Tackling variety in event based systems

Tutorial: Tackling Variety in Event-
Based Systems
Souleiman Hasan, Edward Curry
DEBS 2015

About Us
Souleiman Hasan
PhD Researcher
Insight @ NUI Galway
Interests:
Semantic Event Processing
Internet of things
Semantic Web
souleiman.hasan@insight-centre.org
http://www.souleimanhasan.org/
29 June- 03 July 2015, Oslo, NorwayDEBS ’15 2
Dr. Edward Curry
Unit Leader
Insight @ NUI Galway
Interests:
Event Systems
Energy Intelligence
Internet of things
ed.curry@insight-centre.org
http://edwardcurry.org/

Overview
13:40-14:00 Part I: Events in the IoT
14:00-14:20 Part II: Computational Paradigms
14:20-14:40 Part III: A Theory for Event Exchange
14:40-15:10 Part IV: Semantics and Approximation
15:10-15:30 Break
15:00-16:00 Part V: Approaches to Semantic Coupling
16:00-16:20 Part VI: Thematic Event Processing
16:20-16:40 Part VII: Building IoT Event Systems
16:40-17:00 Part VIII: Future Research Challenges

Events in the IoT
PART I

Current Trends
Smart Homes, Cities, Internet of Things, Big
Data
By 2020 50 billion devices
connected to mobile networks
(OECD, 2012)
OECD, 2012. Machine-to-Machine Communications: Connecting Billions of Devices. OECD Digital Economy Papers, No. 192.

From Internet of Things to Internet of
Everything

From Rigid Schemas to Schema-less
Heterogeneous, complex and large-scale data
Very-large and dynamic “schemas”
Open Environments: distributed, decoupled data sources,
anonymous users, multi-domain, lack of global order of
information flow
10s-100s attributes
1,000s-1,000,000s attributes
circa 2000
circa 2014Slide Credits: Andre Freitas

Fundamental Decentralization
Multiple perspectives (conceptualizations) of the reality.
Ambiguity, vagueness, inconsistency.
Slide Credits: Andre Freitas

Shift in Data Production/Consumption
Number of Information Sources: increases
mobile subscribers 2,205 millions in 2005 to 6,662 millions in 2013, 200% (ITU, 2014)
Data Heterogeneity: increases
Falcons discovers 4,000 ontologies 2008 to 6,400 in 2015 (Cheng et al., 2008)
Number of non-technical Users: increases
Internet users 1,024 millions in 2005, to 2,710 millions in 2013, 160% (ITU, 2014)
USA 3% in STEM disciplines in 2010 (NSF, 2010)
Organization of Users: distributed & decoupled
Wikipedia crowdsourcers global diversity (Ross et al., 2010)
Timeliness: required
important to filter important data items as early as possible (Jagadish at al., 2014)
Information Completeness: Uncertain
Uncertainty, errors, and missing values are endemic, and must be managed (Jagadish at al., 2014)

Current Trends
Small scale,
controlled
environments
Large scale, open environments
Information
sources
10s to 100s 1000s to millions
Data
heterogeneity
Small number of
schemas
High number of schemas
Users Small number
Know the environment
Large number
Not quite know the environment
Users
organization
Users know each others
Top-down hierarchies
(e.g. enterprises)
Decoupled and distributed
Dynamism Low High (sources and users join and
leave often)
Domain Domain specific Users interest range from domain
specific to domain agnostic

INTERNET OF THINGS (IOT)
CASE STUDY

Internet of Things
Sensing and Communication Layer
(RFID, 6LoWPAN, IPv6,…)
Middleware Layer
(MoM, SOA, …)
Applications Layer
(logistics, healthcare, smart envs, analytics,
robo-taxis, virtual reality, …)
ATZORI, LUIGI, ANTONIO IERA, AND GIACOMO MORABITO. "THE INTERNET OF THINGS: A SURVEY." COMPUTER
NETWORKS 54.15 (2010): 2787-2805.

“SmartSantander proposes a unique way in the
world city-scale experimental research facility
in support of typical applications and services
for a smart city”
http://www.smartsantander.eu/
Smart City- How Real?

IoT in Light of Big Data
• Significant efforts in IoT come from Sensing and
Communication communities
• Challenges of the IoT will be more prevalent at the data
level (Aggarwal et al., 2013)

IoT in Light of Big Data
• Datasets so large and complex to be processed
by current data processing applications.
• Volume
 Terabytes, Petabytes…
• Variety
 Many sources, syntax and semantics.
• Velocity
 Near real-time, real-time
Stonebraker, Michael. "What Does' Big Data'Mean." Communications of the ACM, BLOG@ ACM (2012).

Volume
• Distributed and parallel processing
• Hadoop, MapReduce, …
• SPARK
Example: Word count in Spark
file = spark.textFile("hdfs://...")
file.flatMap(lambda line: line.split())
.map(lambda word: (word, 1))
.reduceByKey(lambda a, b: a+b)

Velocity
• Stream and event processing
• Lambda architecture’s speed layer
• SPARK Streaming
• Storm

Variety
• Mainly heterogeneous data models
• Heterogeneous schema in DB terminology
• Currently targeted by
 ETL
 Data Integration
 Semantic Web technologies
• Example
 SSN Ontology

DISCUSSION POINT
• What current paradigms do you think will
struggle with IoT?

Why Do We Really Have Variety?
• Symbolization process in computing systems
 Easier to process symbols by computers
• We use symbols to reduce meanings
• Mostly words
• E.g. Unique Name Assumption in DB
community:
 “Different things and concepts are denoted using
different names”
 California and CA denote different things according to
this assumption
 The whole processing model is symbolic

Why Do We Really Have Variety?
• Same assumption is usually made in event
systems
 Data models in Middleware is symbolic
• Decoupled, distributed parties symbolize
differently!
Variety Happens!

COMPUTATIONAL PARADIGMS
PART II

Information Flow Processing (IFP)
• Users need to collect information
 Produced by multiple distributed sources
 For timely way processing
 To extract knowledge asap
CUGOLA, G. AND MARGARA, A., 2011. Processing flows of information: From data stream to complex
event processing. ACM Computing Surveys Journal.
Financial Continuous
Analytics
RFID Inventory
Management
Environmental
Monitoring

• Processing information as it flows
 No intermediate storage
 New information produced
 Raw information can be discarded
Information
Flow
Processing
Engine
Producers Consumers
Rule managers

• Requirements
 Real-time or near real-time processing
 Expressive language for rules
 Scalability to large number of producers and consumers

Active Databases
• Traditional database systems
 Passive
 Store data and wait for user’s interaction
 Reactive behaviour in the application layer
DAYAL, U., BLAUSTEIN, B., BUCHMANN, A., CHAKRAVARTHY, U., HSU, M., LEDIN, R., MCCARTHY,
D., ROSENTHAL, A., SARIN, S., CAREY, M. J., LIVNY, M., AND JAUHARI, R. 1988. The hipac project:
Combining active databases and timing constraints. SIGMOD Rec. 17, 1, 51–70.
LIEUWEN, D. F., GEHANI, N. H., AND ARLEIN, R. M. 1996. The ode active database: Trigger semantics
and implementation. In Proceedings of the 12th International Conference on Data Engineering
(ICDE’96). IEEE Computer Society, Los Alamitos, CA, 412–420.
GATZIU, S. AND DITTRICH, K. 1993. Events in an active object-oriented database system. In
Proceedings of the International Workshop on Rules in Database Systems (RIDS), N. Paton and H.
Williams, Eds. Workshops in Computing, Springer-Verlag, Edinburgh, U.K.
CHAKRAVARTHY, S. AND ADAIKKALAVAN, R. 2008. Events and streams: Harnessing and unleashing
their synergy! In Proceedings of the 2nd International Conference on Distributed Event-Based Systems
(DEBS’08). ACM, New York, NY, 1–12.

