Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Provenance Analysis and RDF Query Processing: W3C PROV for Data Quality and Trust

971 views

Published on

The tutorial will be of interest to: (a) academic researchers who are incorporating provenance metadata in their research for data quality; (b) developers working on scalable platforms for emerging domain applications, such as IOT, LOD, and healthcare and life sciences. In addition to its meaningful breadth, the tutorial will present key technical topics that have seen significant research. These include, provenance modeling, querying and indexing techniques for W3C RDF datasets for provenance querying, and building complex provenance-enabled healthcare informatics platforms. The tutorial will cover the W3C PROV specifications, which are being used to integrate provenance in information systems, including the PROV Data Model (PROV-DM), PROV Ontology (PROV-O), and the PROV constraints.

Published in: Data & Analytics
  • Be the first to comment

Provenance Analysis and RDF Query Processing: W3C PROV for Data Quality and Trust

  1. 1. Provenance Analysis and RDF Query Processing Satya S. Sahoo, Praveen Rao October 12, 2015
  2. 2. Plan for the Tutorial •  09:00 – 10:00: Provenance and its Applications o  What is provenance? o  W3C PROV specifications and applications •  10:00– 10:30: Provenance Query and Analysis o  Provenance queries o  Graph operations to support provenance queries •  10:30 – 11:00: Coffee Break, RBC Gallery, First Floor •  11:00 – 12:15: RDF Query Processing o  Centralized approaches o  Parallel approaches •  12:15 – 12:30: Discussion
  3. 3. Provenance in Application Domains: Healthcare •  Patient treatment often depends on their medical history o  Past hospital visits o  Current and past medications •  Outcome of treatment also depends on patient history •  Medical history of patient - provenance
  4. 4. Provenance in Application Domains: Sensor Networks •  Sensor properties needed for data analysis o  Location of sensor (geo-spatial) o  Temporal information of sensor observations o  Sensor capabilities (e.g., resolution, modality) •  Provenance in sensor networks o  Find all sensors located in geographical location? o  Download data from wind speed sensors for snowstorm * Patni, H., Sahoo, S.S., Henson, C., Sheth, A., “Provenance aware linked sensor data”, Proceedings of the Second Workshop on Trust and Privacy on the Social and Semantic Web, 2010
  5. 5. Provenance in Application Domains: In silico Experiments •  Provenance information helps explain how results from in silico experiments are derived •  Supports scientific reproducibility •  Helps ensure data quality * Zhao, J., Sahoo, S.S., Missier, P., Sheth, A., Goble, C., “Extending semantic provenance into the web of data”, IEEE Internet Computing, 15(1), pp. 40-48. 2011
  6. 6. Research in Provenance Management •  Provenance: derived from the word provenir - “to come from” •  Provenance metadata is specific category of metadata •  The W7 model: who, when, why, where, what, which, how •  Provenance tracking in relational databases o  Result set → + query (constraints) + time value •  Provenance in scientific workflow ID   Name   1   Joe   2   Mary   GeneToPathwayGene Pathway
  7. 7. Provenance and Semantic Web Layer Cake •  Proof layer aka Provenance •  Trust is derived from provenance information
  8. 8. Provenance Management •  Provenance Modeling using Semantic Web technologies •  Provenance models used as input to the W3C PROV Data Model !  Open Provenance Model (OPM) !  Provenir Ontology !  Proof Markup Language (PML) !  Dublin Core !  Provenance Vocabulary
  9. 9. Provenance Management •  Provenance Querying and Access using Semantic Web technologies •  Access provenance of resources on the Web using standard Web protocols (HTTP) •  Two access mechanisms o  Direct access: Dereferencing URIs o  Provenance query service •  Mechanism for content negotiation
  10. 10. W3C PROV Family of Specifications: Provenance Modeling •  W3C Recommendations o  PROV Data Model (PROV- DM) o  PROV Ontology (PROV-O) o  PROV-Constraints o  PROV Notations (PROV- N) •  PROV Working Group Notes (selected) o  PROV-Access and Querying (AQ) o  PROV Dictionary o  PROV XML o  PROV and Dublin Core Mappings (PROV-DC) o  PROV Semantics (using first-order logic) (PROV-SEM)
  11. 11. W3C PROV: PROV Data Model •  Three primary terms •  Entity: A real or imaging thing with fixed aspects •  Activity: occurs over a period of time and acts on entities •  Agent: bears responsibility for activity, entity, or another agent
