Using Unspecified Ontology to Search and Place Ads in Pe

Document Sample
Using Unspecified Ontology to Search and Place Ads in Pe Powered By Docstoc
					Semantic Platform for Context-
Aware Publish/Subscribe
M-Commerce

     Shlomo Berkovsky
     Yaniv Eytani
Outline

 Mobile context-based systems
 Publish/subscribe paradigm

 Semantic data management

 Context-aware publish/subscribe
  services
Mobile systems

   Portable ubiquitous computing
   M-Commerce
   Limited interactivity of mobile devices
       No keyboard
       Tiny screen
       Low bandwidth
   Limits amounts of data sent to the user
   Need for personalized applications
Context-aware systems

   Context is:
       Collection of information characterizing the
        situation of a person or relevant object [Dey]
       Subset of physical and conceptual states of
        interest to a particular entity [Pascoe]
   In relation to mobile application:
       Situational = location, environment
       Profile = interests, informational needs
       Technical = end-user device, network state
Publish/Subscribe paradigm (P/S)

   Mediator between provider (publisher) and
    consumer (subscriber)
   Checks for matching and triggers notification
   Properties:
       Anonymity: no need to identify the other side
       Robust: stable for sudden changes (connections
        and disconnections)
       Multicasting: sent once, accepted everywhere
   Well suited for mobile wireless environment
Semantic data management

   P/S over mobile environment require
    deployment of smart semantic matching
   Ontology – definitions:
       A specification of a conceptualization
       A shared formalization of a conceptualization of a
        domain
   Two ontology-based data management
    approaches: HyperCup and UNSO
HuperCup [Schlosser et al.]

   Implemented over hypercube topology
   Efficient broadcast and search can               6       0       7
                                             2                   2

    be achieved within a logarithmic     3       0
                                                     1
                                                         2
                                                                     1



    number of steps                      1
                                             2
                                                     4
                                                         1
                                                             0

                                                                 2
                                                                     5



   Search of multiple keys              8       0       1



     according to a fixed predefined ontology
HuperCup [Schlosser et al.]

   Example: query = C2(C0C1)
   Creates “concept clusters” for a sets of
    common keys
                                      1011                                              1111
                              3                     4             1            4
                                                        0011          0111
                                                                                    3
                       1010                                                  1110
                                                                                           2
                                  4          0010              0110
                                                                      4
      C0 C 1 C 2 C 3
                                                        0001          0101

                                      1001                                              1101

                                             0000              0100
                                                                                    3
                       1000                                                  1100
UNSO                  [Ben-Asher and Berkovsky]

   UNSO - ontology which is not fully specified
   Parts of ontology are dynamically specified
    by users
   Each object is described as an uspecified
    vector of (prop:val) pairs
       [ object:car | manufacturer:BMW | color:red |
        mileage:5000 ]
       [ object:book | author:Tolkien | name:”Lord of the
        Rings” | pages:360 | edition:3 ]
UNSO               [Ben-Asher and Berkovsky]

   Hash functions are used to map
    ontology keys to numeric values
   The prop’s are hashed to a
    coordinate indices and the val’s are hashed to
    numeric values forming a new UNSO vectors
   Dynamically generates structure of Multi-
    Layered Hypercube (MLH)
   Supports semantic routing of HyperCup
Semantic context-aware P/S

   Low interactivity of mobile users
   Explicit subscription is replaced with
    automatic subscription based on context
   Mapping of both publishers and subscribers
    to an MLH:
       High layers – location
       Low layers – interests
   Context is translated to
    semantic query
Semantic context-aware P/S

   Matching and notifications on the MLH
   Fast adaptation to changes
   Change of context  automatic re-subscription
     new semantic query  new matchings
   Wide commercial use:
       Content services
       Advertisement
       Recommender systems
       …
   Example scenario




                           NY, relaxation, pizzeria




LA, relaxation, pizzeria
Prototype demonstration
Future research

   Dynamic learning of the user’s profile
   Ranking received information
   Developing dynamic routing algorithm
   Large scale experiments
   Additional M-Commerce functionalities
Q&A

      Thank You!

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:0
posted:10/3/2011
language:English
pages:16