Active Databases
• Reactive behaviour moved to database layer
• Event-Condition-Action (ECA) rules
 Event: source. E.g. tuple inserted
 Condition: post event. E.g. inserted.value > 5
 Action: what to do. E.g. modify the DB
• Cons
 Persistent storage model
 Suitable when updates not frequent and few rules

Data Stream Management Systems
• Streams unbounded (not like tables)
• No arrival order assumptions
• Typically no storage
• Use continuous, or standing, queries
• Reactive in nature
CHANDRASEKARAN, S., COOPER, O., DESHPANDE, A., FRANKLIN, M. J., HELLERSTEIN, J. M., HONG, W., KRISHNAMURTHY, S.,
MADDEN, S. R., REISS,
F., AND SHAH, M. A. 2003. Telegraphcq: Continuous dataﬂow processing. In Proceedings of the ACM SIGMOD International Conference on
Management of Data (SIGMOD’03). ACM, New York, NY, 668–668.
CHEN, J., DEWITT, D. J., TIAN, F., AND WANG, Y. 2000. Niagaracq: A scalable continuous query system for Internet databases. SIGMOD Rec.
29, 2, 379–390.
LIU, L., PU, C., AND TANG, W. 1999. Continual queries for internet scale event-driven information delivery. IEEE Trans. Knowl. Data Eng. 11, 4,
610–628.
ARASU, A., BABU, S., AND WIDOM, J. 2006. The CQL continuous query language: Semantic foundations and query execution. VLDB J. 15, 2,
121–142.

Data Stream Management Systems
• Continuous queries semantics
 Answer: append only stream or update store
 Exact or approximate answer
• Cons
 Atomic item is the stream
 Not possible to detect sequencing or causal patterns

Publish/Subscribe Systems
• Information items are notifications
• Indirect addressing-based communication
scheme
• Ancestors
 Message Passing
 Remote Procedure Call (RPC)
 Shared spaces
 Message Queueing
Eugster, P.T., Felber, P.A., Guerraoui, R. and Kermarrec, A.M., 2003. The many faces of
publish/subscribe. ACM Computing Surveys (CSUR), 35(2), pp.114–131.
MUHL , G., FIEGE, L., AND PIETZUCH, P. 2006. Distributed Event-Based Systems. Springer

Message Queues

Messaging Models
• Two main message models are commonly
available
 point-to-point
 publish/subscribe
• Both are based on the exchange of
messages through a channel (queue)
• Typical system will utilize a mix of these
models to achieve different messaging
objectives

Point-to-Point Model
• Straightforward asynchronous exchange of
messages
 message routed to consuming clients via a queue
 no restriction on number of publishing clients
 usually only a single consuming client
 message is delivered only once to only one receiver

Point-to-Point Model
• Messages are always delivered and will be
stored in the queue until a consumer is ready
to retrieve them

Publish Subscribe Models
• One-to-many and many-to-many distribution
mechanism
 allows single producer to send a message to one user or
potentially hundreds of thousands of consumers
• Clients "publish" to a specific topic or
channel

Publish Subscribe Models
• Channels are “subscribed” to by clients to
consume messages
• No restriction on the role of a client
 may be both a producer and consumer of a channel

• Topic-based pub/sub
 Topics are groups or channels
 Events of a topic are sent to the topic’s subscribers
ALTHERR, M., ERZBERGER, M., AND MAFFEIS, S. 1999. iBus—a software bus middleware for the
Java platform. In Proceedings of the International Workshop on Reliable Middleware Systems. 43–53.
TIBCO. 1999. TIB/Rendezvous. White paper. TIBCO, Palo Alto, CA.

• Content-based pub/sub
 Matching by message filters
 Publishers and subscribers channels are defined by the
content and the subscriptions
David S. Rosenblum and Alexander L. Wolf. 1997. A design framework for Internet-scale event
observation and notification. SIGSOFT Softw. Eng. Notes 22, 6 (November 1997), 344-360.
DOI=10.1145/267896.267920 http://doi.acm.org/10.1145/267896.267920

• Type-based pub/sub
 Matching on type hierarchy
 Type hierarchy can come from programming language
inheritance hierarchies
EUGSTER, P. AND GUERRAOUI, R. 2001. Content based publish/subscribe with structural reflection. In
Proceedings of the 6th Usenix Conference on Object-Oriented Technologies andSystems (COOTS’01).

Publish Subscribe Systems
• Events and matching
 Ordered tuples
 Attribute-values
 XML documents
• Cons
 Single event matching only

• Detection of complex patterns
 Sequencing
 Causal
 Ordering in general
 Of multiple events
 And generate complex,
or derived, events
Complex Event Processing Systems
8-12 Dec 2014, Bordeaux, France
LUCKHAM, D., 2002. The Power of Events: An Introduction to Complex Event Processing in Distributed
Enterprise Systems, Addison-Wesley Professional.

Complex Event Processing Systems
Adapted from CUGOLA, G. AND MARGARA, A., 2011. Processing flows of information: From data stream to
complex event processing. ACM Computing Surveys Journal.
SOULEIMAN HASAN AND EDWARD CURRY. 2014. APPROXIMATE SEMANTIC MATCHING OF EVENTS FOR
THE INTERNET OF THINGS. ACM TRANS. INTERNET TECHNOL. 14, 1, ARTICLE 2 (AUGUST 2014), 23
PAGES.
ADAPTED FROM (CUGOLA AND MARGARA)

DISCUSSION POINT
• What paradigms are the audience used to?
• What is their experience with each?

A THEORY FOR EVENT EXCHANGE
PART III

3 Traits of Large-Scale Event Processing
Systems
• 1- Distribution
• Two complementary aspects:
• The first is the placement of processing workloads on different nodes and
thus making use of parallel computing.
• The second aspect is that large-scale environments are inherently
distributed with event production and consumption happening at distributed
components.
• Thus, even when dealing with a centralized event
processing engine, considerations of the innate nature of
distribution of the environment of event producers and
consumers shall be taken into account.

Systems
• 2- Heterogeneity
• Differences in hardware components, protocols, operating
systems, middleware, and data
• Muhl et al. “Syntax and semantics of notifications are
likely to vary and there are inevitably different data models
in use.”
• We deal here with semantic heterogeneity
• Semantics discussed in PART IV

Problem
• Event producers and consumers are semantically
coupled
 Consumers need prior knowledge of event types, attributes and
values.
 Limits scalability in heterogeneous and dynamic environments
due to explicit dependencies
 Difficult development of event processing subscriptions/rules in
heterogeneous an dynamic environments.
Space
Time
Synch
Producer Consumer
Semantic

Exact Matching Model
Type Energy
Consumption
Place Room 202e
Amount 40 kWh
Type Electricity
Consumption
Location Room 202e
Amount 70 kWh
Type Electricity
Utilized
Venue Room 202e
Amount 600 kWh
e1
Event
Producers
e.g. Sensors
Type =“Energy Consumption”
Place =“Room 202e”
Type =“Electricity Consumption”
Location =“Room 202e”
Type =“Electricity Utilized”
Venue =“Room 202e”
Traditional
Event
Processing
e1
Consumer
e1e2
e1e3

Semantic Matching
Type Energy
Consumption
Place Room 202e
Amount 40 kWh
Type Electricity
Consumption
Location Room 202e
Amount 70 kWh
Type Electricity
Utilized
Venue Room 202e
Amount 600 kWh
e1
Event
Producers
e.g. Sensors
e1
e1e2
e1e3
Semantic
Event
Processing
Type =“Energy Consumption”~
Location =“Room 202e”
Consumer

Systems
• 3- Openness
• The term “open” has been used frequently in the literature
to describe distributed event systems at large scales, it
has not been defined precisely
• Draw upon the definition used in systems theory
“system that has external interactions in form of information,
energy, or matter transfer through its boundary.”
• An open event system from the semantics perspective as
the event environment where an agent can exchange
events with other agents that use different event
semantics

The Principle of Decoupling
• Eugster et al. : decoupling as “removing all explicit
dependencies between the interacting participants.”
• Implicit Interaction: the control over an event-based
system is decentralized into an autonomous version

The Principle of Decoupling
• Event Processing
•Exchange atomic items called events
•Scalable by decoupling
•Eugster et al.
Producer Consumer
event
Space
No Addresses
Time
Active or Not
Synchronization
No Blocking

DISCUSSION POINT
• The hypothesis is that removing explicit
dependencies between event producers and
consumers leads to an increased scalability
• Where did the dependencies go?

