweets using Latent Dirichlet Allocation (LDA). Associations among many topics and fewer aspects are built using a small set of labeled tweets. Aspect scores for unknown tweets are calculated using the associations among the topics and the aspects based on the extracted terms. Appropriate aspects are labeled for unknown tweets by averaging of the aspect scores. Using a large amount of actual tweets, our sophisticated experimental evaluations demonstrate the high efficiency of our proposed multi-label classification method. When an aspect score is much larger than others, that aspect is estimated against the tweet. When several aspect scores are large within similar values, these aspects are estimated. Based on the experimental evaluation results, our prototype system demonstrates that our proposed method can appropriately estimate some aspects of each unknown tweets.