  12. 12. PROV-DM: Additional Terms •  PROV core terms can be extended to model domain- specific provenance o  Subtyping: programming is a specific type of activity •  PROV allows modeling provenance of provenance •  Bundles: named set of provenance descriptions o  For example, provenance of medical record is important to evaluate its accuracy •  Collections: structured entities o  For example, ranked query results
  13. 13. PROV-DM: Relationships •  Generation: completion of production of an entity •  Usage: beginning of utilization of entity by an activity •  Derivation: transformation of an entity into another entity •  Attribution: ascribing entity to an agent •  Association: assignment of responsibility of an activity to agent •  Delegation: assignment of authority or responsibility to agent … prefix prov: http://www.w3.org/ns/prov# prefix tut: <http://www.iswctutorial.com/> Entity(tut:mapreduceprogram) Activity (tut:programming) wasGeneratedBy(tut:mapreduceprogram, tut:programming, 2015-10-12:09:45) …
  14. 14. A Provenance Graph: Medical History of Patients •  Exercise: Identify subtypes of PROV terms in the graph ClassInstance
  15. 15. PROV Ontology (PROV-O) •  Models the PROV Data Model using OWL2 •  Enables creation of domain-specific provenance ontologies
  16. 16. PROV-O: Qualified Terms •  Qualified terms are used to model ternary relationships using the “Qualification Pattern” •  Uses an intermediate class to represent additional description associated with the relationship •  Additional qualifications: o  Time of generation o  Location
  17. 17. PROV Constraints: Provenance Validation and Inference •  PROV Constraints is used to validate PROV instances using a set of definitions, inferences, and constraints •  Support consistency checking and also reasoning over PROV dataset •  Also allow normalization of PROV data •  For example, o  Uniqueness constraint: If two PROV statement describe the birth of a person twice, the two statements will have same timestamp o  Event ordering constraint: A person cannot be released from hospital before admission
  18. 18. PROV Constraints: Inference •  Support for simple and complex inferences Inference 15: IF actedOnBehalfOf(id; ag2, ag1, _a, attrs) THEN wasInfluencedBy(id; ag2, ag1, attrs) Inference 13: IF wasAttributedTo(_att; e, ag, attrs) THEN wasGeneratedBy(_gen; e, a, _t, attrs) AND wasAssociatedWith(_assoc; a, ag, _p1, []).
  19. 19. Summary of First Session •  We have covered: ! What is provenance? ! Why is provenance important? ! How does it fit into the Semantic Web? ! Which models of provenance can be used by domain applications? ! When to use PROV Entity, Agent, and Process? ! Who delegates authority or responsibility to Agent (PROV-DM Relationships)? ! Where can we apply PROV constraints and inference rules to validate provenance data?
  20. 20. Plan for the Tutorial •  09:00 – 10:00: Provenance and its Applications o  What is provenance? o  W3C PROV specifications and applications •  10:00 – 10:30: Provenance Query and Analysis o  Provenance queries o  Graph operations to support provenance queries •  10:30 – 11:00: Coffee Break, RBC Gallery, First Floor •  11:00 – 12:15: RDF Query Processing o  Centralized approaches o  Parallel approaches •  12:15 – 12:30: Discussion
  21. 21. Provenance Query and Analysis: Data-driven Research Source: http://renewhamilton.ca Source: www.comsoc.org/blog Source: www.nature.com Human Connectome Project PAN-STARRS Project Neptune
  22. 22. Provenance Query and Analysis •  Challenges in data-driven research o  How to reliably store and transfer data between applications, users, or across institutions? o  How to integrate data while ensuring consistency and data quality? o  How to select subsets of data with relevant provenance attributes o  How to rank results of user queries based on provenance values? •  Provenance queries o  Directly query provenance o  Query provenance of provenance
  23. 23. Classification Scheme for Provenance Queries •  Type 1: Querying for Provenance Metadata o Has this patient undergone a heart surgery in the past 1 year? •  Type 2: Querying for Specific Data Set o Find all financial transactions conducted by John Doe in the past 3 years involving amount > $1 million? •  Type 3: Operations on Provenance Metadata o What are the difference in the medical history of two patients – one had better outcome than other? 23  
  24. 24. I. Provenance Trails: Query for Provenance of Entity •  Provenance trails consists of all the provenance related information of an entity o  Hospital admissions of the patient o  Medication information o  Diagnosis information •  Involves graph traversal o  May involve recursive graph traversal o  All provenance information associated with specific hospital admission
