| Summary: | 碩士 === 國立臺北科技大學 === 資訊工程系研究所 === 103 === A popular topic is defined as a seminal event or activity along with all directly related events and activities. Topic Detection and Tracking (TDT) techniques could automatically capture the topics from large quantity of the words; however, most of the existing TDT techniques were limited to English news dataset, but differences in languages and post styles makes TDT techniques less applicable to posts on social networking sites.
In this paper, we propose a popular topic detection method in social media. With word segmentation to any posts during a period of time, the proposed method used in this paper could achieve their feature vectors and thus cluster similar posts altogether. Key terms with highest scores are chosen to represent the topic of a particular cluster, making it able to detect popular topic during any time, tracking variation of topics along time.
The test data we collected were posts from Gossiping Broad of PTT between Jan. 2015 and June. 2015. The results show that the proposed method could effectively detect popular news and popular topics shared discussed by the public during a time interval, bringing more convenience for journalists and all Internet users who want to focus on only the most discussed topics.
|
|---|
