Design of Improved Social Event Story Board From Image Click-Through Data

  • Unique Paper ID: 144859
  • Volume: 4
  • Issue: 5
  • PageNo: 109-113
  • Abstract:
  • Traditional sites had been pushed by human edited situations which lead to big web search traffic. This paper is actually a survey conductedfor identifying the different event detection approaches that are beneficial for event mining.While traditionalsites can just present human edited functions, in this particular paper weprovide a novel system to immediately identify events from searchlog data and produce storyboards in which the events are actually arranged chronologically. We selected image search log as the source forevent mining, as search logs could even represent people's interests.In order to find happenings from log data, we show a Smooth Nonnegative Matrix Factorization framework (SNMF) that combinesthe info of query semantics, temporal correlations, research logs as well as time continuity. Additionally, we think about the time factoran essential component since many different events will build in various time tendencies. Additionally, to make a media-rich&visually attractive storyboard, each and every event is actually related with a set of representative pictures set up along a timeline. Theserelevant images are instantly selected from image search engine results by analyzing image content features. Celebrities are used by us as our test domain, that takes a great percent of image search traffics.
email to a friend

Cite This Article

  • ISSN: 2349-6002
  • Volume: 4
  • Issue: 5
  • PageNo: 109-113

Design of Improved Social Event Story Board From Image Click-Through Data

Related Articles