Вы находитесь на странице: 1из 8

2013 IEEE 10th International Conference on e-Business Engineering

Towards an Efficient Sales Pitch with the Web of Things


Sujith Samuel Mathew

Yacine Atif

Quan Z. Sheng

Zakaria Maamar

School of Computer Science


University of Adelaide
Adelaide, South Australia
sujith@adelaide.edu.au

College of IT
UAE University
Al Ain, UAE
yacine.atif@uaeu.ac.ae

School of Computer Science


University of Adelaide
Adelaide, South Australia
michael.sheng@adelaide.edu.au

College of IT
Zayed University
Dubai, UAE
zakaria.maamar@zu.ac.ae

world things or things and their proprietary applications.


Moreover, any approach that is adopted must retain the
openness of the Web and build a generic framework that
encompasses things and their operations. To provide a
gradient scale to the problem we defined Ambient Spaces
(AS) to be homogeneous virtual representation of replicating
things in spatial patterns [3, 4]. For example, every parking
spot in a parking lot has the same type things like parking
sensors, lights, and LCD displays. Within an application
context each parking spot is represented as an AS with one
or more things and Web services are used to communicate
the states and functions of things. The integration of AS
creates the possibility of realizing the WoT.

Abstract New innovations in the Web of Things (WoT) and


the use of the social Web in business scenarios are paving the
path for real-world commodities or things to inherently sell
themselves to potential customers on the Web. An important
challenge to realize this vision of personalized sales initiated by
things is the adoption of massive instances of potential things
into the virtual world. We show in this paper how this issue is
alleviated by creating community of things. We describe how
the presentation of things on the Web as Web Smart things
supports such synergies. We also describe how a cluster of
things represented as Ambient Spaces (AS) increases the
possibility of identifying potential customers on the social Web.
The community framework is used to drive things into
populating the Web to reach relevant customer profiles leading
to friendships with things and people, which could nurture
future adoptions and purchase opportunities. Hence, these
social relationships between people and physical commodities
are exploited to elevate their value, promote adoption and
induce an opportunistic commercial transaction.

However, the realization of WoT by composing AS has


two main problems. Firstly, things need to be elevated,
abstracted and represented within an AS. Secondly, it would
be impossible to manually adopt the large proliferation of
Web-enabled things when they are inherently dynamic in
type, context and also possibly mobile. As an initial step
towards tackling these problems, we proposed an ontological
classification of things in AS. Our capability based
classification of candidate elements requires things to be
Web Smart to participate within AS [5]. We provide a
summary of this work in Section II, of this paper. Next, we
aimed at rapidly deploying a solution to realize an
application of our structure and study the feasibility of
realizing Web Smart things. We proposed a framework to
manage things within AS where we exploited the use of
existing Web technologies like HTTP and XML, and Wi-Fi
to realize it [4].

Keywords: Web of Things, Social Web of Things, Community of


Things, Emergent Semantics

I.

INTRODUCTION

Technology is disappearing into our surroundings with


computing becoming exponentially powerful, smaller, and
disappearing into the fabric of our everyday environments
leaving only interfaces to unleash inner information and
operational services. This trend generates real-time
information that helps businesses make quick operational
decisions. The impetus of this change is also because of the
falling cost of technology, the open application platform of
the World Wide Web (or the Web), and the increasing
availability of communication infrastructure like Wi-Fi,
broadband 3G, Cloud, and Radio Frequency Identification
(RFID). The emergence of the Web of Things (WoT) [1]
represents a paradigm shift in how Web users behold and
interact with the real-world. With cars that email their
owners about tires that need to be changed and sports
companies connecting their training shoes to the Web to
compare performances [2], the opportunity to transform the
value of products with augmented digital services boosts
product experiences and hence businesses. With things
communicating with other things on the Web and
participating in the decision-making process, novel
applications are foreseen to enhance businesses and society.

Having understood the feasibility and potential of the


WoT, in this paper we focus on two subsequent steps i.e., to
provide a semantic representation of things within an AS and
to automate the Web-based participation of things in new
business scenarios. The ability of things to group, re-group,
compose, or make decisions by themselves connote to
things behavior. The behavior of a thing within the
proximity of other things alters their group dynamics and
facilitates business functions like decision-making, selection,
tracking and logistics. With the propagation of Web Smart
things, an understanding of a things behavior mandates the
effectiveness of future businesses. We use the social
networking platform (social Web) to model things behavior
and discuss the benefits of these interactions within a
scenario whereby physical commodities intrinsically
organize themselves and navigate across the social Web
pathways to reach potential customers and incite purchases.

