Upload
others
View
18
Download
0
Embed Size (px)
Citation preview
RISE SICS, Electrum Kista Stockholm, Sweden
RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Towards Ontology Based Event Processing
1
R. Tommasini - Politecnico di Milano
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected] !2
PhD Student @ Politecnico di Milano
Research Interests:
-Semantic Web & Reasoning
-Stream Processing
-Programming Languages
-Distributed Systems
ME
@rictomm
rictomm.me
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected] !3
- Assistant Professor at DEIBPolitecnico di Milano
- Expert in semantic technologies and stream computing
- Brander of stream reasoning
- 17 years of experience in research and innovation projects
- Startupper: http://www.fluxedo.com
My Advisor
@manudellavalle
h2p://emanueledellavalle.org
h2p://streamreasoning.org
h2p://fluxedo.com
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
What is Stream Reasoning?
!4
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Can we detect fire?
!5
*Expected Answer: YES
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Can we (actually) detect fire?
!6
Expected Reaction: Perplexed Audience
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected] !7
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected] !8
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected] !9
100°
70°
20°
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected] !10
100°
70°
20°
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected] !11
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected] !12
70%
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected] !13
30%
OWLED 16RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
DISCANSmoke Detection
Humidity Variations (decreases)
Temperature Variations (increases)
!14
SummaryWorkarounds
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
This is Stream Reasoning!
!15
RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Structural Heterogeneity
!16
RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected] !17
Semantic Heterogeneity
!18IBM 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Incomplete
!19IBM 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Vast
!20IBM 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Noisy
!21IBM 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Complex Domain
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected] !22
Can we make sense in real-time of heterogeneous, vast, incomplete, and inevitably noisy and data streams in
order to support the decision
processes of extremely large numbers
of concurrent users?
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected] !23
Requirement Analysis
x
x
x
x x
x
x
x x
x
Volume
Velocity
Variety
Veracity
•handle massive datasets
•process data streams
•cope with heterogeneous data
•cope with incomplete data
•cope with noisy data
•provide reactive answers
•access fine-grained information
•model complex domains
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected] !24
Stream Processing vs Semantic Technologies
Requirement SP STmassive datasetsdata streamsheterogeneous dataset incomplete datanoisy data reactive answers fine-grained information access complex domain models
✓✓
✓✓
✓✓✓ ✓
✓
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Stream Reasoning
!25
Stuckenschmidt, H., Ceri, S., Della Valle, E., & Van Harmelen, F. (2010). Towards expressive stream reasoning
OWLED 16RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
DISCAN!26
Cascading Reasoning
Stuckenschmidt, H., Ceri, S., Della Valle, E., & Van Harmelen, F. (2010). Towards expressive stream reasoning
DL
LOGIC PROGRAMMING
RDF STREAM PROCESSING
RAW STREAM PROCESSING
Matching
Selection
Interpretation
Reasoning
Querying
Rewriting
PTime
2NEXPTime
104Hz
1 Hz
Cha
nge
Freq
uenc
y
Com
plex
ity
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
RDF Stream Processing (RSP)
!27
Continuous Data Integration
OWLED 16RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
DISCAN-An RDF Stream is an partially ordered sequence of pairs (Gi,ti) where
-Gi, is a [named] RDF graph and
-ti is a timestamp.
!28
RDF Streams
OWLED 16RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
DISCAN( { :s1 :observes :o1 ; :o1 :value 20C }, 1)
( { :s1 :observes :o2 ; :o2 :value 20C }, 2)
( { :s1 :observes :o3 ; :o3 :value 30C }, 3)
( { :s1 :observes :o4 ; :o4 :value 50C }, 4)
!29
An Example
OWLED 16RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
DISCAN- A Reference Model fo Continuous SPARQL
- Extends CQL to process RDF Graphs
- Introduces the notions of Window and Event Pattern
!30
RSEP-QL
OWLED 16RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
DISCANREGISTER STREAM <fire>
CONSTRUCT { ?o a :FireObservation ; :sensedBy ?s .}
FROM NAMED WINDOW <w1> [RANGE 5m,STEP 5m] ON STREAM <temp>
WHERE { WINDOW <w1> {
?s :observes ?o ; ?o :value ?t
FILTER (?t > 50C) }}
!31
An Example
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Continuous Reasoning
!32
Deductive
OWLED 16RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
DISCAN-An Ontology Stream is an partially ordered sequence of pairs (Ai,ti) where
-Ai, is a set of a ABox axioms w.r.t. a static TBox T.
