| Summary: | 碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 101 === Twitter, one of the most popular microblogging services allow users (twitterer) to post messages (tweet) through the website interface, SMS, or a range of Apps for mobile devices. It has over 140 million active users, generating over 340 million messages per day. As a result of rapidly increasing numbers of tweets, Twitter becomes a valuable source for mining people’s opinion and sentiment. Tradition opinion mining from general contents (Blog articles, review articles) on the web focuses on specific domains such as movie, book or consumer product. They gather lots of articles regarding the particular product, and analysis the opinion in the articles. Different from mining the opinion from particular objects, we want to know whether object A can be found when mining opinions from object B, and conclude the competition between object A and object B.
In this paper, we introduce the competition relation between two objects, and propose the idea of the relative opinion to identify the competition. We construct a framework to identify the competition in Twitter. Our method consists of two stages: First, the data process stage processes the raw data from Twitter and constructs the opinion conversions. The ranking stage mines the relative opinion between objects. This stage also aggregates the opinions to identify the competition.
Our result shows that the relative opinion is a good indicator to identify the competition. The opinion aggregation algorithm we proposed can help us to improve the result of the identification.
|
|---|