How Good are Our Paradigms?
• Scale
 Big volume
 Big Velocity
 Big Variety
• Distributed sources and consumers
• The big challenge is now in the exchange of
knowledge at a very large-scale

Shannon-Weaver Model

Cross-Boundaries Exchange
P. R. Carlile. Transferring, translating, and transforming: An integrative framework for managing knowledge across boundaries. Organization science, 15(5):555{568, 2004.
Known
environmen
t
Syntactic
Semantic
Pragmatic
Producer Consumer
Boundaries
Open
environment
Known
environment

Syntactic Boundary
• Transfer is the most common type of
information movement across this boundary
• A common lexicon exists
 Move and process syntax (0’s and 1’s)
 Dominant form of Shannon Weaver’s theory
• Examples
 Different data models of events
 E.g. Transfer RDF events over HTTP

Semantic Boundary
• Common lexicon doesn’t exist
• Lexicon evolve
• Ambiguities exist
• Translation is the process to cross this
boundary
• Examples
 Different ontologies for sensors
 Ontology alignment for RDF events

Pragmatic Boundary
• Actors on the sides of the boundary have
 Different contexts
 Different perspectives
 Different interests
• Transformation is the process to cross this
boundary
• Example
 Temp sensor reading of 35 celsius is acceptable from
outdoor sensors but not from indoor

Cross-Boundaries Exchange
P. R. Carlile. Transferring, translating, and transforming: An integrative framework for managing knowledge across boundaries. Organization science, 15(5):555{568, 2004.
Known
environmen
t
Syntactic
Semantic
Pragmatic
Producer Consumer
Boundaries
Open
environment
Known
environment

Transfer-Translate-Transform
• Current approaches in event processing
• Transfer
 Common event models
 Common language models
• E.g. RDF over HTTP

• Translate
 Agreements on schemas/thesauri/ontologies
• E.g. DERI Energy ontology for building energy
events
Curry, Edward, et al. "Linking building data in the cloud: Integrating cross-domain building
data using linked data." Advanced Engineering Informatics 27.2 (2013): 206-219.

• Transform
 Dedicated enrichers, joins in event languages
• CQELS language for Linked Stream Data
mashups

Decoupling for Scalability
Patrick Th. Eugster, Pascal A. Felber, Rachid Guerraoui, and Anne-Marie Kermarrec. 2003. The
many faces of publish/subscribe. ACM Comput. Surv. 35, 2 (June 2003), 114-131.
Event Processing
Space
Time
SynchronizationEvent
source
Event
consumer

A Trade-Off
• Current decoupling scale at lower boundaries
• Human in the loop to cross higher
boundaries, introducing coupling, limiting
scalability
De/Coupling
Pragmatics
(contexts)
Semantics
(meanings)
Syntactic
Consum
er
Producer
Boundaries
HeterogeneousDistributed
OpenEnvironments
De/Coupling
Formats
Space
Time
Synchronization
Agree-
ments
Ext.
data

A Trade-off
Publisher
Alice
Syntactic boundary
Semantic normalization for Bob
Semantic boundary
Consumer Bob
Publisher
Alice
Syntactic boundary
Semantic boundary
Consumer Dan
Consumer Bob
Consumer Erin
A. Small scale known environment
Low cost to cross boundary
(agreements, number of rules)
B. Large scale open environment (e.g. IoT)
High cost to cross boundary
(agreements, number of rules)
type energy consumption
increaselocation university street
e Semantic normalization for Bob
type electricity usage
riseplace university street
eB
type energy usage
riselocation university road
eE
type energy usage rise
place university road
eD
type electricity usage
riseplace university street
eB
type energy consumption
increaselocation university street
e
Souleiman Hasan and Edward Curry. 2015. TACKLING VARIETY IN INTERNET OF THINGS EVENTS, IEEE
Internet Computing (In Press)

Semantic Coupling
Event Processing
Space
Time
Synchronization
Event
source
Event
consumerSemantic Coupling
type, attributes, values
Souleiman Hasan and Edward Curry. 2014. Approximate Semantic Matching of Events for the
Internet of Things. ACM Trans. Internet Technol. 14, 1, Article 2 (August 2014), 23 pages

DISCUSSION POINT
• How significant is semantic coupling
compared to other decoupling dimensions?
• What are possible solutions?

SEMANTICS AND APPROXIMATION
PART IV

What is Semantics?

What is Semantics?
• Relationship between two spaces (or worlds or sets): the
meanings, and the symbols
Peter Gardenfors
• Semiotics and sign systems

What is Semantics?
• Semiotics (Chandlers, 2001)

Semantic Heterogeneity
• Semantics maps a language L to meanings M

The Meanings Space
The meanings space:
• Objects are individuals like a specfic laptop used by Alice.
• Properties are a “way of abstracting away redundant
information about objects”. E.g. Alice's laptop is “black”
which is a property.
• Concepts are the most generic form of objects and
properties. A concept clusters similar properties and objects
such as the concept “Laptop”.

Three Main Levels of Semantics
1. Symbolic
2. Conceptual or Sub-Symbolic
3. Non-Symbolic

Symbolic Semantics
• Information is represented by symbols, e.g. “apple”
• Processing of information is by definition a manipulation
of symbols through Rules.
• Symbols can be gathered into sentences of a language of
• thought.
• What a sentence means is a belief of an agent.
• Various beliefs are connected by logical or inferential
relations such as first-order logic in Artificial Intelligence
(AI)
• Meanings are purely the result of logical, syntactic
relations of symbols, rather than the states they refer to.

Symbolism and Computationalism
“One of the fundamental contributions to knowledge of
computer science has been to explain, at a rather basic
level, what symbols are. This explanation is a scientific
proposition about Nature. It is empirically derived, with a
long and gradual development. Symbols lie at the root of
intelligent action, which is, of course, the primary topic of
artificial intelligence. For that matter, it is a primary
question for all of computer science.”
Newell and Simon

Symbolism and Computationalism
• Fundamental tenet in AI
• But also in databases and hence event systems
• E.g. Codd’s relational model
• E.g. Unique name assumption in databases

Symbolic Semantics
• Extensional Semantics where a property is defined by the
set of objects in the world that have the property.
• Intensional Semantics alters the concept of one world to
the case of multiple possible worlds, to tackle properties
like small or big
• Situation Semantics uses one world model, but instead of
truth functions from symbols or sentences to possible
worlds, it uses a polarity function from symbols or
sentences to a subset of the world, called situation.

Critiques to Symbolic Semantics
1. They do not explain how a person can perceive two properties to
be similar.
2. Their limited account for inductive reasoning.
3. The frame problem which states that representing all necessary
knowledge about the world requires a combinatorial explosion of
logical axioms and inferences.
4. The symbol grounding problem which states that in the symbolic
paradigm the meanings of symbols are actually grounded in
symbols themselves.
5. Symbolism does not largely separate the symbolic level from the
meaning level. When event agents need to agree on the meanings
they have to agree on which is a highly costly process and thus
hinders the loose semantic coupling requirement.

Conceptual Sub-Symbolic Semantics
• Fundamentally leverage topological features of meanings,
e.g. Apple is closer to Orange than to Car
• Geometrical nature of the meaning space
• Distances and closeness between meanings can be
established.
• E.g. Gardenfors conceptual spaces
•concepts are structured into domains, e.g. the domain of colors, the
spatial domain, etc.
•Conceptual spaces are then built up from quality dimensions which
serve the purpose of building the domains. For instance, the colors
domain can be built up from three dimensions: hue, chromaticness or
saturation, and brightness.