Though the WoT vision promises many opportunities, it


appears to be an impractical problem to realize when
contemplating the enormity and the heterogeneity of real978-0-7695-5111-1/13 $26.00 2013 IEEE
DOI 10.1109/ICEBE.2013.58

377

scalable framework to encompass the heterogeneity of realworld things and their functions by abstracting things within
AS [3]. While the adoption of real-world things into the Web
becomes a reality and the mashup of things are being
explored [10, 11], a suitable application platform for the
mashup of people, things and their interactions is now of
prime importance. The behavior of things in relation to their
social interactions needs to be studied for the benefit of
industry and businesses.
The social networking platform is ideal for the
composition and use of Web services [12]. The adoption of
the social networking platform to integrate real-world things
was proposed [13], but the approach is restricted to
exploiting the access restrictions offered by existing social
networking platforms to ensure privacy of things and their
users. Moreover, the approach does not look at things as
autonomous entities of the social network without human
intervention. Here, we propose to use the social networking
platform to realize the Social Web of Things (SWoT) and
create communities of things based on similarities found
among things. We also explore the behavior of things within
a community to enhance future business scenarios.

We present two scenarios where WoT directly affects the


future of supply chain management and product marketing.
A. Motivational Scenario
Sams Fast Cars (Sams) are car dealers. The business
decided to invest in a new software application that manages
the functions of the store which is built on a social
networking platform called Sams Ambient Space Manager
(SASM). The business wants to retain their existing
customers and is looking at selling them a recent model of
the cars that they currently own.
Things group themselves: In the morning a container with
50 cars arrived at Sam's. The showroom manager receives a
friend request from a social community in SASM called
Hon5A3Series. The manager inspects the request to see that
there are 50 members in the community where each member
is a new car and realizes that they are part of a shipment of
60 cars that were ordered last month. The store manager
accepts the friend's request and authorizes the unloading, and
preparations for the cars to the sold within the next few days.
Things market themselves: Today morning, Mr. Gibb
received a friends request to join a new virtual community
in SASM called Hon5A3Series. Mr. Gibb inspects the
friends request and is immediately taken to a page to view
the details of a new model of his sedan in the local outlet of
Sams. He is presented with the specification to study the
car. He also compares with other models in SASM. Mr.
Gibbs accepts the request and tags one of the cars with I
like the black one! and sends an invitation to Mrs. Gibbs to
check out his selection.
The above scenario represents an ideal situation where
our approach and framework is implemented and used. In the
first case it would have ideally taken three to four days to
unload, and prepare the new cars before the sales team can
contact potential customers. Moreover, the order was placed
for 60 cars and only 50 were received. We see here that the
lead time to initiate sale is reduced drastically as the cars
automatically grouped themselves to an existing community
of cars in SASM and declared their presence to the manager.
While in the second case we see that the cars have started
contacting potential customers themselves while they are
being unloaded. The information for contacting Mr. Gibb
was retrieved from an existing virtual community on SASM
where Mr. Gibb's and his present car are members. Such
scenarios boost the sales efforts of Sams as the cars
autonomously market themselves.
With WoT domain maturing, we foresee a complete
makeover of business functions like supply chain
management, marketing, and sales.

II.

GETTING REAL-WORLD THINGS READY FOR THE


SOCIAL WEB OF THINGS

A classification of real-world things harnesses the


complexity of modeling context-aware scenarios with a
common approach for interfacing with a wide range of
things. This would increase the ease of deploying things on a
large scale into WoT as Web resources and as Web services.
However, the classification of all things with potential
information to be shared on the Web is impossible, unless
they are based on the required capabilities of things to
potentially participate on the Web. Such a classification
provides a detailed understanding of the necessary
characteristics that would contribute to integrating them into
the Web.
A. Semantics of things on the Web
We proposed an OWL and SPARQL based ontological
approach [3, 4] to classify and capture the semantics of realworld things in an AS. The virtualization of real-world things
is achieved through a finite-state machine representation of
ontological
structure.
We
call
this
model
WebObjectMetadata (WOM), which defines concepts like
capability, location or friends of a thing in the form of OWL
classes as well as related SPARQL rules, which gear the
thing's behavior as a candidate for the social WoT. These
include available capabilities in some states, or reaction to
given inputs and the transition to a new state in WOM. This
model is instantiated through the specification of WOM
individuals, corresponding to the capabilities, context, state
and rules shaping their behavior as a new social thing. By
providing these details, this ontology enables social partners
to know about a things availability, capability, and when
and how to use it.
The metadata of a Web Smart thing is composed of
various ontological structures as shown in Figure 1. The
ontology for WOM-Annotations provides rich semantic

