Semantic Context Specification

1
Semantic Context Specification

• April 28, 2004
• Amit Nanda, Reiman Rabbani, Maciej Janik

2
Introduction

• Context What is it?
• Semantic context.
• Benefits from having a context ?

3
Example Hotel check-in

• David
• business trip
• has to get up at 7am
• does not like AC
• has to watch CNN finance news
• teetotaller

• Robert
• leisure trip
• can sleep till 9am
• likes to have cool room
• watches Sports
• likes to drink champaign

4
Example Hotel check-in
David
default room settings
alarm set at 7am AC turned off TV starts with
CNN non-alcoholic drinks in fridge
default room settings
alarm set at 9am AC turned to 70 F TV starts with
Sport channel champaign in fridge
default room settings
Robert
5
Context f( task, ontology, data )
Our Proposed Model
6
Context TASK

• Operation that has to be performed.
• Goal to accomplish.
• Assumption Task has to be application specific
  for best results.
• Example Task travel. On expedia.com and on
  amazon.com, results differ.

7
Context DATA

• Vast store of information, covering different
  domains, a part of which constitutes the end
  result for user task.
• In semantic web, all data corresponds to real
  world objects and objects have different types of
  relations between each other
• Denoted in the form of triples, encoded as graphs
  in RDF files.

8
Context ONTOLOGY

• Specification of a conceptualization.
• Can be visualized as skeleton or framework for
  the data.
• Specified by OWL or DAML-OIL.
• Assumption a task that needs to be executed,
  should have a well defined ontology.

Gruber, T.R. A Translation Approach to Portable
Ontology Specifications. 1993.
9
Profile Ontology Dynamic Profile Update

• Profile as an ontology
• Assumption Everything in the ontology is a
  universally defined object
• When new object are required the ontology and
  relationships of that object can be downloaded
• User profile may or may not be updated.
• If task is executed using the users context,
  profile is updated based on rules settings.
• If users profile has no information regarding
  task and if its still executed, profile is
  updated using current results.
• Profile is not updated if task is not executed

10
Profile Ontology
11
Profile Ontology
lt?xml version"1.0"?gt ltrdfRDF
xmlns"http//a.com/ontology"
xmlnsrdf"http//www.w3.org/1999/02/22-rdf-syntax
-ns" xmlnsrdfs"http//www.w3.org/2000/01/rd
f-schema" xmlnsowl"http//www.w3.org/2002/0
7/owl" xmlbase"http//a.com/ontology"gt
ltowlOntology rdfabout""/gt ltowlClass
rdfID"Historical_data"gt ltrdfssubClassOfgt
ltowlClass rdfabout"Temporal_data"/gt
lt/rdfssubClassOfgt lt/owlClassgt ltowlClass
rdfID"Address"gt ltrdfssubClassOfgt
ltowlClass rdfabout"Spatial_Enviroment"/gt
lt/rdfssubClassOfgt lt/owlClassgt ltowlClass
rdfID"Private_address"gt ltrdfssubClassOf
rdfresource"Address"/gt lt/owlClassgt
ltowlClass rdfID"Interests"gt
ltrdfssubClassOfgt ltowlClass
rdfabout"Preferences"/gt lt/rdfssubClassOfgt
lt/owlClassgt ltowlClass rdfID"Business_addre
ss"gt ltrdfssubClassOf rdfresource"Address"/
gt lt/owlClassgt ltowlClass rdfID"Privacy_rule
s"gt ltrdfssubClassOfgt ltowlClass
rdfabout"Rules"/gt lt/rdfssubClassOfgt
12
Context
Context sensitive results
Questions .
Task Application
User Application
Task
Answers .
Profile Data
13
Camera Ontology
14
Camera Ontology - sample

• ltrdfRDFgt
• ltowlOntologygt
• ....
• ltMoney rdfID"Eos_body_price"gt
• ltcurrencygt550lt/currencygt
• lt/Moneygt
• ltBody rdfID"Canon_Eos"gt
• ltcost rdfresource"Eos_body_price"/gt
• lt/Bodygt
• ltDigital rdfID"Olympus_Camedia_C2040"gt
• ltcostgt
• ltMoney rdfID"Olympus_price"gt
• ltcurrencygt500lt/currencygt
• lt/Moneygt
• lt/costgt
• lt/Digitalgt
• ltMoney rdfID"Zeiss_price"gt
• ltcurrencygt450lt/currencygt
• lt/Moneygt

15
Example of camera context ...
buy camera
my cameras ...
User App
Merchant App
cameras to offer personalized view
profile query what cameras do you have ?
ltMoney rdfID"Eos_body_price"gt
ltcurrencygt550lt/currencygt lt/Moneygt ltBody
rdfID"Canon_Eos"gt ltcost rdfresource"Eos_b
ody_price"/gt lt/Bodygt ltDigital
rdfID"Olympus_Camedia_C2040"gt ltcostgt
ltMoney rdfID"Olympus_price"gt
ltcurrencygt500lt/currencygt lt/Moneygt
lt/costgt lt/Digitalgt ltMoney rdfID"Zeiss_price"
gt ltcurrencygt450lt/currencygt lt/Moneygt ltSLR
rdfID"Canon_Eos_10"gt ltbody
rdfresource"Canon_Eos"/gt lt/SLRgt ltLens
rdfID"Zeiss"gt ltcompatibleWith
rdfresource"Canon_Eos"/gt ltcost
rdfresource"Zeiss_price"/gt lt/Lensgt
ltMoney rdfID"Eos_body_price"gt
ltcurrencygt550lt/currencygt lt/Moneygt ltBody
rdfID"Canon_Eos"gt ltcost rdfresource"Eos_b
ody_price"/gt lt/Bodygt ltDigital
rdfID"Olympus_Camedia_C2040"gt ltcostgt
ltMoney rdfID"Olympus_price"gt
ltcurrencygt500lt/currencygt lt/Moneygt
lt/costgt lt/Digitalgt ltMoney rdfID"Zeiss_price"
gt ltcurrencygt450lt/currencygt lt/Moneygt ltSLR
rdfID"Canon_Eos_10"gt ltbody
rdfresource"Canon_Eos"/gt lt/SLRgt ltLens
rdfID"Zeiss"gt ltcompatibleWith
rdfresource"Canon_Eos"/gt ltcost
rdfresource"Zeiss_price"/gt lt/Lensgt
what is a camera ?
P R O F I L E
Context
KnowledgeBase
InventoryDB
Camera Ontology
16
Security Issues

• Greater the details provided by user application
  from users profile, better the result.
• But the user application is required not to
  provide user details regarding salary, SSN, bank
  account etc
• This is a constraint or drawback.

17
Conclusions

• In our model, context is created on application
  side, using information provided by user or
  profile.
• Specialized application is able to create a
  better context, than general user application.

18
Conclusions

• Because of the vastness of application areas,
  difficult to categorize them into definite
  types.
• Generalization of context is difficult and not
  always feasible.
• It is not possible to create a unified context
  that suits every need.

19
Conclusions

• Profile may grow to enormous proportions
  (including vast amount of information) and become
  hard to manage.
• Security issues may arise, because of complexity
  of data.
• Excessive transparency of profile will lead to
  abuse by malicious applications
• Sensitive information can be inferred and rules
  can be bypassed by use of circular logic by
  dishonest programmers
• Future work apply this model of context to
  different applications.

20
Questions?

• Thank you ! ?