Conceptual Sub-Symbolic Semantics
• Fundamentally leverage topological features of meanings,
e.g. Apple is closer to Orange than to Car
• Geometrical nature of the meaning space
• Distances and closeness between meanings can be
established.
• E.g. Gardenfors conceptual spaces
•Computationally challenging due to need to build and agree on quality
dimensions

Distributional Semantic Model
• Distributional hypothesis: the context surrounding a given word
in a text provides relevant information about its meaning.
• Simplified semantic model.
• Associational and quantitative.
• Explicit Semantic Analysis (ESA) is the primary distributional
model used in this work.
A wife is a female partner in a marriage. The term "wife" seems to be a
close term to bride, the latter is a female participant in a wedding
ceremony, while a wife is a married woman during her marriage.
...
Slide Credits: Andre Freitas (http://andrefreitas.org/), [Freitas et al., 2013]

Distributional Semantic Model
c1
function (number of times that the words occur in c1)
c1
child
husband
spouse
cn
c2
function (number of times that the words occur in c1)
0.7
0.5
Commonsense is here
Slide Credits: Andre Freitas (http://andrefreitas.org/), [Freitas et al., 2013]

Semantic Relatedness
θ
c1
child
husband
spouse
cn
c2
Works as a semantic ranking function
E.g. esa(room, building)= 0.099
E.g. esa(room, car)= 0.009
Slide Credits: Andre Freitas (http://andrefreitas.org/)
[Freitas et al., 2013]

Critiques to Sub-Symbolic Semantics
• Compositionality: they mainly concern lexical meanings,
i.e. meanings of individual terms, rather than complex
sentences.
• We argue that the compositionality problem is not an issue
for event matching. That is due to the fact that linguistic
structures and syntax is not the kind of data model used in
event processing systems to represent events and
subscriptions.

Non-Symbolic Semantics
• Artificial Neural Networks (ANNs)
• Connectionism
• Fundamentally can be abstracted into geometrical models
• Difficult to build an interpret

Semantic Models

Tagging
• Inspired by works in social
tagging, i.e. folksonomies
• Folksonomies are bottom
up approach to semantics
• Free words
• Tag events such as in IoT
• Leads to the concepts of
thingsonomies

Free Tagging and Thingsonomies
• Top-down taxonomies difficult to impose
• Folksonomies successful in social tagging

Approximation
• Coupling is important to cross semantic and pragmatic
boundaries, but it limits scalability.
• Loosening coupling at these levels is a compromise to
tackle the trade-of between decoupling for scalability and
crossing the boundaries.
• The cost of this compromise is a loss in effectiveness
while crossing the boundaries, i.e. loss of some precision
and context when processing the events.
• In literature approximation has been used to address
•Time efficiency, e.g. approximation algorithms
•Full integration, e.g. uncertain schema matching, Gal et al.

Approximation
• Exact matching assumes full agreements
• Approximation flexible with uncertainties

APPROACHES TO SEMANTIC COUPLING
PART V

Loosening the Semantic Coupling
• Approach 1: Content-Based with Semantic
Decoupling
• Approach 2: Content-Based with Implicit Shared
Agreements
• Approach 3: Concept-Based
• Approach 4: Loose Semantic Coupling +
Approximation
Hasan, S. and Curry, E., 2014. Approximate Semantic Matching of Events for The Internet
of Things. ACM Transactions on Internet Technology (TOIT).
• Approach 5: Theme-Based
Hasan, S. and Curry, E., 2014. Thematic Event Processing. Middleware 2014.

Approach 1: Content-Based with Semantic
Decoupling
7-11 July 2014,
Rhodes, Greece
EarthBiAs2014
Producer Consumer
event
Semantic De-
Coupling
Happened
Publish:
A
Happened
Interested in
Subscribe:
Interested in
B
• Very low detection rate
 High false positives/negatives
 Low precision/recall

Current Approaches
Semantic Decoupling
Effectiveness & Efficiency
Content-based
Concept-based
Bottom-up
Semantics

Approach 1: Content-Based with Semantic
Decoupling
7-11 July 2014,
Rhodes, Greece
EarthBiAs2014
Producer Consumer
event
Semantic De-Coupling
Happened
Publish:
A
Happened
Interested in
Subscribe:
Interested in A
Interested in B
Interested in C
• Use many rules to improve detection
 Time and effort
 Affects scalability to heterogeneous environments

Approach 2: Content-Based with Implicit
Shared Agreements
7-11 July 2014,
Rhodes, Greece
EarthBiAs2014
Producer Consumer
event
Semantic Coupling via
Implicit Agreements
Happened
Publish:
A Happened
Interested in
Subscribe:
Interested in A
Face-to-face, or via
documentation
e.g. Use symbol A to
describe

Shared Agreements
• Implicit semantics
 Top-down approach to semantics
 Granular on the level of concepts
Producer Consumer
event
Implicit Agreements
Happened
Publish:
A Happened
Interested in
Subscribe:
Interested in A

Shared Agreements
• Need for shared agreements
 Time and effort
Producer Consumer
event
Implicit Agreements
Happened
Publish:
A Happened
Interested in
Subscribe:
Interested in A

Approach 3: Concept-Based
7-11 July 2014,
Rhodes, Greece
EarthBiAs2014
Producer Consumer
event
Ontologies
Happened
Publish:
A Happened
Interested in
Subscribe:
Interested in B
C
D
B
E
A
F
subClassOf

• Explicit semantics
 Top-down approach to semantics
 Granular on the level of concepts
Producer Consumer
event
Ontologies
Happened
Publish:
A Happened
Interested in
Subscribe:
Interested in B

• Need for shared agreements
 Time and effort
Producer Consumer
event
Ontologies
Happened
Publish:
A Happened
Interested in
Subscribe:
Interested in B

Approach 4: Loose Semantic Coupling +
Approximation
7-11 July 2014,
Rhodes, Greece
EarthBiAs2014
Producer Consumer
event
Loose Semantic
Coupling via Large
Text Corpora
Happened
Publish:
A Happened
Interested in
Subscribe:
Interested in B
A d1 d2 d3 d4 d5 d6 d7 d8 ….
B d1 d3 d4 d17 d25 d26 d77 d78 ….
~
Souleiman Hasan and Edward
Curry. 2014. Approximate
Semantic Matching of Events for
the Internet of Things. ACM Trans.
Internet Technol. 14, 1, Article 2
(August 2014), 23 pages

Approximation
• Bottom-up model of semantics
• Global semantics: distribution vs. granular
7-11 July 2014,
Rhodes, Greece
EarthBiAs2014
Producer Consumer
event
Loose Semantic
Coupling via Large
Text Corpora
Happened
Publish:
A Happened
Interested in
Subscribe:
Interested in B
~

Approximation
• Low cost to scale to heterogeneous
environments
• Slightly lower detection rate
7-11 July 2014,
Rhodes, Greece
EarthBiAs2014
Producer Consumer
event
Loose Semantic
Coupling via Large
Text Corpora
Happened
Publish:
A Happened
Interested in
Subscribe:
Interested in B
~

Approach 5: Thematic Event Processing
7-11 July 2014,
Rhodes, Greece
EarthBiAs2014
Producer Consumer
event
Loose Semantic
Coupling via Large
Text Corpora
Happened
Publish:
A Happened
Interested in
Subscribe:
Interested in B
~
• Can we exchange better approximations of
meanings rather than mere symbols to
improving detection rate?

Approach 5: Thematic Event Processing
7-11 July 2014,
Rhodes, Greece
EarthBiAs2014
Producer Consumer
event
Loose Semantic
Coupling via Large
Text Corpora
Happened
Publish:
(A+T1)
Happened
Interested in
Subscribe:
Interested in
(B+T2)
A d1 d2 d3 d4 d5 d6 d7 d8 ….
B d1 d3 d4 d17 d25 d26 d77 d78 ….
~
Souleiman Hasan and
Edward Curry. 2014.
Thematic event processing.
In Proceedings of the 15th
International Middleware
Conference (Middleware '14).