B. Related Work
The vision of ubiquitous computing [6] is being realized
with the abstraction of real-world things into the Web and its
effect is driving research trends towards realizing ubiquitous
computing [7, 8]. The impact that this has when integrating
with enterprise applications [9] has been suggested with
possible architecture to realize the benefits. However the
large-scale impact and the behavior of these things within
such environments are yet to be studied. We proposed a

378

Processing, Communication, and Storage, referred to as the


IPCS capability set. The WOM-Profile provides things
specification and integrates all variables defined by the
ontologies and divides the information as static and dynamic
descriptors of a thing.

content to capture user experiences and feedback. The


WOM-Location provides a record of how an object is traced
from the virtual space to its physical whereabouts. The
WOM-Capability ontology recognizes the four capability
dimensions of candidate elements to be Identity (ID),
http://samasm.com/people/Gibb.owl
WOM-Object

WOM-Annotations

s[0
ag
sT
WOM-Object

ha

hasFriend[0..n]

o
isL

cat

WOM-Processes
n]
0. .
[
s
s
ce
ro
sP
a
h
WOM-Communication
[0..n]
hasMedium
WOM-Capability
hasS
torag
e[0..
n]
WOM-Storage
hasID [1..n]

WOM-Location

..n
]

FOAF

ed

-In

isCapable-O

hasInformation
WOM-Profile

WOM-ID

Figure 1: Model of Web Object Metadata

updatable. Moreover, many real-world things might be


resource constrained. Hence the Web representation formats
must satisfy three criteria, (1) it must be understandable to
other Web Smart things, (2) it must be light-weight, and (3)
it must be represented for human perception. We represent
things states and functions in XML which ensures easy
interoperability between Web Smart things. The
representation in HTML enhances human perception of realworld objects. The dynamic context of real-world things is
reflected in an XML and then the HTML is updated in realtime. Both XML and HTML are light weight and provides
structured constructs for things representation. Ambient
Spaces provides a mashup of the Web-object Handlers of
various things within an application context. The WOM
provides the necessary elements for semantic structure of a
Web Smart thing within an AS which is represented in the
XML. An example of an XML file with few descriptors that
represent the WOM-Profile of a car from our scenario is
shown below.

The WOM-Capability ontology mandates the minimum


requirement for a thing to participate in an AS to be a unique
ID within the application context of the AS and refers to
things as Core when it has only an ID. The taxonomy refers
to things as Smart when it has all four IPCS capabilities.
However, our context requires virtualizing the capabilities of
things within AS for its use on the Web. For this we
extended our ontology to define things to be Web Smart. A
Web Smart thing inherits the dimensions of Smart thing i.e.,
using an object oriented connotation a Web Smart thing is-a
Smart thing. The IPCS capability dimensions take on a new
meaning when representing WOM-Objects in AS and are
described below.
Identity: A thing is uniquely identified with a URL. The
URL defines the namespace for all states and functions of a
thing. This is a mandatory requirement for things to
participate within AS.
Processing: RESTful adaptations of the URL provide
necessary semantics to access the state and operations of
things. Requests for state and operations of things are
processed and either a respective representation is returned
or converted to operations on the thing. Web services
provide interfaces to the state and functions of a thing.
Communication: Communication with a thing is defined
by the parameters that indicate the medium of interaction.
Within the context of Wi-Fi communication this would
define details like the SSID, the encryption methods, and
keys.
Storage: A thing must have the capabilities to cache Web
resources and status information. The storage could be
onboard or remote like on a private cloud. Parameters define
the access to the location and format of the information.