  25. 25. II. Query for Entity Satisfying Provenance •  Retrieve all entities that satisfy specific provenance constraints o  Involves identification and extraction of subgraph o  Conforms to the provenance constraints •  May involve multiple SPARQL queries •  Require aggregation of result subgraphs
  26. 26. III. Aggregation or Comparison of Provenance •  Compare the provenance trails of two sensor data entities to identify source of data error •  Provenance graph comparison can be related to subgraph isomorphism used in SPARQL query execution o  Covered in the RDF query processing segment •  A patient’s medical history spans multiple hospital admissions o  Requires aggregation of individual provenance graphs corresponding to hospital admissions
  27. 27. RDF Reification Approach lipoprotein inflammatory_cellsaffects
  28. 28. Provenance Context •  Provenance contextual information defines the interpretation of an entity •  Provenance context is a formal object defined in terms of Provenir ontology lipoprotein inflammatory _cells affects derives_from PubMed_ Source derives_from Entity rdf:type derives_from PROV-O Provenance Context * Sahoo et al., 22nd SSDBM Conference, Heidelberg, Germany, pp. 461-470, Jun 30 - Jul 2, 2010
  29. 29. Provenance Context Entity (PaCE) Approach •  A provenance context is used for entity generation - S, P, O of a RDF triple •  Allows an application to decide the level of provenance granularity Exhaus've  approach  (E_PaCE)   Minimal  approach   (M_PaCE)   Intermediate  approach  (I_PaCE)  
  30. 30. PaCE Inferencing and Evaluation Result •  85 million fewer RDF triples using PaCE Asserted   Inferred   •  Extends existing RDFS entailment •  Condition: Equivalence of provenance context * Sahoo et al., 22nd SSDBM Conference, Heidelberg, Germany, pp. 461-470, Jun  30  -­‐  Jul  2, 2010
  31. 31. Provenance Context Entity (PaCE) Results Query:   List   all   the   RDF   triples   extracted   from  a  given  journal  ar'cle     Query:  List  all  the  journal  ar'cles  from  which  a   given  RDF  triple  was  extracted     Query:  Count  the  number  of  triples  in  each  source   for  the  therapeu'c  use  of  a  given  drug     Query:   Count   the   number   of   journal   ar'cles   published   between   two   dates   for   a   given   triple   * Sahoo et al., 22nd SSDBM Conference, Heidelberg, Germany, pp. 461-470, Jun  30  -­‐  Jul  2, 2010
  32. 32. Time Series Analysis using Provenance Information Query: Count the number of journal articles published over 10 years for a given triple (e.g., thalidomide → treats → multiple myeloma) * Sahoo et al., 22nd SSDBM Conference, Heidelberg, Germany, pp. 461-470, Jun  30  -­‐  Jul  2, 2010
  33. 33. Summary of Second Session •  We covered: ! Provenance queries ! Different categories of provenance queries ! Graph operations in context of provenance queries ! Provenance of RDF triples ! Comparison of Provenance Context Entity approach and RDF Reification approach ! What’s next?
  34. 34. Plan for the Tutorial •  09:00 – 10:00: Provenance and its Applications o  What is provenance? o  W3C PROV specifications and applications •  10:00 – 10:30: Provenance Query and Analysis o  Provenance queries o  Graph operations to support provenance queries •  10:30 – 11:00: Coffee Break, RBC Gallery, First Floor •  11:00 – 12:15: RDF Query Processing o  Centralized approaches o  Parallel approaches •  12:15 – 12:30: Discussion
  35. 35. Semantic Web Layer Cake  
  36. 36. What will we cover? •  Oracle-RDF, SW-Store •  RDF-3X, Hexastore •  BitMat •  DB2RDF •  TripleBit •  RIQ Centralized approaches •  Scalable SPARQL querying •  HadoopRDF •  Trinity.RDF •  H2RDF+ •  TriaD •  DREAM Parallel approaches RDF query processing
  37. 37. Resource Description Framework (RDF) •  Each RDF statement is a (subject, predicate, object) triple o  Represents an assertion or a fact <http://xmlns.com/foaf/0.1/Alice> <http://xmlns.com/foaf/0.1/name> “Alice”
  38. 38. RDF Quadruples (Quads) •  A quad is denoted by (subject, predicate, object, context) o  Context (a.k.a. graph name) can be used to capture provenance information (e.g., origin/source of a statement) o  Triples with the same context belong to the same RDF graph @prefix foaf: <http://xmlns.com/foaf/0.1/> foaf:Alice foaf:name “Alice” <http://ex.org/John/foaf.rdf> . foaf:Bob foaf:name “Bob” <http://ex.org/John/foaf.rdf> . foaf:Alice foaf:knows foaf:Bob <http://ex.org/graphs/John> . foaf:Alice foaf:knows foaf:Bob <http://ex.org/graphs/Mary> .