Summary
Simple
Content-
based
Content-
based +
Many Rules
Concept-
based
Simple
Distributional+
Approximation
Thematic
Matching exact string
matching
exact string
matching
Boolean semantic
matching
approximate
semantic matching
approximate
semantic matching
Semantic
Coupling
term-level full
agreement
term-level full
agreement
concept-level
shared agreement
loose agreement loose agreement
Semantics not explicit not explicit top-down
ontology-based
statistical model
based on
distributional
semantics
statistical model
based on
distributional
semantics +
themes
Effectiveness very low 100% depends on the
domains and
number of
concept models
depends on the
corpus
depends on the
corpus + theme
representatives
Cost defining a small
number of rules
defining a large
number of rules
establishing
shared agreement
on ontologies
minimal agreement
on a large
textual corpus
minimal
agreement on a
large
textual corpus +
good theme
representatives
Efficiency high high medium to high medium to high Medium to high

DISCUSSION POINT
• Is the audience familiar with the symbolic vs.
non-symbolic debate in AI?
• What position do they take?
• How relevant to our current discussion?

RDF EVENT PROCESSING
CASE STUDY

From The Web The Semantic Web
• 1989 Tim Berners Lee
proposed what became the
Web
• Hypertext over HTTP and URI
• 2001 TBL et al. published their
Scientific American paper The
Semantic Web
• A structured Web for machines
• Builds over the Web
architecture

From The Web The Semantic Web
• Initial Semantic
Web vision
• Standards: W3C
• Research: mainly
ISWC, ESWC
• Recently attention
moved to the lower
layers of the stack
• Called: Linked Data

Linked Data is Web-based
• Leverages the architecture of the Web to
make sharing data easier
• Linked Data is a method of exposing, sharing,
and connecting data (via dereferenceable
URIs) on the Web.
• Provides a Data (RDF) and Naming (URI)
model for the Web
• W3C Web-based Standards
• Adaptive Ontologies

Linked Data is Web-based
1. Use URIs as names for things
2. Use HTTP URIs so that people
can look up those names.
3. When someone looks up
a URI, provide useful
information, using the
standards (RDF*, SPARQL)
4. Include links to other URIs. so
that they can discover more
things
http://www.w3.org/DesignIssues/LinkedData.html

Linked Data Cloud
US government
UK government
BBC
New York Times
LinkedGeoData
BestBuy
Overstock.com
Facebook
Media
Government
Geo
Publications
User-generated
Life sciences
Cross-domain
Over 200 open data sets with more than 25 billion facts,
interlinked by 400 million typed links, doubling every 10 month!
Linking Open Data cloud diagram, by Richard Cyganiak and Anja Jentzsch.

Two Key Ingredients
• RDF – Resource Description Framework
Graph based Data – nodes and arcs
 Identifies objects (URIs)
 Interlink information (Relationships)
• Vocabularies (Ontologies)
 provide shared understanding of a domain
 organise knowledge in a machine-comprehensible way
 give an exploitable meaning to the data

Example: Linked Building Data
http://www.deri.ie/about/team/member/edward_curry/
http://lab.linkeddata.deri.ie/2010/deri-rooms#r202e
http://vocab.deri.ie/rooms#occupant
Resource Description Framework (RDF)
subject - predicate – object
Edward Curry is the Occupant of Room 202e
Edward Curry is the Occupant of Room 202e

Example: Linked Building Data

Why Linked Data for the IoT?
• Many communities struggle with closed
approaches
 E.g., pervasive computing, embedded systems, IoT, …
• Cyber-Physical Systems are inherently “open
world”
 Prof. David Karger (MIT) in his ESWC 2013 keynote:
“Semantic Web technologies support and open world
assumption where millions of unforeseeable schemas
may have to be integrated.”

Why Linked Data for the IoT?
• Simple integration with existing LOD data sets
 Geo-spatial, governmental, media, …
• Manageable integration effort with other graph
data, e.g., Google Knowledge Graph,
Facebook Graph, etc.

• How to classify RDF event processing
according to previous classification?
• What are the pros and cons?
• How to make RDF event processing more
suitable for large scale IoT with variety?
DISCUSSION POINT

• RDF Event Processing
 Based on a top-down semantic model, i.e. ontologies
 Use RDF as data model
 Use URIs for interlinking
 SPARQL-like query languages
 Graph matching
DISCUSSION POINT

 Top-down semantics difficult to agree on
 Granular semantics difficult to agree on
 RDF graph model powerful
 URIs good for linking
 SPARQL-like query languages, expressive
 SPARQL-like query languages, not user friendly
DISCUSSION POINT

 Is an example of Concept-Based approach to semantic
coupling
 Needs some research to address the problems of this
class, but also to gain power of RDF model and URIs
 Efforts in querying RDF graphs using distributional
semantics can largely feed this research, see [Freitas et
al., 2013]
DISCUSSION POINT

TEMATIC EVENT PROCESSING
PART VI

The Thematic Approach
• Decoupling is good for scalability
• Decoupling parties creates challenges to
cross semantic boundaries
• We can solve that by re-coupling parties at
semantic boundaries, but it is not ideal

Problem: How can we achieve effective and
efficient event processing without causing
semantic coupling at the semantic boundaries?

• Exchange approximations of meanings
• Instead of exchanging words, try to convey
their meaning too
• How?Use tags that can help understand their
meaning

• Inspired by works in
social tagging, i.e.
folksonomies
• Folksonomies are
bottom up approach
to semantics
Souleiman Hasan and Edward Curry. 2014. Thematic event processing. In Proceedings of the 15th International
Middleware Conference (Middleware '14). ACM, Bordeaux, France, 109-120. DOI=10.1145/2663165.2663335
http://doi.acm.org/10.1145/2663165.2663335
Image credit http://lowriderlibrarian.blogspot.fr/

Event
Publisher
Alice
Consumer
Bob
Theme the
Payload
Subscription
Theme ths
ExpressionApproximate
matcher
Parameterization
Loose coupling mode: lightweight agreement on themes
No coupling mode: free use of well representative themes
Middleware Conference (Middleware '14). ACM, Bordeaux, France, 109-120. DOI=10.1145/2663165.2663335

Event Representation
Event
energy, appliances, building
type: increased energy consumption event,
measurement unit: kilowatt per hour,
device: computer,
office: room 112
• Thematic tags added to events

Subscription Representation
Subscription
power, computers
type= increased energy usage event~,
device~= laptop~,
office= room 112
• Thematic tags added to subscriptions

Matching Model
• Top-1 and Top-k most probable mappings
{type: increased energy
consumption event,
measurement unit: kilowatt per hour,
device: computer,
desk: desk 112c,
office: room 112,
floor: ground floor,
zone: building,
city: Galway,
country: Ireland,
continent: Europe}
{type = increased energy
consumption event,
device = laptop~,
room~esa = room 112}

Probabilistic Approximate Matcher
• Top-1 and Top-k mappings between an event
and a subscription

Parameterized Similarity
• Thematic tags used to parameterize the
semantic measure

Interpreting Terms
• Project vectors in a
distributional semantic
vector space
• Thematic projection

Evaluation Metrics
• Number of exact rules
• Precision
• Recall
• F1Score
• Throughput
• Standard error

Evaluation Dataset
• Seed events synthesized from IoT sensors
• SmartSantander smart city project
 Luis Sanchez, Jos´e Antonio Galache, Veronica Gutierrez, JM Hernandez, J Bernat, Alex Gluhak, and Tom´as Garcia. 2011.
SmartSantander: The meeting point between Future Internet research and experimentation and the smart cities. In Future
Network & Mobile Summit (FutureNetw), 2011. IEEE, 1–8.
• Sensor Capabilities
 solar radiation, particles, speed, wind direction, wind
speed, temperature, water flow, atmospheric pressure,
noise, ozone, rainfall, parking, radiation par, co, ground
temperature, light, no2, soil moisture tension, relative
humidity, energy consumption, cpu usage, memory
usage
Internet of Things. ACM Trans. Internet Technol. 14, 1, Article 2 (August 2014), 23 pages.
DOI=10.1145/2633684 http://doi.acm.org/10.1145/2633684