Real-world things are inherently dynamic and proprietary


in nature i.e., during the lifespan of a thing it adorns various
context values and also adapts to various ownership.
Moreover, things could also have static or fixed
characteristics like the manufacturer, date of manufacture, or
model number. Hence we propose that any semantic
representation of things must have two sets of elements, one
for the static information <wom: static> and the other for the
dynamic information <wom:dynamic>. As mentioned
earlier, the WOM-Profile defines this structure and captures
the essential semantic content to describe a thing. The
dynamic elements of a things profile are extendable to
contain information like tags that are given to things by the
consumers. WOM-Annotations define the structure for tag
organization and their interpretation.

The representation of Web Smart things should reflect


real-time scenarios and must also be retrievable and

379

<?xml version="1.0"?>
<rdf:RDF xmlns:rdf=http://www.w3.org/1999/02/22-rdf-syntax-ns# xmlns:moat=http://moat-project.org/ns#
xmlns:foaf=http://xmlns.com/foaf/0.1/ xmlns:wom=http://www.samsasm.com/wom#
xmlns:car="http://www.samsasm.com/car#">
<rdf:Description rdf:about="http://www.samsasm.com/dish/Omida">
<wom:profile>
<wom:static>
<car:itemcode>Car5AWS12</car:itemcode>
<car:country>USA</car:country>
<car:company>Ondah</car:company>
<car:year>2013</car:year>
</wom:static>
<wom:dynamic>
<wom:owner>Sams</wom:owner>
<wom:tags>
<moat:Tag>
<moat:name>Great Car!</moat:name>
<moat:hasMeaning>
<moat:Meaning>
<moat:meaningURI rdf:resource="http://dbappliance.org/resource/car"/>
<foaf:maker rdf:resource="http://socialnetpeople.com/gibbs"/>
</moat:Meaning>
</moat:hasMeaning>
</moat:Tag>
</wom:tag>
</wom:dynamic>
</wom:profile>
</rdf:Description>
</rdf:RDF>

A. Leveraging Ambient Spaces to build the Social Web of


Things
The real-world things augmented with necessary
capabilities transforms them into Web Smart things that
participate in AS. The collaborations and compositions of
AS create the Social Web of Things (SWoT). Several
platforms, such as SenseWeb [14] or Cosm1 have attempted
to provide a social platform to share and integrate sensor
feed. However, these approaches are not designed to support
decentralization and direct interaction with things, which is
essentially the focus of WoT. We propose the use of
similarities in things to match them and then group them into
communities to form the Social Web of Things (SWoT).
When considering things that are commodities the
process to create communities of things need to be modeled.
This will help understand the behavior of things within a
community and enhance the possibility of marketing them
among potential customers. For instance, in our scenario
when a new consignment of cars arrives from the supplier to
a warehouse, how do the cars form a community? Also, how
do the cars in the consignment become part of the products
in the warehouse? It is impractical for a person to sit and
enter all the information about each product. When a new
thing is introduced into an Ambient Space (AS), similarity
functions can be used to identify similarities in the WOMProfiles between the new things and the available things in
AS.

B. Enabling Things to be Represented on the Web


The software components that essentially represent Web
Smart capabilities of real-world things are illustrated in
Figure 2. A Web Smart thing is realized by augmenting a
thing with a tiny Web server and providing RESTful Web
services to interact with it. The Web-Object Handler handles
the requests for things services. The Adapter provides the
necessary drivers to interact with a things operations.

Web-Object
Handler

Control
and
Monitor

HTML

Update

State and
Functions

Adapter

View

RESTful URIs
Over Wi-Fi

XML
Web Smart
Tag

Control
Real-World
Thing

Web Smart Thing

Figure 2: Exposing Information and Control of a Real-World


Object as RESTful services

III.

THE SOCIAL WEB OF THINGS

The success of the existing social networking platforms


like Facebook, Google+, and LinkedIn is dependent on its
ability to populate communities based on social links that
exist between people. The social networking platform
suggests possible links between people based on similarities
that are assumed to exist between people. Real-world objects
whether tangible or abstract have heterogeneous properties,
but they can be inherently grouped based on similarities in
profile, spatial similarities, and social similarities.
Communities of real-world things can be quickly birthed in
an effective way to automatically populate social
communities. Creation of such communities is imminent for
the realization of the WoT vision and also enhances future
business and industry as we see in our scenario.

