A Semantic Continuity Based Analysis of Topic Lifecycle on Social Networks

摘要: Analyzing the lifecycle of topics, that are present in user-generated text content, has emerged as a mainstream topic social network research. The literature presently identifies topics on Twitter, prominent online network, either individual hashtags, or burst keywords within short span time, latent concept spaces obtained from sophisticated analysis mechanisms, such Latent Dirichlet Allocation (LDA). first and second approaches fail to recognize do not restrict themselves hashtags likely across (semantically related) keywords, while third does capture user’s intended expressed via hashtags. In current paper, we propose novel methodology addresses these shortcomings. We jointly utilize temporal concurrency contained given tweets space addressed by tweet identify groups representing space—a “topic”—addressed many tweets. A topic, thus, is represented different set representative at times; usage rate change some gain prominence over others time. Unlike literature, where one typically comprises analyzing hashtag, analyze characterize combination multiple semantically temporally related derive insights about lifecyle topics: inception continuity time (expressed hashtags), how morph another, before eventually dying down.