Presentation is loading. Please wait.

Presentation is loading. Please wait.

The Chatty Web : Emergent Semantics Through Gossiping Karl Aberer, Philippe Cudre-Mauroux, Manfred Hauswirth Presented by Yookyung Jo.

Similar presentations


Presentation on theme: "The Chatty Web : Emergent Semantics Through Gossiping Karl Aberer, Philippe Cudre-Mauroux, Manfred Hauswirth Presented by Yookyung Jo."— Presentation transcript:

1 The Chatty Web : Emergent Semantics Through Gossiping Karl Aberer, Philippe Cudre-Mauroux, Manfred Hauswirth Presented by Yookyung Jo

2 Overview :  This paper presents Achieving an effective global semantic agreement starting from local semantic agreements In loosely coupled information sharing systems

3 Overview :  Table of Contents Motivation Problem Definition Model Similarity measures Algorithm Case Study Comments

4 Motivation :  The need for global semantic agreement P2P systems Semantic Web  Their approach global agreement starting from local agreement Self-organizing behavior  Application Scenario Meta-data support for P2P applications federating, loosely-coupled databases

5 Problem Overview :  Loosely coupled information sharing system  Local schema mappings available Missing attributes Possibly erroneous  Mapping graph Transitivity : semantic gossiping Cycle : assess the quality of mappings  End result Routing link assessment => query routing decision Gradual reinforcement of the correct mappings

6 System Overview : Semantic networkSemantic neighbors and Semantic translations

7 Model :  Each peer : a single relational table  Basic relational algebra :  Query :

8 Model  Translation operator :  Query format :  Translated query :

9  Neighbor schema : A_1’’(city), A_2’’(addr), A_3’’(job), …, A_k’’  A_1’(addr), A_2’(title),…,A_j’(phone)…,A_r  Own schema : A_1’(addr), A_2’(title), A_i,A_j’(phone)…, A_m’  A_1, A_2, A_3, …, A_s, …, A_n  A_1, A_3, …, A_x  A_2, A_3, …, A_y Original query translation

10 Query routing  Query routing based on similarity measures “similarity measures”, given q, TT, candidate translation link How promising is it to route a query through the translation link Good semantic agreement => good similarity measures The reverse is not true!  Similarity measures Syntactic similarity  For selection in query  For projection in query Semantic similarity  At the schema level  At the data level

11 Syntactic Similarity (1) :  Not all attributes in ap are preserved

12 Syntactic Similarity (2) :  Not all attributes in as are preserved

13 Semantic Similarity : data level  Semantic agreement => preservation of data dependency  Cycle (in query routing) => Checks the functional dependency  Given an attribute, outgoing translation link

14 Semantic Similarity : schema level  Cycle analysis : what happened to the original attributes of the query?

15 P1 : the likelyhood of the cycles, given translation being correct P2 : the likelyhood of the cycles, given translation being incorrect

16 Gossiping Algorithm :  Upon reception of a query message: Detect any semantic cycles Forward it to the local neighbor, if needed Return potential results  For each outgoing translation links : Apply the translation to the query Update the similarity measures Perform similarity tests forward the query only to the links that pass all similary tests

17 Case Study : T(A->C) : no translation for “title” T(A->D) : erroneous translation for “title” -> “acronym”

18 Case Study : Query = FOR $project IN “project_A.xml”/* RETURN $project/title

19 Case Study : CycleT(A->D) erroneousT(B->D) erroneous A,B,D,E,A+- A,B,D,E,F,A+- A,B,E,A++ A,B,E,F,A++ A,B,F,A++ A,D,E,A-+ A,D,E,B,F,A-+ A,D,E,F,A-+ Syntactic similarity (selection) : not applicable Syntactic similarity (projection) : T(A->C) is not used Semantic similarity (data level) : not applicable (no FDs) Semantic similarity (schema level) : as shown in the table

20 Comments :  Interesting problem framework : Assumption : Deriving a global semantic agreement based on local semantic agreements Solution : Similarity measures based on syntactic, semantic agreement, No global processing  Questions on technical details Similarity measure :  the size of FV does not matter? Semantic similarity at schema level  Serious empirical evaluation over a specific domain is desired

21 Comments :  Related Work OBSERVER, KRAFT, EDUTELLA Schema matching at the local level : GLUE  A perspective : Local relationships exploited to derive the global assessment on their quality  Semantic Interoperability : Important in loosely coupled information sharing system A key issue to the success of Semantic Web


Download ppt "The Chatty Web : Emergent Semantics Through Gossiping Karl Aberer, Philippe Cudre-Mauroux, Manfred Hauswirth Presented by Yookyung Jo."

Similar presentations


Ads by Google