Invited talk by Dr. Jiaul Paik on High Quality Search: Ranking Models and System Design

Title of the talkHigh Quality Search: Ranking Models and System Design
Speaker:Dr. Jiaul Paik
TimeAugust 26th (Wednesday) from 3:00 pm - 4:00 pm
Venue: Room 121
Host Faculty: Dr. Manish SIngh


ABSTRACT
In this talk, I will describe my recent work on ranking models as well as  the  design  of  system  that will  facilitate  large  scale  data processing.  I  will start by pinpointing  the significant  limitations of the ranking functions widely used for decades and then describe the development of  my work  that addresses  the problem  effectively.  To that  end, I  will  present  a model  that  is  guided by  information theoretic principle.

Although, the existing models are  known to be reasonably effective in controlled  setup  with  smaller   data  set,  their  brittleness  are strikingly  evident on  very large  modern datasets  with millions  of documents.   The next  part  of  my talk  will  introduce yet  another probabilistic ranking model  based on extreme value  theory that gives state of the  art results by outperforming  the existing probabilistic models by a significantly large margin. Finally, the  last part  of my  talk will focus  on system  design for retrieving   high  quality   information   from   massive  amount   of data. Specifically,  I will  talk about my  ongoing work  on massively parallel  text  processing  algorithms  that are  built  on  SPARK,  a distributed framework for  large scale data processing. I  will show a case  study  where  a  parallel   algorithm  executed  on  a  scale-up architecture with significantly lesser computing resources outperforms a powerful scale-out architecture with hundreds of cores and terabytes of memory.

SPEAKER BIO
I  am a  postdoctoral researcher at  the University  of Maryland, College Park,  USA (http://www.umiacs.umd.edu/~jiaul/). I work  on web search, information retrieval and "big data", with a particular focus on  system building  and large-scale  distributed algorithms  for text processing.  Before moving to University of Maryland, I was a visiting scientist  at  Indian  Statistical  Institute, Kolkata  from  where  I received my PhD in Computer Science in 2013.

Dates: 
Wednesday, August 26, 2015 - 15:00