1 Giuseppe Riccardi, Marco Ronchetti University of Trento
1
Giuseppe Riccardi, Marco RonchettiUniversity of Trento
Giuseppe Riccardi University of Trento 2
Outline
�Searching Information
�Next Generation Search Interfaces
�Needle
�E-learning Application
�Multimedia Docs Indexing, Search and Presentation
�Demo
�Conclusion
Giuseppe Riccardi University of Trento 3
Searching Information(Past)
�Query model
�State-of-the-art -> bag of words
�Loyal Users
�Indexing
�Web pages, pdf, doc...
�State-of-the-Art ->index data structures
�Evaluation
�$$$
�Increasing the quality of the retrieved docs through
�Ranking
�Crawling
Giuseppe Riccardi University of Trento 4
“Where is Mississippi”(11/30/2006)
Giuseppe Riccardi University of Trento 5
Outline“Microsoft Reorg a bulwark against Google”
(11/30/2006)
Giuseppe Riccardi University of Trento 6
Searching Information(Present)
�Meta-Engines
�Docs are retrieved by de-facto standard search engines
�Query-Answer pairs extraction (e.g. ask.com)
�Docs are
�Clustered (many-to-one) (e.g. vivissimo.com)
�Visualized via multiple views (many-to-many)
Giuseppe Riccardi University of Trento 7
Outline“Microsoft Reorg a bulwark against Google”
(11/30/2006)
Giuseppe Riccardi University of Trento 8
Companies“Microsoft Reorg a bulwark against Google”
Giuseppe Riccardi University of Trento 9
Topics“Microsoft Reorg a bulwark against Google”
Giuseppe Riccardi University of Trento 10
Stories“Microsoft Reorg a bulwark against Google”
Giuseppe Riccardi University of Trento 11
Read/Center Story“Microsoft Reorg a bulwark against Google”
Giuseppe Riccardi University of Trento 12
“Microsoft Reorg a bulwark against Google”
Giuseppe Riccardi University of Trento 13
What is next?
�Typical tasks of web users
�Transactional
�Navigational
�Informational
�Task-Driven search
�Information search is part of a given task
�Business Intelligence (e.g. Decision Making)
�E-Learning ( e.g. Student E-Tutoring)
�Vertical Search Engines
Giuseppe Riccardi University of Trento 14
Next Generation Information Search Interfaces
� Search Multimedia Documents� Indexing, Ranking� Large Scale � Real-time
� Query� Multimodal Input (text, gesture, speech)
� Vertical Engine� Limited Domain (e.g. business, education)� Structured & Annotated Content (not free!)� Certification of the results
� “What is the success rate of medication X?”� Results
� Multimedia Presentation� Bandwitdth
Giuseppe Riccardi University of Trento 15
Needle Research Program
� Search Multimedia Content� Audio, Video, Metadata streams
� Indexing� Video� Automatic Speech Recognition (Unlimited-CSR)� Semantic Segmentation � Topic segmentation� Domain Ontology
� Input� Natural Language Query (Spoken or Text or
Multimodal)� Presentation
� Multimedia Presentation� Usability
Giuseppe Riccardi University of Trento 16
Large amount of different kinds of educational resources.
Video lecturesBooksSlidesInteractive whiteboard streams
Goal : Information Search Interfaces for e-Learning
Needle: E-Learning Domain
Giuseppe Riccardi University of Trento 17
Where is the content?Domain: Education Domain
� MSRI� Math-CS research/advanced topics (skewed)� Video/Audio lectures� Presentation vgs from video close shots
� MIT� Courseware (syllabus, lecture notes)� Video/Audio lectures� Wide range of topics
� University of Trento� Video/Audio lectures� Synch powerpoint presentation-video-audio� Skewed topics (CS & other)
Giuseppe Riccardi University of Trento 18
System Components
� LODE� Content Creation � video lecture acquisition and synchronization with the
learning materials, and of their reproduction in a web browser.
