JIKD EmailSpeech Update

1
JIKD Email/Speech Update

• Doug Oard
• October 26, 2005

2
JIKD Timeline

• March 10, 2005 Year 1 start
• June 15 Kickoff
• September 9 Site visit
• Oct 26/Nov 30!/Dec 21? Updates
• mid-January? Site Visit
• March 10, 2006 Year 2 start

3
Research Progress

• Yejun TREC Enterprise (W3C mailing lists)
• Topic development (5)/relevance judgment (9)
• Pro/con search (good)/known item search (ok)
• Adam Visualization (Shneiderman emails)
• Time/Participant thread enrichment
• Rank by Feature Framework
• David end-to-end summarization (Enron)
• Sandeep Close-set speaker ID (NIST 98)
• Tamer Quoted text threading

4
Near-Term Research Plans

• Annotation planning meeting (Raj, Oct 31)
• Email/phone cross-references
• Sentence extraction (Sergei and Tamer)
• Rouge test set (David and Erin)
• Mentioned name resolution (Tamer)
• Machine-gen email detection (Tamer)
• Encrypted indexing (Ashwin)
• Ranked retrieval, efficiency, integration
• Netlens data flow/history (Hyunmo)
• Speaker ID on Enron (Sandeep)

5
Research Infrastructure

• ODBC/Web services
• ISI speech act annotations being integrated
• Text indexing using Lemur (with Yejun)
• UIMA
• Ryen White / Ayelet Goldin at IBM this week
• Andreas is looking at this for speech
• First effort will be named entity section
• Ordering full collection from Aspen Systems

6
Collaboration Opportunities

• Johns Hopkins APL
• Chris Diehl (with Lise Getoor)
• Christine Platko Jason Eisner
• Loyola University (Baltimore)
• Roberta Sabin (speech acts)
• Dawn Lawrie (summarization)?

7
Recent JIKD Publications

• Adam JASIST article
• Yejun SIGIR Enterprise Track report
• Mara LAMP technical report
• Galileo SDM conference paper submitted
• Sandeep ICASSP paper submit (not Enron)

8
Centers of Gravity
HCI Catherine Plaisant
Identity Lise Getoor
Data TJ Rogers