An Extension of SPARQL with Fuzzy Navigational Capabilities for Querying Fuzzy RDF Data Olivier Pivert, Olfa Slama, Virginie Thion

Transcription

1 An Extension of SPARQL with Fuzzy Navigational Capabilities for Querying Fuzzy RDF Data Olivier Pivert, Olfa Slama, Virginie Thion 2016 IEEE International Conference on Fuzzy Systems JULY 2016, VANCOUVER, CANADA

2 Outline Motivations 1 Motivations 2 The RDF data model The F-RDF data model SPARQL 3 4 Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 2 / 21

3 Motivations Retrieve artists such that the albums he/she recommends are low-rated and created by a close friend. Previous extensions of SPARQL with Boolean navigational capabilities (e.g., SPARQLeR, SPARQ2L, PSPARQL, nsparql, SPARQL 1.1); with fuzzy querying capabilities, on node attribute values only (e.g. fsparql); but not both, and no fuzzy querying on the structure of the graph. Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 3 / 21

4 Goal: extension of the SPARQL Query Language 1 query a fuzzy RDF model containing gradual information better model real-world concepts. 2 express fuzzy user preferences queries better reflect the intention of the user; rank-orders the retrieved items; provides a relaxed version of the query. Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 4 / 21

5 Outline Motivations The RDF data model The F-RDF data model SPARQL 1 Motivations 2 The RDF data model The F-RDF data model SPARQL 3 4 Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 5 / 21

6 The RDF data model The F-RDF data model SPARQL The RDF data model RDF (Resource Description Framework, W3C) RDF is a model for representing (linked) data on the web. s, p, o : the subject s has a property p with a value o. Mariah Subject creator Predicate Butterfly Object RDF supports only Boolean relationships (True or False). Beyonce friend Shakira Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 6 / 21

7 The RDF data model The F-RDF data model SPARQL The F-RDF data model [Mazzieri and Dragoni, 2008], [Vaneková et al., 2005], [Straccia, 2009], [Lv et al., 2008] F-RDF data model supports gradual relationships friend (close, best, normal,...), recommends (strongly, weakly,...). Beyonce friend (0.8) Shakira F-RDF graph is a tuple (T, ζ): T is a finite set of triples of (U B) U (U L B), ζ is a membership function on triples ζ : T [0, 1]. Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 7 / 21

8 friend (0.7) JustinT Shakira friend (0.4) friend(0.5) creator friend(0.6) friend(0.5) friend(0.3) Justified EnriqueI friend(0.3) rating recommends(0.6) 6 creator recommends(0.7) friend(0.8) 9 rating Euphoria Beyonce recommends (0.8) friend(0.7) recommends(0.8) rating 4 MariahC creator Butterfly friend(0.4) recommends(0.4) Figure: Fuzzy RDF graph inspired by MusicBrainz 1 1

9 The RDF data model The F-RDF data model SPARQL SPARQL (SPARQL Protocol and RDF Query Language, W3C) Query language for querying RDF data based on graph pattern matching. Basic graph pattern (BGP): set of triples patterns containing variables.?artist creator?album?album has rating 6?Artist 6 has rating creator?album Alternative (UNION), optional (OPTIONAL), filter (FILTER) GP. Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 9 / 21

10 The RDF data model The F-RDF data model SPARQL Example SELECT?art1 WHERE {?art1 friend+?art2?art2 creator?alb.?alb rating?r.?art1 recommends?alb. FILTER (?r < 4) } A SPARQL query contains only: Boolean conditions (e.g.?r < 4), Boolean regular expressions (SPARQL 1.1) (e.g. friend + ). Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 10 / 21

11 Outline Motivations 1 Motivations 2 The RDF data model The F-RDF data model SPARQL 3 4 Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 11 / 21

12 Contribution: More flexible SPARQL Express fuzzy user preferences queries to F-RDF graph on the values of the nodes, on the structure of the graph. on the values of the nodes : fuzzy conditions (e.g. low rating) low: fuzzy term based on fuzzy set theory [Zadeh, 1965] degree µ low rating Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 12 / 21

