Knowledge Base Building Project 3 rd meeting 2008. 08. 30.
-
Upload
aubrey-sherman -
Category
Documents
-
view
212 -
download
0
Transcript of Knowledge Base Building Project 3 rd meeting 2008. 08. 30.
Knowledge Base Building Project3rd meeting
2008. 08. 30
Copyright 2008 by CEBT
What We Did In Last Week
1st meeting (2008. 08. 25)
Surveying several papers and applications
Presenting and discussing the survey results
2nd meeting (2008. 08. 27)
Discussing several initial issues in our project
– Additional presentation about GoogleBase
– Motivation (users, services)
– Initial data structure
– Benchmark systems and application
– Project maintenance
– Attachment with structured Naver Kin service
Copyright 2008 by CEBT
Additional Survey on GoogleBase
Google Base 의 모델에 대한 보충 Presentation
Storage 는 attribute, Description, Picture or file(Xml, HTML, so on) 등의 정보로 이루어짐
Basic Attribute 는 item type 에 따라 정해짐
Category, Item, Attribute 는 사용자가 추가 가능
Item 들간의 Relationship 은 없는 것 같음
File 은 대략 15 개 정도까지 올릴 수 있음 . 20MB 정도를 한계점으로 지정
API 리스트는 Google base Homepage 에서 찾아볼 수 있음 . 그것을 보고 어떤 기능을 제공해야 할 지를 생각해 볼 수 있을 것 같음
Copyright 2008 by CEBT
Motivation
“What is the main purpose of this project?”
Assistant (product) knowledge base which is able to provide some richer information about domain information to existing applications
“Who is the main target users?”
Should be application designer – they can attach easily our knowledge base to their application by API for enrich the informa-tion
“What we have to do?”
Designing the general data structure of our own
– 1st goal : building the database that is similar to Freebase
Providing graph visualization of knowledge base
Supposing general APIs to retrieve the information in knowledge base
Deciding initial data structure and framework model
Copyright 2008 by CEBT
Initial Data Model
First step
Adapting flat scheme of Freebase data model
– Refer to structure of PPS ontology and GoodRelation
Using the PPS ontology as initial data instance
Providing the way to identify the special relation between the objects (products)
– Supporting to navigate the standard classification of the product : UNSPSC, eCl@ss, EOTD, …
Second step
Considering the expansion to accommodate the public web re-sources as target knowledge in general way
Regarding how to gather and reflect users’ collective intelli-gence in our system easily
Attaching the structured Naver Kin service into the system
Copyright 2008 by CEBT
Abstract Data Model in Freebase
Simple and flexible
Can be transformed into RDF or XML datasets with post processed tag
We are trying to add the set of ontological properties like GoodRelation in the RDBMS schema, and adapt the data schema of PPS ontology on the basis of Freebase data structure
Copyright 2008 by CEBT
Initial Data Structure
OBJECTcategoryrelation
CATEGORY
instancerelation
INSTANCE
relationmap
attributerelation
ATTRIBUTE
RELATION
objectrelation
OBJECT
Copyright 2008 by CEBT
Initial Data Structure: Example
Obj:CAMERARel:
hasCategoryObj:OpticalElectronics
hasInstance
Obj:CANON
Rel:subClassOf
Rel:hasAttribute
Att:Size
Obj:hasRelation
Rel:hasWiki
Obj:DocumentURL
Att:Pixel
Rel:hasCategory
Obj:MultimediaDevices
Att:CategoryID
A set of relation can be adapted from
product ontology (cf. GoodRelation)
Be able to specify a category forStandard categorization (cf. UN-
SPSC)
Att and Rel type also
has attribute type
Entity (type)
Relation (type)
Connection
Edge has their own weight
Which means the probability
of confidence
Copyright 2008 by CEBT
Initial Data Structure: Example (cont’d)
Obj:Dark Knight
Rel:Genre
Obj:Action
Rel:hasRelation
Rel:hasDirector
Att:MovieDirector
Obj:Relation
Rel:wikiURL
Obj:DocumentURL
Att:ReleaseDate
Rel:Genre
Obj:Crime
Entity (type)
Relation (type)
Connection
Rel:ReleaseDate
Obj:Christopher
Nolan
Rel:hasValue
Obj:6 August
2008
Copyright 2008 by CEBT
Data Schema: Example
Obj_Pro
Id
Obj_Rel
Id
Obj_Att
Id
Obj_Cat
Id
Rel_ProAtt
Pro_Id
Att_Id
Weight
Rel_RelAtt
Att_Id
Rel_Id
Weight
Rel_RelAtt
Cat_Id
Rel_Id
Weight
Rel_AttVal
Att_Id
Value_Id
Value
Weight
Rel_ProVal
Pro_id
Value_Id
Weight
Rel_Pro
Pro_id
Pro_id
Weight
Rel_ProCat
Pro_Id
Cat_Id
Weight
Copyright 2008 by CEBT
Collaboration with structured Naver Kin
Object Question
Answer
QuestionURL
AnswerURL
Author UserId
hasAtt URL
CategoryIdRel:
Genre
Category
Object_ID Relation_ID Question_URL
Answer_URL
Question_User_Id
Answer_User_Id Att_Id
00000001 00000010 http://kin.naver.com/q1.html
http://kin.-naver.com/a1.html
questioner Walkdic 00000100
00000002 00000011 http://kin.-naver.com/q2.html
http://kin.-naver.com/a2.html
newbie masterofkin 00000101
…… …… …… …… …… …… ……
Copyright 2008 by CEBT
Simple System Framework
Knowledge Base
End-user Applications
Service Interface
Web Browser
API Engine
Mobile App
ServiceController
Data ExchangeModule
Storage Engine
InferenceModule
Physical I/OHandler
VersioningModule
Rule Engine
Direct API
Optimizer
Other KB Ser-vice
FreebaseGRDDL
Microfor-mat
Physical Storage
LoggingModule
Log
Logical Data Model
Object Attribute CategoryRelationInstance
Wikipedia
RDBMS
NavigationModule
Copyright 2008 by CEBT
Another View of System Frame-work
Knowledge Base
ServiceController
LoggingModule
App. #1
App. #2
App. #3
Data Structure
RuleEngine
Storage Engine
DataVersioning
AP
I Mod
ule
OptimizerI/O
Handler
Log
RDBMS
DataExchangeModule
Other Knowledge Base
Freebase WikipediaStructured
NaverKin
OutsourcingData
Usage Log
Structured Log Data
ServiceRequest
ServiceResult
RequestMessage
API set
Stored Knowledge
ExtractedData
RequestedKnowledgeResult
KnowledgeRequestQuery
Copyright 2008 by CEBT
Issues and ToDo
ToDo
Detailed Data Structure
– Types of object, attribute, relation and category
– Set of attribute and relation
Implementing data schema in RDBMS
– Setting table and columns up
Collaborating with structured Naver Kin
– Refining the target objects in Naver Kin
Issues
Data gathering method
Data Navigate UI