  39. 39. SPARQL Query PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#> PREFIX foaf: <http://xmlns.com/foaf/0.1/> PREFIX movie: <http://data.linkedmdb.org/resource/movie/> SELECT ?g ?producer ?name ?label ?page ?film WHERE { GRAPH ?g { ?producer movie:producer_name ?name . ?producer rdfs:label ?label . OPTIONAL { ?producer foaf:page ?page . } ?film movie:producer ?producer . }} Basic Graph Pattern (BGP) matching Triple pattern
  40. 40. Open-Source and Commercial Tools Sesame, Apache Jena, 3store, Mulgara, Kowari, YARS2, … Virtuoso, AllegroGraph, Garlik 4store/ 5store, … SYSTAP’s Blazegraph, Stardog, Oracle 12c, Titan, Neo4j, MarkLogic, Ontotext’s GraphDB
  41. 41. Reported Large-scale Deployments 1+ trillion triples Oracle 12c 8 database nodes (192 cores) and 14 storage nodes (168 cores), 2 TB total RAM and 44.8 TB Flash Cache AllegroGraph 240 core Intel x5650, 2.66GHz, 1.28 TB RAM 10+ billion triples OpenLink Virtuoso (15+ Billion) 8-node cluster, two quad-core processors per node, 16 GB RAM Ontotext’s GraphDB (13 Billion) Dual-CPU server with Xeon E5-2690 CPUs, 512 GB of RAM and SSD storage array Stardog (50 Billion) Single server, 32 cores, 256 GB RAM Blazegraph (50 Billion) Single server, GPU-acceleration Source: http://www.w3.org/wiki/LargeTripleStores
  42. 42. Triples Table, Vertical Partitioning •  SQL-based RDF querying scheme [Chong et.al. VLDB ‘05] o  IDTriples table, URIMap table; use of self-joins; subject- property matrix •  SW-Store [Abadi et.al., VLDB ’07, VLDBJ ‘09] o  Vertical partitioning of RDF data •  Triples with the same property are grouped together: (S,O) o  Use of a column-store; materialization of frequent joins •  MonetDB/SQL [Sidirourgos et.al., PVLDB ’08] o  Triplestore on a row-store vs vertical partitioning on column- store D. J. Abadi, A. Marcus, S. R. Madden, K. Hollenbach, “Scalable Semantic Web Data Management Using Vertical Partitioning,” in Proc. of the 33rd VLDB Conference, 2007, pp. 411-422. L. Sidirourgos, R. Goncalves, M. Kersten, N. Nes, S. Manegold, “Column-store support for RDF data management: not all swans are white,” in PVLDB, 1(2), 2008. E. I. Chong, S. Das, G. Eadon, J. Srinivasan, “An efficient SQL-based RDF querying scheme,” in Proc. of the 31st VLDB Conference, 2005, pp. 1216-1227.
  43. 43. Exhaustive Indexing •  Early approaches o  Kowari [Wood et.al., XTech ‘05], YARS [Harth et.al., LA-WEB ‘05] •  RDF-3X [Neuman et.al., PVLDB ‘08, VLDBJ ‘10] o  6 permutations: (SPO), (SOP), (POS), (PSO), (OSP), (OPS) o  Clustered B+-tree indexes; leverages merge joins; compression o  New join ordering method using a cost model based on selectivity estimates •  Hexastore also builds similar indexes [Weiss et.al., PVLDB ‘08] o  Merge joins; no compression T. Neumann, G. Weikum, “RDF-3X: a RISC-style engine for RDF,” in Proc. of the VLDB Endowment 1 (1) (2008), pp. 647-659. C. Weiss, P. Karras, A. Bernstein, “Hexastore: Sextuple indexing for Semantic Web data management,” in Proc. VLDB Endow. 1 (1) (2008), pp. 1008-1019.
