Seed Selection for Genre Specific Search

This poster was presented at IIIT-H RnD show case

Transcript

1. Seed Selection for Genre Specific Search
   Search and Information Extraction Lab P Nikhil Priyatam Krish Perumal Dharmesh Kakadia
   Vasudeva Varma
   International Institute of Information Technology, Hyderabad
   AIM  
   •  This work aims to get a set of diverse seed URLs
   for genre specific search using Twitter data.
   SYSTEM  ARCHITECTURE  
   PROPOSED  ALGORITHM  
   WORKING  OF  ALGORITHM  
   EVALUATION  ARCHITECTURE  
   EXPERIMENTAL  RESULTS  
   MOTIVATION  
   •  Coverage and diversity are crucial aspects of
   genre specific search engines. These depend
   largely on the initial set of seed URLs. There is
   no existing work that automates the process of
   seed URL selection with a focus on diversity.
   CONCLUSION  
   •  First work to automate the process of seed URL
   selection for genre specific search
   •  Addressed the issue of crawl diversity, which was
   hitherto neglected.
   DIVERSITY  SCORES  
   SIMILARITY  MEASURES  FOR  EDGES  
   •  Content overlap
   •  URL n-gram overlap
   •  Timestamp similarity
   •  Follower-followee relations or Retweets
   

   View Slide