Abstract The growth of social media sites, such as Twitter, which can provide a visual record of the daily interests and concerns of people in the form of their tweets and tweeting behaviors, has led to an increasing demand among enterprise users, to be able to identify those users who are interested in the services and products that these enterprises offer. However, accurately determining whether people who receive information, such as tweets, from enterprise users have a genuine interest in it can be difficult. In this study, a method for extracting feature words and phrases from the past users’ tweets using temporal patterns of sequential pattern evaluation indices and phrase importance evaluation indices is developed. In this method, a variety of the followers interests are first analyzed using the feature words and phrases retweeted by the followers. Next, the temporal patterns of each evaluation index that are created based on the usage frequencies of feature words and phrases obtained from the historical followers’ tweeting behaviors are extracted. An experimental result has shown that this method successfully extracted the sets of words and phrases based on the followers’ tweeting behaviors as the temporal patterns for each evaluation index and the following retailer’s account. These sets of words and phrases lead to understand the variety of the followers’ interests with more clues.
展开▼