  44. 44. Reducing the Cost of Join Processing •  BitMat [Atre et.al., WWW ‘10] o  A triple is uniquely mapped to a cell in a 3D cube o  Compressed bit matrices are loaded and processed in memory during join processing •  Intermediate join results are not materialized •  DB2RDF [Bornea et.al., SIGMOD ‘13] o  Direct Primary Hash, Reverse Primary Hash •  Wide table layout to reduce joins for star-shaped queries •  Only subject and object indexes o  SPARQL-to-SQL translation M. A. Bornea, J. Dolby, A. Kementsietsidis, K. Srinivas, P. Dantressangle, O. Udrea, B. Bhattacharjee, “Building an efficient RDF store over a relational database,” in Proc. of 2013 SIGMOD Conference, 2013, pp. 121-132. M. Atre, V. Chaoji, M. J. Zaki, J. A. Hendler, “Matrix "Bit" loaded: A scalable lightweight join query processor for RDF data,” in Proc. of the 19th WWW Conference, 2010, pp. 41-50.
  45. 45. Reducing the Cost of Join Processing •  TripleBit [Yuan et.al., PVLDB ‘13] o  Represents triples as a 2D bit matrix called Triple Matrix •  Compression for compactness o  For each predicate •  SO and OS ordered buckets of triples •  Conceptually, only two indexes are needed instead of six: POS, PSO o  Reduction in the size of intermediate results during join processing P. Yuan, P. Liu, B. Wu, H. Jin, W. Zhang, L. Liu, “TripleBit: A fast and compact system for large scale RDF data,” in Proc. VLDB Endow. 6 (7) (2013), pp. 517-528.
  46. 46. Join Processing on Large, Complex BGPs Too many join operations "
  47. 47. RIQ •  Fast processing of SPARQL queries on RDF quads: (S,P,O,C) Decrease-and-conquer V. Slavov, A. Katib, P. Rao, S. Paturi, D. Barenkala, “Fast Processing of SPARQL Queries on RDF Quadruples,” in Proc. of the 17th International Workshop on the Web and Databases (WebDB 2014), Snowbird, UT, 2014.
  48. 48. RIQ’s Architecture
  49. 49. Performance Comparison: Single Large, Complex BGP BTC 20121 ~ 1.4 billion quads LUBM ~ 1.4 billion RDF statements 1http://challenge.semanticweb.org Y. Guo, Z. Pan, J. Heflin, “LUBM: A benchmark for OWL knowledge base systems,” Web Semantics: Science, Services and Agents on the World Wide Web 3 (2005) 158–182.
  50. 50. Performance Comparison: Multiple BGPs BTC 20121 ~ 1.4 billion quads
  51. 51. Parallel RDF Query Processing in a Cluster •  Early approaches o  YARS2 [Harth et.al., ISWC/ASWC ’07], SHARD [Rohloff et.al., PSI EtA ‘10], Virtuoso1 •  Hash partition triples across multiple machines •  Parallel access during query processing o  Work well for simple index lookup queries o  For complex SPARQL queries, need to ship data during query processing K. Rohloff and R. Schantz, “High-performance, massively scalable distributed systems using the MapReduce software framework: The SHARD triple-store.” International Workshop on Programming Support Innovations for Emerging Distributed Applications, 2010. 1OpenLink Software. Towards Web-Scale RDF. http://virtuoso.openlinksw.com/dataspace/dav/wiki/Main/VOSArticleWebScaleRDF. A. Harth, J. Umbrich, A. Hogan, S. Decker, “YARS2: A Federated Repository for Querying Graph Structured Data from the Web,” in Proc. of ISWC'07/ASWC'07, pp. 211-224, 2007.
  52. 52. Scalable SPARQL Querying •  Vertex partitioning using METIS1 •  Triples in each partition are placed together on a machine o  Replication of triples on the partition boundaries •  n-hop guarantee o  PWOC queries •  No data shuffling between machines o  Uses RDF-3X on each machine •  Uses Hadoop for certain tasks o  E.g., data partitioning, communication during query processing J. Huang, D. J. Abadi, K. Ren, “Scalable SPARQL querying of large RDF graphs,” in Proc. of VLDB Endow. 4 (11) (2011), pp. 1123-1134. 1METIS. http://glaros.dtc.umn.edu/gkhome/views/metis
  53. 53. HadoopRDF •  Split triples by predicate •  For rdf:type, split by distinct objects •  Store the splits as HDFS files •  MapReduce-based join processing to process SPARQL queries o  Heuristics-based cost model M. Husain, J. McGlothlin, M. Masud, L. Khan, B. Thuraisingham, “Heuristics-Based Query Processing for Large RDF Graphs Using Cloud Computing,” in IEEE Transactions on Knowledge and Data Engineering 23(9), pp. 1312-1327 (2011).