Evaluation Dataset
• Seed events synthesized from IoT sensors
• Linked Energy Intelligence platform
 Edward Curry, Souleiman Hasan, and Sean O’Riain. 2012. Enterprise energy management using a linked dataspace for Energy
Intelligence. In Sustainable Internet and ICT for Sustainability (SustainIT), 2012. IEEE, 1–6.
• Car brands from the yahoo directory
 Yahoo! 2013. Yahoo! Directory: Automotive - Makes and Models. (2013). http://dir.yahoo.com/recreation/ automotive/makes and
models/
• Home based appliances from BLUED dataset
 Kyle Anderson, Adrian Ocneanu, Diego Benitez, Derrick Carlson, Anthony Rowe, and Mario Berges. 2012. BLUED: A Fully
Labeled Public Dataset for Event-Based Non-Intrusive Load Monitoring Research. In Proc. SustKDD.
• Rooms from DERI Building
 Richard Cyganiak. 2013. Rooms in the DERI building. (2013). http://lab.linkeddata.deri.ie/2010/deri-rooms
Internet of Things. ACM Trans. Internet Technol. 14, 1, Article 2 (August 2014), 23 pages.
DOI=10.1145/2633684 http://doi.acm.org/10.1145/2633684

Evaluation Methodology
Souleiman Hasan and Edward
Curry. 2014. Thematic event
processing. In Proceedings of
the 15th International
Middleware
Conference (Middleware '14).
ACM, Bordeaux, France, 109-
120.

Evaluation
• FScore up to 95% and 1000s events/sec
Hasan, S. and Curry,
E., 2014.
Approximate
Semantic Matching
of Events for The
Internet of Things.
ACM Transactions
on Internet
Technology (TOIT).

Evaluation
Middleware Conference (Middleware '14). ACM, Bordeaux, France, 109-120.

Evaluation
The use of less terms to describe events, around 2-7, and more to
describe subscriptions, around 2 -15, can achieve a good matching
quality and throughput together with less error rates.
Lightweight amount of tags.
Middleware Conference (Middleware '14). ACM, Bordeaux, France, 109-120.

BUILDING IOT Event Systems
PART VII

Challenges for Building IoT Systems
• Vastly heterogeneous, decoupled, and
distributed nodes
• Lack of central coordination, reference
requirements, or data model
• High overhead associated with software
design associated with establishing
agreements between parties

Building IoT Event Systems
Indexing
Collector
Semantic
relatedness
web service
Textual
corpus
Vector
space
index
Consumer Bob
(user)
Publisher Alice
Publish + thematic
tags
Thematic event processing engine(s)
Approximate single event matching
Subscribe
+ thematic
tags
IoT sensors
Terms +
themes
pairs
Relatednes
s score
CollectorPublisher Carol
Publish + thematic
tags
CollectorPublisher Dave
Publish + thematic
tags
Consumer Dan
(application developer)
Consumer Erin
Heterogeneous IoT Events
Relevant
events
normalized
for Bob
Subscribe
+ thematic
tags
Relevant
events
normalized
for Dan
Subscribe
+ thematic
tags
Relevant
events
normalized
for Erin
5
3
1
2
4 6
Souleiman Hasan and
Edward Curry. 2015.
TACKLING VARIETY IN
INTERNET OF THINGS
EVENTS, IEEE Internet
Computing

Collector
Semantic
relatedness
web service
Consumer Bob
(user)
Publisher Alice
Publish + thematic
tags
Subscribe
+ thematic
tags
IoT sensors
Terms +
themes
pairs
Relatednes
s score
Publish + thematic
tags
Publish + thematic
tags
Consumer Dan
Consumer Erin
Relevant
events
normalized
for Bob
Subscribe
+ thematic
tags
Relevant
events
normalized
for Dan
Subscribe
+ thematic
tags
Relevant
events
normalized
for Erin
5
3
2
4 6
1
Textual corpus
Vector
space
index
Indexing
• Step 1: Build the semantic model
 To enable the system establish relationships between
various terms such as ‘computer’ vs. ‘laptop’.
 Use distributional semantics: bottom-up, not fine-grained
model
 Revise model with software iterations
 update corpus
 Wikipedia or a subset of it is a good start
 Use enterprise Wikis if available

1
Textual
corpus
Vector
space
index
Indexing
Collector
Consumer Bob
(user)
Publisher Alice
Publish + thematic
tags
Subscribe
+ thematic
tags
IoT sensors
Publish + thematic
tags
Publish + thematic
tags
Consumer Dan
Consumer Erin
Relevant
events
normalized
for Bob
Subscribe
+ thematic
tags
Relevant
events
normalized
for Dan
Subscribe
+ thematic
tags
Relevant
events
normalized
for Erin
5
3
4 6
• Step 2: Avail semantic measure
 REST and JSON
 Request:
http://example.com/esa?term1=energy&term2=electricity
 Response {“relatedness” : 0.154}
Semantic
relatedness
web service
Terms +
themes pairs
Relatedness
score
2

Semantic
relatedness
web service
Terms +
themes
pairs
Relatednes
s score
2
1
Textual
corpus
Vector
space
index
Indexing
Consumer Bob
(user)
Subscribe
+ thematic
tags
Consumer Dan
Consumer Erin
Relevant
events
normalized
for Bob
Subscribe
+ thematic
tags
Relevant
events
normalized
for Dan
Subscribe
+ thematic
tags
Relevant
events
normalized
for Erin
5
4 6
• Step 3: Publishers associate sensor events
with tags
 Thematic tags to represent domain and meaning
 Associate with attribute-value events
 Tags {energy, appliances, building} to accompany
increased energy consumption event
CollectorPublisher Alice
Publish + thematic
tags
IoT sensors
Publish + thematic
tags
Publish + thematic
tags
Heterogeneous IoT
Events
3

Relevant
events
normalized
for Bob
5
6
CollectorPublisher Alice
Publish + thematic
tags
IoT sensors
Publish + thematic
tags
Publish + thematic
tags
3
Semantic
relatedness
web service
Terms +
themes
pairs
Relatednes
s score
2
1
Textual
corpus
Vector
space
index
Indexing
Consumer Dan
Consumer Erin
Subscribe
+ thematic
tags
Relevant
events
normalized
for Dan
Subscribe
+ thematic
tags
Relevant
events
normalized
for Erin
• Step 4: Subscribers associate subscriptions
with tags
 Thematic tags to represent domain and meaning
 Associate with attribute-value events
 Tags {power, computers} to accompany increased
energy consumption event of a device
Subscribe
+ thematic
tags
4
Consumer Bob
(user)

4
Relevant
events
normalized
for Bob
6
Consumer Bob
(user)
Subscribe
+ thematic
tags
Publisher Alice
Publisher Carol
Publisher Dave
1
Textual
corpus
Indexing
Semantic
relatedness
web service
Terms +
themes
pairs
Relatednes
s score
2
Vector
space
index
Consumer Dan
Consumer Erin
Subscribe
+ thematic
tags
Relevant
events
normalized
for Dan
Subscribe
+ thematic
tags
Relevant
events
normalized
for Erin
• Step 5: Middleware matches events to
subscriptions
 Approximate, probabilistic semantic normalization
Publish +
thematic tags
Publish +
thematic tags
Publish +
thematic tags
3
5
Collector
Collector
Collector

Subscribe
+ thematic
tags
4
5
Publisher Alice
Publisher Carol
Publisher Dave
Publish + thematic
tags
Publish + thematic
tags
Publish + thematic
tags
1
Textual
corpus
Indexing
3
Semantic
relatedness
web service
Terms +
themes
pairs
Relatednes
s score
2
Vector
space
index
Collector
IoT sensors
Collector
Collector
Consumer Dan
Consumer Erin
Subscribe
+ thematic
tags
Relevant
events
normalized
for Dan
Subscribe
+ thematic
tags
Relevant
events
normalized
for Erin
• Step 6: Middleware distributes matching
results to subscribers
 Relevant events are returned
 Uncertainty scores are associated
 Uncertainty reflects semantic normalization
Relevant
events
normalized
for Bob
6
Consumer Bob
(user)

FUTURE RESEARCH CHALLENGES
PART VIII

Future Research Challenges
• Investigation of user tagging behavior of
sensor events in zero-coupled open
environment
• Investigation of the limitations of evolution of
semantic agreements in open environments

• Investigation of event dynamic enrichment for
pragmatic boundaries within loose coupled
environments
Hasan, S., O’Riain, S. and Curry, E., 2013.
Towards Unified and Native Enrichment in
Event Processing Systems. In The 7th ACM
International Conference on Distributed
Event-Based Systems (DEBS 2013).
Arlington, Texas, USA: ACM

• Dynamic enrichment
Hasan, S., O’Riain, S. and Curry, E., 2013.
Towards Unified and Native Enrichment in
Event Processing Systems. In The 7th ACM
International Conference on Distributed
Event-Based Systems (DEBS 2013).
Arlington, Texas, USA: ACM

• Support complex pattern detection
• Uncertainty propagation
E., 2014.
Approximate
Semantic Matching
of Events for The
Internet of Things.
ACM Transactions
on Internet
Technology (TOIT).

