WING%20Research%20Group

1
WING Research Group

• Demos and Posters

2
Demos

• SlideSeer (M.-Y. Kan)
• Coordinating presentation slides and scholarly
  papers
• Align set of slides to set of paragraphs
• Difficult as alignment is not monotonic, nil
  alignments possible
• JCDL 2007
• LINC 2.0 (J. Prabawa)
• - How to build a better library catalog interface
  using AJAX
• - 4 key areas of user design change overview
  details, AJAX grid, tabs, suggestion bar,
• - Usability studies for evaluation
• - JCDL 2007 short paper

3
Posters

• Scenario Template Generation (L. Qiu)
• Joint work with Tat-Seng Chua
• Create template of slots for articles from
  multiple different event reports
• Uses EM based context sensitive clustering
• Phrase-based Statistical MT (H. Setiawan)
• Joint work with Haizhou Li (I2R)
• Focus on function words to reorder arguments
  (e.g., English preposition above to Chinese
  postposition ? )

• Record Linkage using Web resources (Y. F. Tan)
• - Joint work with Dongwon Lee (PSU)
• - How to best combine disparate info from various
  sources of web evidence
• Math IR (J. Zhao)
• - Working on indexing, retrieving mathematical
  information
• - Key problem co-ref between formula name
  Pythagorean theorem and equation form and
  variants.
• Graph-Based Text Summarization (Z. Lin)
• - Joint work with Wee Sun Lee
• - Using incrementally built graphs and PageRank
  to build better generic and update summarization
  systems

4
(No Transcript)
5
SlideSeer Aligned Presentation and Document
Alignment

• Min-Yen Kan
• Web IR / NLP Group (WING)
• http//wing.comp.nus.edu.sg

6
Aligning slides to scholarly papers

• Why do it?
• Slides can be seen as a dual and/or summary of
  papers
• Useful as a learning and comprehension aid
• See OCW example in Leslie and Tomas lecture
  notes
• Whats the goal?
• A digital library of slides/papers and their
  fine-grained alignment to scholarly papers (an
  extension of CiteSeer or Rexa)
• What data to use?
• Seeded with CiteSeer data (thanks to CL Giles),
  merged in DBLP
• Searched for corresponding ltppt, pdfgt
  presentation using Google API

7
SlideSeer Demo

• Proof of concept
• 10 presentation/paper pairs manually aligned
• Allows focus on one medium with other in context
• Allows full slideshow mode
• Navigation using mouse, keyboard shortcuts
• Search using Lucene IR package

8
Next steps

• Planning to hook up current work in progress
• 2 stage CRF/SVM re-ranking citation segmentation
  algorithm
• Automatic keyphrase extraction program
• Automatic synthetic image classification
• Automatic de-deduplication module
• Partnering with Simone Teufel (Cambridge, UK) to
  do argumentative zoning of documents
• What is a citation used for?