B. Community of Things formed by their Similarities


As illustrated in Figure 3, SWoT integrates people and
things within communities represented by AS. The profile of
1

380

www.cosm.com

a thing can be parsed to determine their context. Similarities


with other things within SWoT can be used to determine
their participation within an AS. A community of things is
formed using different types of relationships like
complementary relationship or similarity of things. We
determine the similarities between things to form
communities.
T3

T4

T2

T5

T16
T15

T14

T13

has within the Social Web are exploited to determine


possible adoption of things in AS3. Such associations are
bound to introduce privacy and security issues and the
various possibilities are discussed in [13]. AS3 community
integrates things based on the similarity in the <womdynamic> part of the WOM-Profile, for instance, the same
owner. The clustering process is also used to project
relationships with people in the Social Web. In this stage,
only things that are part of existing communities created in
first stage are considered. These AS are created with at least
one socially-connected member, for example T5 (manually
inserted), which acts as a seed or centroid for nurturing
social connections of future members. The clustering process
matches the WOM-Proles (<wom:dynamic>) of candidate
things and groups them into AS. If a new thing is not
matched into any cluster i.e., none of the WOM-Profiles of
cluster seeds match with the new thing's WOM-Profile, then
the thing is manually administered as a new seed to form a
new AS. Social connections to things are suggested to
members (people) on the Social Web based on the social
connections of the seed and in subsequent iterations the
social connections of other things in the AS are also used.
Hence, the clustering process here creates AS of things
which are related to social parameters like <wom:owner> or
<foaf:maker> found in the <wom:dynamic> part of the
WOM-Profiles. These social similarities are used to suggest
friend relationships to people on the Social Web.
Hence, SWoT furnishes a platform for people and realworld things to coexist. Here, we do not explain the process
of connecting to a specific social networking platform as this
has been studied and demonstrated previously [13, 15].
Instead, we model the creation of communities that automate
the process of identifying associations in the Social Web and
also reduce the overhead of manually registering all things
with social networking platforms.

AS 3

Social
Web

AS 131

T12
T11

T17

AS: Ambient Space


T: Thing
P: Person on Social Web
: Association between things and with people
: Association between people in Social Web

Figure 3: Platform Connecting things and People through Ambient


Spaces

In the first stage, we create AS like AS131 (see Figure 3).


AS131 community represents a cluster of things that are
similar i.e., T11 to T16. For example, the similarity attributed
to create AS131 could be the same manufacturer which is an
attribute found in the <wom-static> part of the WOMProfile. Clusters like AS131, created in the first stage, allows
the automatic categorization of a new thing like T17 into a
community like AS131 that has things similar to T17. These
groups are initiated with at-least one member (manually
inserted) which acts as a seed or centroid to adopt future
members. Similarities that exist between the static parts
(<wom:static>) of the things WOM-Profiles are used to
create clusters around the predefined seed. During a
clustering process, every new thing that is Web-enabled but
not in an AS is adopted into an AS by comparing similarities
of its WOM-Profile (<wom:static>) with the available cluster
seeds. If a thing is not matched into any cluster, then the
thing would be adopted during another clustering cycle or
manually administered as a new seed to form a new AS. The
seed in each AS provides a common representation for the
similar things that are grouped within them. The clustering
process ensures the re-election of the seed i.e., the clustering
process may change the centroid of the cluster. Hence, over a
period of time the seed becomes a purified representation of
a cluster. This stage is executed first, to ensure that all
possible candidates are part of existing communities. This
ensures that a new community is formed only after it is
initiated by authorized personnel, thus providing a level of
control whereby things are not allowed social associations
with people if they do not match into an existing community
of things.
In the second stage, we focus on similarities between
things like T16, T2, T3, T4, and T5, to create a community like
AS3. Our approach suggests relationships for all things in
AS3 with person P because of existing relationship between a
thing (T5) in the community and P. The associations that P

C. Clustering Algorithm
In our approach to cluster real-world things, we propose
the use of k-means algorithm [16, 17] to compare and cluster
the WOM-Profiles of Web Smart things. The k-means
algorithm is iterative and commonly used for clustering in
applications with large data sets. There have been various
attempts to improve the efficiency of k-means algorithm
[18]. Here, we are not focusing on the efficiency of the
algorithm but rather on the possibility of creating clusters
from large data sets that the WoT would generate.
The algorithm is used to partition a given set of WOMProfiles into k clusters. Here, the algorithm is applied to the
grouping of things that are represented by an n-dimensional
vector space. The algorithm clusters a set of vectors that
represent WOM-Profiles,  =  |  = 1, 2, ,  , where

, which is the  things in an AS.
In k-means, each projected cluster (AS) is represented by
a cluster centroid or seed. We denote the set of cluster
representatives as =   |  = 1, 2, ,  . The k-means
algorithm attempts to minimize the total Euclidean distance
between each thing   and its closest seed  . The algorithm
for clustering m un-clustered things is shown below.