  54. 54. Trinity.RDF •  Uses a distributed in-memory key-value store o  Hashing on vertex-ids, random partitioning on machines o  RDF graphs are stored natively using key-value pairs •  Parallel graph exploration, optimized exploration o  Lower communication cost o  Reduction in the size of intermediate results K. Zeng, J. Yang, H. Wang, B. Shao, Z. Wang, “A distributed graph engine for Web Scale RDF data,” Proc. VLDB Endow. 6 (4) (2013), pp. 265-276. 2models (vertex-id, <in-adjacency-list, out-adjacency-list>) (vertex-id, <in1, …, ink, out1, …, outk>), (ini, <adjacency-listi>), (outi, <adjacency-listi>) Adjacency list is partitioned on i machines
  55. 55. H2RDF+ •  Uses HBase to build indexes on triples o  6 permutations of (SPO) •  Triples are stored as rowkeys o  Aggressive compression •  MapReduce-based multi-way merge and sort-merge joins o  Sort-merge join is used when joining unordered intermediate results N. Papailiou, D. Tsoumakos, I. Konstantinou, P. Karras, N. Koziris, “H2RDF+: An Efficient Data Management System for Big RDF Graphs,” in Proc. of the 2014 SIGMOD Conference, Snowbird, Utah, 2014, pp. 909-912. N. Papailiou, I. Konstantinou, D. Tsoumakos, P. Karras, N. Koziris, “H2RDF+: High-performance Distributed Joins over Large- scale RDF Graphs,” in Proc. of the IEEE International Conference on Big Data, 2013.
  56. 56. TriAD •  Master node o  Global summary graph – concise summary of RDF data •  Graph partitioning; a supernode per partition •  Worker/slave nodes o  Locality-based sharding •  Triples belonging to a supernode are stored on the same horizontal partition o  Local indexes – 6 permutations of (SPO) •  Query processing o  Use the summary graph for join-ahead pruning o  Distributed query execution via asynchronous inter-node communication (MPICH2) S. Gurajada, S. Seufert, I. Miliaraki, M. Theobald, “TriAD: A Distributed Shared-nothing RDF Engine Based on Asynchronous Asynchronous Message Passing,” in Proc. of the 2014 SIGMOD Conference, Snowbird, Utah, 2014, pp. 289-300.
  57. 57. DREAM •  RDF data is not partitioned across different machines o  Each machine stores the entire RDF data •  Adaptive query planner o  Partitions a query graph into sub-queries o  Sub-queries are executed in parallel on M (≥1) machines o  No data shuffling •  Machines exchange auxiliary data (e.g., ids of triples) for joining intermediate data and producing the final result M. Hammoud, D. A. Rabbou, R. Nouri, S.M.R. Beheshti, S. Sakr, “DREAM: Distributed RDF Engine with Adaptive Query Planner and Minimal Communication,” in Proc. VLDB Endow. 8 (6) (2015), pp. 654-665.
  58. 58. What did we cover? •  Oracle-RDF, SW-Store •  RDF-3X, Hexastore •  BitMat •  DB2RDF •  TripleBit •  RIQ Centralized approaches •  Scalable SPARQL querying •  HadoopRDF •  Trinity.RDF •  H2RDF+ •  TriaD •  DREAM Parallel approaches RDF query processing
  59. 59. Open Challenges in Provenance •  Large scale storage of provenance o  Limited work in real world provenance management for Big Data applications •  Standardization of provenance query APIs •  Integration of provenance analysis with RDF query processing systems •  Efficient provenance analysis using state of the art approaches in SPARQL query execution •  Visualization of provenance data
  60. 60. Acknowledgement •  Tutorial Website o  https://sites.google.com/site/provenancetutorial/ •  Acknowledgements o  National Science Foundation (NSF) Grant No. 1115871 o  National Institutes of Health (NIH) Grant No. 1U01EB020955-01 •  Contact o  Satya Sahoo, satya.sahoo@case.edu o  Praveen Rao, raopr@umkc.edu

×