PowerPoint-Pr

1
(No Transcript)
2
Contents

• Cross media aspects
• in media production (NM2 project)
• in media monitoring (DIRECT-INFO project)
• Conclusions
• Outlook

3
NM2 (New Media for a New Millennium) - Motivation

• Traditional distribution channels
• have developed their specific compelling media
  forms
• art, science and technology of production,
  direction and editing in non-linear narrative are
  poorly formed
• NM2 investigates on
• non-linear, interactive narratives
• tools for production and delivery
• Main Objectives of NM2
• proof of concept to produce compelling non-linear
  narratives
• production of compelling media
• aesthetically pleasing
• truly engaging (30 minutes)
• plausible from a business perspective

4
Basic Technical Concepts

• Narrative Objects (NO)
• MPEG-7 to describe audiovisual features
• OWL to describe semantics of NO
• Narrative Structure Language
• internal representation of a story world
• basic narrative structures

5
Media Semantic Mapping Utility (Monsoon)
Tabs for objects, concepts and relations
Definition of an object by using other objects
Definition of an object by low level features
6
System Architecture
7
Main Project Results

• Definition of a Narrative Structure Language
• describing multimedia objects in narrative
  context
• Bridging semantic gap
• MPEG-7 OWL
• linking lower level features to higher level
  concepts
• Six NL productions with compelling media
• Targeted for three end-user devices
• Windows XP Media Centre and TV card
• Game consoles
• Mobile phones

8
DIRECT-INFO - Background Media Monitoring
Companies

• Measurement of advertisement expenditure
• Find out
• how much a company has spent for advertising
• on a product level
• over different media
• Typical business cases
• advertisement information
• creativity (new campaigns) (targeted in
  MediaCampaign)
• sponsorship tracking (targeted in DIRECT-INFO)

9
Sponsorship Tracking

• Sport teams need sponsors to finance their
  activities
• Measure efficiency of sport sponsorship
• broadcasting print media
• convince sponsors to invest again
• currently done manually
• Technically very complex task
• context information needed
• multimodal analysis necessary

10
Main Technical Innovations

• Identification of logos within visual content
• Multimodal event modelling
• Genre classification
• Text analysis (pos. / neg. mentions)

• Semantic correlation fusion of multi-modal
  content analysis tosemantically meaningful
  reports

11
Technical Workflow (1/2)
12
Technical Workflow (2/2)
13
BrandDetector Detection of Logos
14
MPEG-7 Result Viewer/Editor
Module parameters
Player
Key frames shots
Basic events
15
Conclusions

• Standardized metadata (MPEG-7, OWL) is a good
  choice
• otherwise a lot of time will be lost with
  specification and extensions
• MPEG-7, OWL already provided the needed
  structures
• one expert for MPEG-7, OWL is necessary
• Web services not so straightforward as always
  thought
• managed vs. unmanaged code
• not all SOAP protocol features supported
• Do not underestimate integration efforts!
• take security issues into account
• remote access essential

16
Outlook MediaCampaignRD project, Started
04/2006

• Identification and tracking of new campaigns
• in different media (Press, TV, Internet)
• utilizing different modalities (audio, text,
  image, video)
• Semantic cross-market advertis. interlinking
• relate similar advertisements over different
  media countries
• Define Ontology for cross-media modeling of media
  presence and campaigns
• Media Presence and Campaign Ontology (MEPCO)
• includes media presence information

17
Relevant Links

• NM2 website
• http//www.ist-nm2.org/
• DIRECT-INFO website
• http//www.direct-info.net
• MediaCampaign website
• http//www.media-campaign.eu
• JOANNEUM RESEARCH MPEG-7 Library
• available for free!
• full support for parts 3, 4 and 5 of the MPEG-7
  standard
• http//mpeg-7.joanneum.at

18
Dr. Herwig Rehatschek
JOANNEUM RESEARCH Institute of Information
Systems Information Management Steyrergasse
17 A-8010 Graz / AUSTRIA Voice 43 316
876-1194 Mobile [phone] Fax 43 316
876-1191 E-mail Herwig.Rehatschek_at_joanneum.at Web
http//www.joanneum.at/iis
19
(No Transcript)
20
Delivery
21
MPEG-7 Automatic Detection Preview (UI)
22
System Architecture
23
Genre Classification

• Fully automatic classification of
• commercials
• news
• studio/speaker
• soccer
• credits
• movie
• Real-time analysis
• Semantic block classification
• 90 precision
• 90 recall

24
Multimodal Event Modelling

• Event detection in audio and video
• Integration of basic events
• visual shots
• audio segments
• overlaid text
• visible frontal faces
• Pseudo-scene/story segmentation
• Classified segments

25
Text Analysis

• Detect positive negative mentions of
• brand names
• teams
• SCHUG text analysis engine
• Shallow and Chunk-based Unification Grammar
• robust chunk parser for linguistic analysis
• Topic detection
• Sources
• transcripts and text captions of TV audio
• newspaper texts
• supported languages English, Italian

26
Semantic Data Fusion

• Combing data from various analysis subsystems
• Based on time-stamped analysis results of all
  analysis modules
• Rule-based derivation of new facts
• basic appearances
• complex appearances
• Two main user interfaces
• consolidation of appearances
• set-up application for delivery system

27
Semantic Data FusionDisplay of Results
28
Delivery system for Customers