Web Service Reputation Conceptual Model Presentation on November 26th 2008 Team Members Sowmya J Bandalli Karen Doore Preeti GaneshMohan Title Conceptual Model of Web Service Reputation Authors E Michael Maximilien IBM and NCSU maxim us ibm com Munindar P Singh North Carolina State University singh ncsu edu URL http delivery acm org 10 1145 640000 637417 p36 maximilien pdf key1 637417 key2 0037377221 coll GUIDE dl GUIDE CFID 12837775 CFTOK EN 70178819 Web Service Reputation Conceptual Model A Web service represents a set of functions addressing a particular domain Web Service Attributes Generic Attributes Domain Specific Attributes Reputation of a web service Is a function of a web service that depends on attributes It depends on the relative importance of each attribute with respect to the agent using the service Web Service Reputation Important Factors Relative Weights given to attributes Dependant on Agent Biases Dependant on Specific Domain Attribute Aggregation Algorithm Simple Majority Other Algorithms Endorsers of the Service Trust level of Endorsers History of Service Services that are new won t have as many endorsers Older services have more history data Dampening Factors for Ratings Allows a web service s reputation to increase with improved service More recent activity matters more than old activity WASP Web Service Agent Proxy Conceptual Model of Web Service Reputation Conceptual Model UML Static Model Figures 2 3 1 A Web Service has 1 associated Reputation for an agent considering the service 2 A Reputation of a service is a vector of attribute values 3 4 5 6 A Reputation is determined by a reputation algorithm that aggregates attributes that are selected to be considered by the agent A Reputation is also affected by a History that keeps previous ratings for a particular service The rating for a particular service is determined for a particular principal agent using the RatingAlgorithm which takes into consideration the agent s bias about the relative importance of each attribute Endorsements by other Principals agents are also considered for determination of Reputation the weighting given to an endorsement is dependant on the trust level of the Principal giving the endorsement Conceptual Model of Web Service Reputation Conceptual Model of Web Service Reputation Attributes Model Domain are collections of attributes for a specific type of service Cross Domain attribute Domain specific attribute Attributes Each Attribute has a range or set of allowed values can be integer percentage Boolean etc Relative Weight of an Attribute Relative to Domain User preferences Characteristic function from Attribute value to Rating S shaped lower threshold and upper saturation level Monotonically Increasing Step wise function Temporal Characteristic Decaying value of an attribute Example travel delay if a plane is delayed 10 minutes is tolerable however a delay of 45 minutes may be intolerable because it causes problems with connecting flights Generic and Domain Specific Attributes Web Service Attribute Domains For Web services that provide similar services they will share common attributes perhaps weighted differently For different domains Example Car rental service Domains Travel Price Flexibility of Reservation Changes Retail Price Inventory Flexibility of Reservation Changes may be used by some services to differentiate their service Determining Attributes for a Domain Suggestion let the attributes for a domain be determined by community of users and providers for each domain Ideally want a system where new attributes can be added to domains and where services can distinguish themselves by focusing on certain attributes So that service doesn t need to have values for all attributes Conceptual Model of Web Service Reputation Conceptual Model Web Services Agent Proxy Client Desires Service Client uses WSAP to access web services Web Services Agent Proxy Consults with outside registries reputation and endorsement agencies to find an appropriate provider of a desired web service WSAP can use information about the user s priorities with information in the form of reputation attributes and endorsements to determine the service which is optimum for the client The WSAP can provide feedback to the reputation and endorsement agencies based on the WSAP and user experience to keep reputation and endorsement data up to date
View Full Document
Unlocking...