13 Contribution: More flexible SPARQL on the structure of the graph (fuzzy regular expressions Ϝ) F ::= ɛ, υ, F F, F.F, F, F+, F cond F cond : paths satisfying the pattern F with a condition cond (F distance is short ) We limit the path fuzzy structural properties to ST (strength) and distance. distance(x, y) = min p Paths(x,y) length(p) the length of a path p in a fuzzy graph: length(p) = t p 1 ζ(t) ST (x, y) = max p Paths(x,y) ST _path(p) the strength of the path p in a fuzzy graph: ST _path(p) = min({ζ(t) t p}). Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 13 / 21

14 Contribution: More flexible SPARQL A formal redefinition of the syntax and the semantics of the SPARQL graph pattern introduced in [Pérez et al., 2008]. Fuzzy graph pattern: allows to express fuzzy preferences into SPARQL. Syntax of a fuzzy graph pattern is a fuzzy triple (U V) (U F V) (U L V). (P 1 AND P 2 ), (P 1 UNION P 2 ), (P 1 OPT P 2 ) and (P 1 FILTER C) are fuzzy graph patterns. A fuzzy condition C is a logical combination of fuzzy terms: bound(?x),?x θ c and?x θ?y, where θ is a fuzzy or crisp comparator,?x IS Fterm (?age IS young), ( C 1 ) and (C 1 C 2 ), where is a fuzzy connective. Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 14 / 21

15 FURQL (FUzzy RDF Query Language) Fuzzy extension of the SPARQL Query Language: the occurrence of fuzzy graph patterns in the W H E R E clause, the occurrence of fuzzy conditions in the F I L T E R clause. Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 15 / 21

16 FURQL (FUzzy RDF Query Language) Example: Retrieve artists (?art1) such that the albums (?alb) he/she recommends are low-rated and created by a close friend (?art2). SELECT?art1 WHERE {?art1 (friend+ distance IS short)?art2.?art2 creator?alb.?alb rating?r.?art1 recommends?alb. FILTER (?r IS low) } CUT 0.3 (friend + ) distance IS short.creator rating?art1?alb?r recommends low Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 16 / 21

17 Implementation Storage of Fuzzy RDF graphs: using the reification mecanism Idea: attach fuzzy degrees to triples. Shakira friend(0.7) MariahC Blank node Shakira subject predicate Reification friend object degree MariahC 0.7 Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 17 / 21

18 Implementation Adding a software (called FURQL) layer over a standard and possibly distant classical SPARQL engine (endpoint) FURQL query Q Client (user) Answers of Q Compiling SPARQL query (Q crisp ) Software add-on layer Qualitative cut function Sat. degrees function Cutting and Ranking Fuzzy treatment Answers of Q crisp Classical SPARQL querying SPARQL query evaluator engine Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 18 / 21

19 Outline Motivations 1 Motivations 2 The RDF data model The F-RDF data model SPARQL 3 4 Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 19 / 21

20 To sum up Definition of a fuzzy extension of the SPARQL Query Language: deal with fuzzy RDF data, express fuzzy structural conditions and fuzzy conditions on the values of the nodes. Definition of a language ( FURQL) based on the notion of fuzzy graph pattern. Future Work Implementing a prototype with real-world RDF databases. Extending the language with more sophisticated fuzzy conditions (fuzzy quantifiers...). Introducing some quality-related metadata (about freshness, reliability and so on) to this framework. Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 20 / 21

21 Thank you for your attention Questions?

22 Lv, Y., Ma, Z. M., and Yan, L. (2008). Fuzzy RDF: A data model to represent fuzzy metadata. In Proc. of FUZZ-IEEE, pages IEEE. Mazzieri, M. and Dragoni, A. F. (2008). A fuzzy semantics for the resource description framework. Springer. Pérez, J., Arenas, M., and Gutierrez, C. (2008). nsparql: A navigational language for RDF. In Proc. of ISWC. Straccia, U. (2009). A minimal deductive system for general fuzzy RDF. In Web Reasoning and Rule Syst., pages Springer. Vaneková, V., Bella, J., Gurskỳ, P., and Horváth, T. (2005). Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 21 / 21

23 Fuzzy rdf in the semantic web: Deduction and induction. In Proceedings of Workshop on Data Analysis (WDA 2005), pages Olivier Pivert, Olfa Slama, Virginie Thion SPARQL with Fuzzy Navigational Capabilities 21 / 21