• Support complex pattern detection
• Statistical monotonicity and top-k
8-12 Dec 2014, Bordeaux, FranceMiddleware ’14 161
E., 2014.
Approximate
Semantic Matching
of Events for The
Internet of Things.
ACM Transactions
on Internet
Technology (TOIT).

• Optimization of approximate matching for
higher throughputs
• Extension of the approximate thematic
matcher into parallel computation
• Investigation of integration of the vector
space model within current data management
systems paradigms

Conclusions
• Coupling necessary for crossing boundaries
• Decoupling necessary for scalable software
• Event-based systems need extension to
address the coupling/decoupling trade-off for
semantics
• Approximate and thematic event processing
exchange approximations of meaning with
loose semantic coupling

DISCUSSION POINT
• Overall questions and answers

References
• CUGOLA, G. AND MARGARA, A., 2011. Processing flows of information: From data stream to complex event processing.
ACM Computing Surveys Journal.
• EUGSTER, P.T., FELBER, P.A., GUERRAOUI, R. AND KERMARREC, A.M., 2003. The many faces of publish/subscribe. ACM
Computing Surveys (CSUR), 35(2), pp.114–131.
• Carlile, Paul R. "Transferring, translating, and transforming: An integrative framework for managing knowledge across
boundaries." Organization science15.5 (2004): 555-568.
• SOULEIMAN HASAN AND EDWARD CURRY. 2015. TACKLING VARIETY IN INTERNET OF THINGS EVENTS, IEEE
Internet Computing (In Press)
• SOULEIMAN HASAN AND EDWARD CURRY. 2014. APPROXIMATE SEMANTIC MATCHING OF EVENTS FOR THE INTERNET OF
THINGS. ACM TRANS. INTERNET TECHNOL. 14, 1, ARTICLE 2 (AUGUST 2014), 23 PAGES. DOI=10.1145/2633684
HTTP://DOI.ACM.ORG/10.1145/2633684
• HASAN, S., O’RIAIN, S. AND CURRY, E., 2013. TOWARDS UNIFIED AND NATIVE ENRICHMENT IN EVENT PROCESSING
SYSTEMS. IN THE 7TH ACM INTERNATIONAL CONFERENCE ON DISTRIBUTED EVENT-BASED SYSTEMS (DEBS 2013).
ARLINGTON, TEXAS, USA: ACM.
• HASAN, S., O’RIAIN, S. AND CURRY, E., 2012. Approximate Semantic Matching of Heterogeneous Events. In 6th ACM
International Conference on Distributed Event-Based Systems (DEBS 2012). Berlin, Germany: ACM, pp. 252–263.
• SOULEIMAN HASAN AND EDWARD CURRY. 2014. THEMATIC EVENT PROCESSING. IN PROCEEDINGS OF THE 15TH
INTERNATIONAL MIDDLEWARE CONFERENCE (MIDDLEWARE '14). ACM, BORDEAUX, FRANCE, 109-120.
DOI=10.1145/2663165.2663335 HTTP://DOI.ACM.ORG/10.1145/2663165.2663335
• HASAN, S., CURRY, E., BANDUK, M., AND O’RIAIN, S. TOWARD SITUATION AWARENESS FOR THE SEMANTIC SENSOR WEB:
COMPLEX EVENT PROCESSING WITH DYNAMIC LINKED DATA ENRICHMENT. THE 4TH INTERNATIONAL WORKSHOP ON
SEMANTIC SENSOR NETWORKS 2011 (SSN11), (2011), 60–72.

Dataset and Software
• Dataset
 Souleiman Hasan, Edward Curry, Thematic event
processing dataset, DOI: 10.13140/2.1.3342.9123
 Available at
http://www.researchgate.net/publication/263673956_Thematic_event_proce
ssing_dataset
• Collider
 Souleiman Hasan, Kalpa Gunaratna, Yongrui Qin, and Edward Curry. 2013.
Demo: approximate semantic matching in the collider event processing
engine. In Proceedings of the 7th ACM international conference on
Distributed event-based systems (DEBS '13). ACM, New York, NY, USA,
337-338. DOI=10.1145/2488222.2489277

More References
• OECD, 2012. MACHINE-TO-MACHINE COMMUNICATIONS: CONNECTING BILLIONS OF DEVICES. OECD DIGITAL ECONOMY PAPERS, NO. 192.
• P. MCFEDRIES, THE COMING DATA DELUGE, IEEE SPECTRUM, 2011.
• CUGOLA, G. AND MARGARA, A., 2011. PROCESSING FLOWS OF INFORMATION: FROM DATA STREAM TO COMPLEX EVENT PROCESSING. ACM COMPUTING SURVEYS JOURNAL.
• EUGSTER, P.T., FELBER, P.A., GUERRAOUI, R. AND KERMARREC, A.M., 2003. THE MANY FACES OF PUBLISH/SUBSCRIBE. ACM COMPUTING SURVEYS (CSUR), 35(2), PP.114–131.
• LUCKHAM, D., 2002. THE POWER OF EVENTS: AN INTRODUCTION TO COMPLEX EVENT PROCESSING IN DISTRIBUTED ENTERPRISE SYSTEMS, ADDISON-WESLEY PROFESSIONAL.
• DAYAL, U., BLAUSTEIN, B., BUCHMANN, A., CHAKRAVARTHY, U., HSU, M., LEDIN, R., MCCARTHY, D., ROSENTHAL, A., SARIN, S., CAREY, M. J., LIVNY, M., AND
JAUHARI, R. 1988. THE HIPAC PROJECT: COMBINING ACTIVE DATABASES AND TIMING CONSTRAINTS. SIGMOD REC. 17, 1, 51–70.
• LIEUWEN, D. F., GEHANI, N. H., AND ARLEIN, R. M. 1996. THE ODE ACTIVE DATABASE: TRIGGER SEMANTICS AND IMPLEMENTATION. IN PROCEEDINGS OF THE 12TH INTERNATIONAL
CONFERENCE ON DATA ENGINEERING (ICDE’96). IEEE COMPUTER SOCIETY, LOS ALAMITOS, CA, 412–420.
• GATZIU, S. AND DITTRICH, K. 1993. EVENTS IN AN ACTIVE OBJECT-ORIENTED DATABASE SYSTEM. IN PROCEEDINGS OF THE INTERNATIONAL WORKSHOP ON RULES IN DATABASE
SYSTEMS (RIDS), N. PATON AND H. WILLIAMS, EDS. WORKSHOPS IN COMPUTING, SPRINGER-VERLAG, EDINBURGH, U.K.
• CHAKRAVARTHY, S. AND ADAIKKALAVAN, R. 2008. EVENTS AND STREAMS: HARNESSING AND UNLEASHING THEIR SYNERGY! IN PROCEEDINGS OF THE 2ND INTERNATIONAL
CONFERENCE ON DISTRIBUTED EVENT-BASED SYSTEMS (DEBS’08). ACM, NEW YORK, NY, 1–12.
• CHANDRASEKARAN, S., COOPER, O., DESHPANDE, A., FRANKLIN, M. J., HELLERSTEIN, J. M., HONG, W., KRISHNAMURTHY, S., MADDEN, S. R., REISS, F., AND
SHAH, M. A. 2003. TELEGRAPHCQ: CONTINUOUS DATAﬂOW PROCESSING. IN PROCEEDINGS OF THE ACM SIGMOD INTERNATIONAL CONFERENCE ON MANAGEMENT OF DATA
(SIGMOD’03). ACM, NEW YORK, NY, 668–668.
• CHEN, J., DEWITT, D. J., TIAN, F., AND WANG, Y. 2000. NIAGARACQ: A SCALABLE CONTINUOUS QUERY SYSTEM FOR INTERNET DATABASES. SIGMOD REC. 29, 2, 379–390.
• LIU, L., PU, C., AND TANG, W. 1999. CONTINUAL QUERIES FOR INTERNET SCALE EVENT-DRIVEN INFORMATION DELIVERY. IEEE TRANS. KNOWL. DATA ENG. 11, 4, 610–628.
• ARASU, A., BABU, S., AND WIDOM, J. 2006. THE CQL CONTINUOUS QUERY LANGUAGE: SEMANTIC FOUNDATIONS AND QUERY EXECUTION. VLDB J. 15, 2, 121–142.
• MUHL , G., FIEGE, L., AND PIETZUCH, P. 2006. DISTRIBUTED EVENT-BASED SYSTEMS. SPRINGER
• ALTHERR, M., ERZBERGER, M., AND MAFFEIS, S. 1999. IBUS—A SOFTWARE BUS MIDDLEWARE FOR THE JAVA PLATFORM. IN PROCEEDINGS OF THE INTERNATIONAL WORKSHOP
ON RELIABLE MIDDLEWARE SYSTEMS. 43–53.
• TIBCO. 1999. TIB/RENDEZVOUS. WHITE PAPER. TIBCO, PALO ALTO, CA.
• A. FREITAS, J. G. OLIVEIRA, S. O’RIAIN, J. C. DA SILVA, AND E. CURRY. QUERYING LINKED DATA GRAPHS USING SEMANTIC RELATEDNESS: A VOCABULARY INDEPENDENT
• APPROACH. DATA & KNOWLEDGE ENGINEERING, 88:126–141, 2013.

