AN APPROACH FOR CONTEXT-AWARE SERVICE DISCOVERY AND RECOMMENDATION
service recommendation
service discovery
Outline
Introduction Our Approach Experiment Conclusion
Outline
Introduction Our Approach Experiment Conclusion
Introduction
Context type location, time
Context Context value New York
Context-aware system: react to a user’s context without their intervention
Problems
Limited support for dynamic adaption to newly added context types
Manually define all the context types Manually establish the relation between
the sensed context scenario and the corresponding services in the form of if-then rules
Outline
Introduction Our Approach Experiment Conclusion
Overview of our approach
Overview of our approach
Ontology
Class: abstract description of a group of concepts with similar characteristics
Individual: instance of a class Property: describes an attribute of class
or individual Relation: ways classes or individuals
associate with each other
Steps of find relevant ontologies
Search with the context
value
Remove the first adj/adv, then search
Annotated the ontology to the context, convert
the remove adj/adv to
constraints
Annotated the ontology
to the context
YES
String is empty
YESNO
Use synonyms of the context
value
NO
Overview of our approach
Identifing context relations
Relations between two Context Values Intersection Complement Equivalence Independence
Identifing context relations
Multiple Context Values: E-R model For each relation of two context values
Convert the two context values into two entities in E-R model
Convert the relation type into a relationship node
Steps of building integrated E-R model
Filter out independence relations Remove equivalence relations Set the integrated E-R model as empty For each relation in the remainder relation list
Convert the relation into an independent E-R model Add the independent E-R model to the integrated E-R
model If exist similarity or equivalence entities, merge them by
keeping the one with the richer information If exist subset or complement relations, add a relation ship node
in the integrated E-R model If two relationship nodes contain the same relation type and
relationship attributes, we merge them into one relationship node
Steps of building integrated E-R model
TravelLos
AngelesInterse
ct
TouristAttractions
Integrated E-R model
Steps of building integrated E-R model
Travel LosAngeles
Intersect
TouristAttractions
Integrated E-R model
NBA
Intersect
Los Angele
s Lakers
Steps of building integrated E-R model
Overview of our approach
Generating searching criteria Suppose are entities in the
integrated E-R model. SharedElementsSet represents the set of a user’s needs.
Generating searching criteria Apply the rules on the E-R model Obtain a SharedElementSet Group the entities in SharedElementSet
Each entity in SharedElementSet is treated as a group
If the entities in one group are a subset of the entities in another group, we combine these two groups together.
Repeat until no groups can be combined Extract keywords from each group as
searching criteria
Outline
Introduction Our Approach Experiment Conclusion
Experiment
Objective Evaluation of the detected context relations Evaluation of Service Recommendation
Precision, Recall
Evaluation of the detected context relations
Five context scenarios Manually examine its context and
identify the potential needs of the user Use our prototype to automatically find
user’s needs
Evaluation of Service Recommendation
Use the keywords in each group as searching criteria to search for online resources.
Use Google and Seekda as the search engine to search for Web pages and Web services
Outline
Introduction Our Approach Experiment Conclusion
Conclusion
Use ontologies to enhance the meaning of a user’s context values
The SharedElementSet reflects user’s needs
Experiment is not clear..