XML and Data Integration

1
XML and Data Integration

• Edward Yau
• (2002/03/27)

2
Fundamental Digital Video Library Functions

• Functions
• Extraction and Discovery
• Searching and Indexing
• Query and Presentation

3
Modalities

• Video Information Processing
• Modality concept
• Video Channel
• Scene break , Face, Logo
• Audio Channel
• Speech -gt Text (Summary, Topic)
• Event Channel
• Mouse click, movement

4
Functions x Modalities
Text Name Feature Vector Extraction Discovery
String Match Exact Match Distance Match Search
ing Indexing
Text Graphic Text Text Query Presentation
Speech
Logo
Modalities
Face
Knowledge
Functions
5
iView System

• Container system for modalities modules
• Easy integration of additional modalities
• Minimal effort if replacing different module of
  same modality

6
XML as Data Glue

• Each modality has it own data format
• Some of them already in XML
• NewsXML, MathXML,
• Design policy to integrate them together

7
Data Integration

• Data model reconciliation
• Common data model
• Time/Frame
• Over specify reduce flexibility
• Schema Reconciliation
• Different representation format
• Semantically the same
• Instance Reconciliation
• Contradictory information

8
ltxmlgt ltmodality nametranscriptgt
lttranscript time000.0 typetextgt
???? lt/transcriptgt lt/modalitygt
ltmodality namefacegt ltface
time015.4 typefacegt
ltscreen_pos x10 y20 width100
height80\gt ltface_vectorgt10.9,
21.9, 23 .lt/face_vectorgt lt/facegt
lt/modalitygt ltmodalitygt ltlogo
time101034gt ltlogo_name
typetextgt TVBlt/logo_namegt
ltlogo_idgt1542lt/logo_idgt lt/logogt
lt/modalitygt lt/xmlgt
Speech Recognizer
Face Recognizer
Logo Recognizer
9
ltxmlgt ltmodality nametranscriptgt
lttranscript time000.0 typetextgt
???? lt/transcriptgt lt/modalitygt
ltmodality namefacegt ltface
time015.4 typefacegt
ltscreen_pos x10 y20 width100
height80\gt ltface_vectorgt10.9,
21.9, 23 .lt/face_vectorgt lt/facegt
lt/modalitygt ltmodalitygt ltlogo
time101034gt ltlogo_name
typetextgt TVBlt/logo_namegt
ltlogo_idgt1542lt/logo_idgt lt/logogt
lt/modalitygt lt/xmlgt
iView XMLRepository
Text Index Engine
Face Index Engine
Logo Index Engine
10
ltxmlgt ltmodality nametranscriptgt
lttranscript time000.0 typetextgt
???? lt/transcriptgt lt/modalitygt
ltmodality namefacegt ltface
time015.4 typefacegt
ltscreen_pos x10 y20 width100
height80\gt ltface_vectorgt10.9,
21.9, 23 .lt/face_vectorgt lt/facegt
lt/modalitygt ltmodalitygt ltlogo
time101034gt ltlogo_name
typetextgt TVBlt/logo_namegt
ltlogo_idgt1542lt/logo_idgt lt/logogt
lt/modalitygt lt/xmlgt
iView Browser (Timing)
Text Presentation
Face Presentaion
Logo Presentation
11
Time Tag Problem

• Common data modal and data reconciliation
• Tag
• lttimegt
• ltfacegt
• Attribute of Tag
• ltface time gt
• Attribute of Tag to Tag translation policy