More References
• DAVID S. ROSENBLUM AND ALEXANDER L. WOLF. 1997. A DESIGN FRAMEWORK FOR INTERNET-SCALE EVENT OBSERVATION AND NOTIFICATION. SIGSOFT SOFTW. ENG. NOTES 22, 6
(NOVEMBER 1997), 344-360. DOI=10.1145/267896.267920 HTTP://DOI.ACM.ORG/10.1145/267896.267920
• EUGSTER, P. AND GUERRAOUI, R. 2001. CONTENT BASED PUBLISH/SUBSCRIBE WITH STRUCTURAL REFLECTION. IN PROCEEDINGS OF THE 6TH USENIX CONFERENCE ON OBJECT-
ORIENTED TECHNOLOGIES ANDSYSTEMS (COOTS’01).
• C. SHANNON AND W. WEAVER. THE MATHEMATICAL THEORY OF COMMUNICATION. UNIVERSITY OF ILLINOIS PRESS, 1949.
• P. R. CARLILE. TRANSFERRING, TRANSLATING, AND TRANSFORMING: AN INTEGRATIVE FRAMEWORK FOR MANAGING KNOWLEDGE ACROSS BOUNDARIES. ORGANIZATION SCIENCE,
15(5):555{568, 2004.
• CURRY, EDWARD, SOULEIMAN HASAN, AND SEÁN O'RIAIN. "ENTERPRISE ENERGY MANAGEMENT USING A LINKED DATASPACE FOR ENERGY INTELLIGENCE." SUSTAINABLE INTERNET AND
ICT FOR SUSTAINABILITY (SUSTAINIT), 2012. IEEE, 2012.
• CURRY, EDWARD, ET AL. "LINKING BUILDING DATA IN THE CLOUD: INTEGRATING CROSS-DOMAIN BUILDING DATA USING LINKED DATA." ADVANCED ENGINEERING INFORMATICS 27.2 (2013):
206-219.
• PATRICK TH. EUGSTER, PASCAL A. FELBER, RACHID GUERRAOUI, AND ANNE-MARIE KERMARREC. 2003. THE MANY FACES OF PUBLISH/SUBSCRIBE. ACM COMPUT. SURV. 35, 2 (JUNE
2003), 114-131.
• A. CARZANIGA, D. S. ROSENBLUM, AND A. L. WOLF. ACHIEVING SCALABILITY AND EXPRESSIVENESS IN AN INTERNET-SCALE EVENT NOTI_CATION SERVICE. IN PROCEEDINGS OF THE
NINETEENTH ANNUAL ACM SYMPOSIUM ON PRINCIPLES OF DISTRIBUTED COMPUTING, PAGES 219{227. ACM, 2000.
• M. PETROVIC, I. BURCEA, AND H.-A. JACOBSEN. S-TOPSS: SEMANTIC TORONTO PUBLISH/SUBSCRIBE SYSTEM. IN PROCEEDINGS OF THE 29TH INTERNATIONAL CONFERENCE ON VERY
LARGE DATA BASES - VOLUME 29, VLDB '03, PAGES 1101-1104. VLDB ENDOWMENT, 2003.
• HASAN, S. AND CURRY, E., 2014. APPROXIMATE SEMANTIC MATCHING OF EVENTS FOR THE INTERNET OF THINGS. ACM TRANSACTIONS ON INTERNET TECHNOLOGY (TOIT). IN PRESS
• HASAN, S. AND CURRY, E., 2014. THEMATIC EVENT PROCESSING. MIDDLEWARE 2014. UNDER REVIEW.
• HASAN, S. AND CURRY, E., 2014. TACKLING EVENT VARIETY IN INTERNET OF THINGS SOFTWARE. IEEE INTERNET COMPUTING 2015. UNDER REVIEW.
• LUIS SANCHEZ, JOS´E ANTONIO GALACHE, VERONICA GUTIERREZ, JM HERNANDEZ, J BERNAT, ALEX GLUHAK, AND TOM´AS GARCIA. 2011. SMARTSANTANDER: THE MEETING POINT
BETWEEN FUTURE INTERNET RESEARCH AND EXPERIMENTATION AND THE SMART CITIES. IN FUTURE NETWORK & MOBILE SUMMIT (FUTURENETW), 2011. IEEE, 1–8.
• EDWARD CURRY, SOULEIMAN HASAN, AND SEAN O’RIAIN. 2012. ENTERPRISE ENERGY MANAGEMENT USING A LINKED DATASPACE FOR ENERGY INTELLIGENCE. IN SUSTAINABLE INTERNET
AND ICT FOR SUSTAINABILITY (SUSTAINIT), 2012. IEEE, 1–6.

More References
• YAHOO! 2013. YAHOO! DIRECTORY: AUTOMOTIVE - MAKES AND MODELS. (2013). HTTP://DIR.YAHOO.COM/RECREATION/ AUTOMOTIVE/MAKES AND MODELS/
• KYLE ANDERSON, ADRIAN OCNEANU, DIEGO BENITEZ, DERRICK CARLSON, ANTHONY ROWE, AND MARIO BERGES. 2012. BLUED: A FULLY LABELED PUBLIC DATASET FOR EVENT-
BASED NON-INTRUSIVE LOAD MONITORING RESEARCH. IN PROC. SUSTKDD.
• RICHARD CYGANIAK. 2013. ROOMS IN THE DERI BUILDING. (2013). HTTP://LAB.LINKEDDATA.DERI.IE/2010/DERI-ROOMS

Tackling variety in event based systems

More Related Content

What's hot (12)

Viewers also liked (11)

Similar to Tackling variety in event based systems (20)

Recently uploaded (20)

Tackling variety in event based systems

Editor's Notes