� Needle : Interface for searching though the multimedia content and generating the multimedia documents.
Giuseppe Riccardi University of Trento 19
LODE
LODE is a software for low-cost acquisition of lectures – no special requirements for the end user.
Streaming or Download
Off-Line(DVD)
•Good quality audio and video +•Images of the slides projected in class •Tools for navigating the lecture(by section title, by other indexes or through a time-slider) •Annotating video lectures withdocuments . •One DVD for a 50-hours class (MP4).
Giuseppe Riccardi University of Trento 20
System Architecture
Transcripts
Slides
InteractiveWhiteboard
Forums
MeetingRecordings
Audio
Video
Giuseppe Riccardi University of Trento 21
DB Structure
5 main entities: Actor ( the
teacher), Event (a lecture) , Series (a course), View (part
of a document) and Document (a MS PowerPoint presentation).
Giuseppe Riccardi University of Trento 22
Lecture Topics
86%
8% 6%
Computer Science
Meteorology
Sociology
Multimedia Database(2003-Present)
Languages
24%
76%
English
Italian
Giuseppe Riccardi University of Trento 23
Database Statistics
277198634
416
201Sociologia del Turismo [2003]
401Programmazione 2 [2003]
401Lab. Programmazione 2 [2003]
151Lab. Sistemi Operativi [2004]
151Lab. di Algoritmi e Strutture Dati [2004]
401Ingegneria del Software [2004]
401Architettura degli Elaboratori [2004]
33Science Faculty Seminars [2005]
401Programmazione 2 [2006]
24862Corso Meteorologia [2005]
401Machine Learning [2006]
401Distribuited Systems - Design [2005]
10SSSW05 [2005]
10SSSW06 [2006]
10WeeNet Summer School [2006]
HTL06
hoursspeakershoursspeakers
ItalianEnglish
Giuseppe Riccardi University of Trento 24
Utterance Length Statistics
Words
Fre
quen
cy
Min : 1Max : 78Average : 19,9
Giuseppe Riccardi University of Trento 25
Multimedia Indexing(speech driven)
“operatore new”Speech
Video
Slides
time
Giuseppe Riccardi University of Trento 26
Multimedia Indexing(Metadata driven)
“operatore new”Slides
Video
time
Giuseppe Riccardi University of Trento 27
Prototype
� Multimedia data streams (Audio, Video, ASR, Metadata)
� Indexing� Multimedia docs search
� Present & Browse
Giuseppe Riccardi University of Trento 28
DemoWith Angela Fogarolli, Alessandro Bertacco (UNITN)
Giuseppe Riccardi University of Trento 29
E-learning evaluationKirkpatrick’s 4 levels� Level 1 Reactions (Qualitative)
� Did they like it? Was the material relevant to their work?
� Level 2 Learning (Quantitative)� formal to informal testing to team assessment
and self-assessment.� Level 3 Behavior (Qualitative)
� Are the newly acquired skills, knowledge, or attitude being used in the everyday environment of the learner?
� Level 4 Results (Quantitative)� measures the success of the program in terms
that managers and executives can understandKirkpatrick, D.L. (1994).
Evaluating Training Programs: The Four Levels. San Francisco, CA: Berrett-Koehler.
Giuseppe Riccardi University of Trento 30
� Multimedia database with resource of other kind (interactive whiteboard recording, discussion, real and virtual meeting registration).
� Ontologies linking to offer knowledge-supported search.� Training of Unlimited-ASR
� Portable (domains)� Spoken Language Understanding (Query/Doc)� Semantic indexing� Evaluation
� E-learning domain� Content Creation
� Inter-University collaborative efforts
Future Research
Giuseppe Riccardi University of Trento 31
� Information Search� Past & Present
� Next Generation Information Search� Needle
� Multimedia Documents� Indexing� Search� Presentation
� Content Creation� Inter-University collaborative efforts
Conclusion