-ti is a timestamp.
!33
Ontology Streams
OWLED 16RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
DISCAN-An Windowed Ontology Stream S[o,c] is the union
of all the Abox axioms Sets Ai with o<i<c
-Continuous Reasoning can be reduced to traditional ontological reasoning over a windowed ontology stream
!34
Windowed Ontology Streams
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Ontology Based Event Processing
!35
Joint work with P.Bonte, E. Mannens, F. De Turck, F. Ongenae
OWLED 16RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
DISCAN!36
Cascading Reasoning Approach
Stuckenschmidt, H., Ceri, S., Della Valle, E., & Van Harmelen, F. (2010). Towards expressive stream reasoning
CEP
DL
RDF STREAM PROCESSING
RAW STREAM PROCESSING
Matching
Selection
Interpretation
Reasoning
Querying
Rewriting
PTime
2NEXPTime
104Hz
1 Hz
Cha
nge
Freq
uenc
y
Com
plex
ity
RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Data Integration
We assume RDF Stream as common data model
!37
Time
(Gi, ti)
RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Events!
first-class objects in the language
!38
`Physical Event
Logical Event
DL Reasoning
RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Logical Modeling
Logical Event
Specifications
EVENT OfficeTemperaturEvent
subClassOf TemperaturEvent
and (observationResult some (hasValue >= 40)) and
(hasLocation some Office)
!39
RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Semantic Complex Event Processing
Patterns
EVENT FireEvent {
MATCH TemperaturEvent
SEQ SmokeDetectionEvent
WITHIN (5m) }
!40
RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Semantic Complex Event Processing
In OBEP
EVENT FireEvent {
MATCH TemperaturEvent
SEQ SmokeDetectionEvent WITHIN (5m)
IF {
EVENT TemperaturEvent {?loc0 hasValue ?v}
EVENT SmokeDetectionEvent {?loc1 hasValue ?v
FILTER (?smokeLevel == 3) }}
!41
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Future Works
!42
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Ontology Based Streaming Data Access
!43
OWLED 16RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
DISCAN!44
Cascading Reasoning
Stuckenschmidt, H., Ceri, S., Della Valle, E., & Van Harmelen, F. (2010). Towards expressive stream reasoning
CEP
DL
RDF STREAM PROCESSING
RAW STREAM PROCESSING
Matching
Selection
Interpretation
Reasoning
Querying
Rewriting
PTime
2NEXPTime
104Hz
1 Hz
Cha
nge
Freq
uenc
y
Com
plex
ity
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
RDF STREAM PROCESSING
RAW STREAM PROCESSING
!45
Rewriting and Interpreting
- including continuous semantics will enable continuous querying over
virtual streaming sources;
- including time operators like windows will enable query rewriting into continuous query languages
DISCANRISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Stream Reasoning Applications
!46
RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
WASP
Web Stream Processing
Application
Anatomy of a Streaming Application
- Input Streams
- Output Streams
- Continuous Tasks
!47
RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
The Web is Streaming
!48
OWLED 16RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
DISCAN- VOCALS allows to describe streams and streaming endpoints in a machine readable form
- VOCALS enables stream services description, fostering interoperability between producers and consumers.
- VOCALS let track stream transformation provenance describing the continuous tasks operating on streams.
!49
VoCaLS - Vocabulary and Catalog for Linked Streams
RISE SICS 2018 - Riccardo Tommasini - @rictomm - rictomm.me - [email protected]
Questions?Email: [email protected] Twitter: @rictomm Github: riccardotommasini Web1: riccardotommasini.com Web2: streamreasoning.org
!51