Detecting Topics from Twitter Posts During TV Program Viewing
Takanobu Nakahara, Yukinobu Hamuro · 2013
This research proposes a method to detect the contents of Twitter posts by analyzing the contents of tweets posted by viewers watching a specific TV program whenever the number of posts increase dramatically and then to summarize that content. First the proposed method creates concepts from clusters based on the co-occurrence of words. Then posts during tweet bursts and posts that match the contents of the TV program dialog are taken to be tweets of interest, and a minimal number of clusters that cover as much as possible those tweets are extracted using a knapsack-constrained maximum covering problem. The extracted clusters are thought to express topics obtained from the tweets of interest, and thus post contents related to specific objectives can be abstracted from a huge amount of tweets. A computational experiment shows the effectiveness of the proposed method with reference to a TV animation program "Space Brothers".