381

Proc createCluster(dn)
Input: Cl: the set of clusters with similar static parameters
Input: Vm: the set of clusters with similar dynamic parameters
Input: $ : static similarity function
Input: % : content similarity function
Input: & : functional similarity threshold
Input: ': tag-based similarity threshold
Auxiliary: i, j, k: integer
Begin
- Initialize k seeds from W
- Use these k seeds as initial set of cluster representatives C
For each wi do
If (  / $ ( ,  ) > &) then
-  =  { }
- Calculate cluster mean cj
Else
If (- // % ( , - ) > ') then
- - = - { }
- Calculate cluster mean vk
Else
= { } and / = / { }
End if
End if
End for
End

Algorithm 1: Creating cluster of things


In general, the higher the purity value, the better the quality
of the cluster is. For every new candidate added to a
community of Web Smart things, we choose the community
that delivers the maximum purity as follows:

Formally, given a threshold , a similarity function ,


and Web Smart things to clusters   , the algorithm
considers the WOM-Profile of each thing  , in turn, and
calculates the corresponding similarity (  ,  ), for each
existing cluster  ,  = 1, 2, , . If no matching cluster is
found (i.e. ( ,  )>, j=1,, ), we increment  by 1 and
create a new cluster ci for wi. Alternatively, wi is assigned to
cluster cj with the highest (wi, cj).

Purity ( , ) =

 max   

Where  = { , . ,  } is the set of all Web Smart


things in an Ambient Space, = {  , . ,  } is the set of
existing clusters or communities of Web Smart things, where
each cluster  contains a set of similar Web Smart things of
W and finally, E = { , . ,  } is the set of tags where each
e represents a set of Web Smart things containing tag e .
We employ a simple term frequency approach to clustering
as follows.

D. Optimizing Content Similarity


The similarity (, ) between a thing w and a cluster c
is derived through comparing the tags of ws WOM-Profile,
to those of other WOM-Profiles in c.
A trivial similarity method to compare w to every object
in the cluster c based on an average score could be defined as
follows:
  (,  )

(, ) =
| |
However, this is not an efficient approach. Instead, we
adopt a quality metric of the clustering, which includes Web
Smart things tags, and then cluster a thing based on the
maximum induced quality.

Let tf (w, t) denote the frequency of term t T in Web


Smart thing w W, where T={t1,...,tm} is the set of distinct
terms occurring in W. T represents the value of tags or the
actual data of the Web Smart things. A thing is then
represented as an m-dimensional vector """
! as follows:
"""
! = (tf(d,t1),...,tf(w,tm))
Where, tf (w, t) denote the frequency of term t T in
thing w W. We use Cosine similarity, which corresponds
to the correlation between the vectors of distinct Web Smart
things. This is quantified as the cosine of the angle between
these two vectors. Cosine similarity is one of the most
popular similarity measure applied to text documents and

Purity measure is commonly used in clustering, due to its


accuracy, robustness, and universality [19]. The purity
measure evaluates the coherence of a cluster, that is, the
degree to which a cluster contains similar Web Smart things
from a single category. For an ideal cluster, which only
contains things from a single category, its purity value is 1.

382

were able to merge things into the existing communities and


attain different levels of purity. Clusters with higher purity
(~1) were more accurate in reaching out to the existing
customers.

used in numerous information retrieval applications and


clustering, as well, as follows.
Given two things t"""0 and """
t 3 , their cosine similarity is:
t"""0 . t"""3
4567 (t""""
0 , t"""
3) =
"""
t 3  t"""3 

In the scenario where things themselves initiate sales in a


social Web context, more number of customers is reached
when clustering is employed and the number of pure clusters
created increases in a gradient. Using a number of iterations
and fixed number of hundred customers we averaged the
number of customers reached with different level of purity.
Our simulation indicates the benefit of creating pure
communities of things to enable business functions like
product sales. Figure 4 illustrates how pure clusters reach out
to more number of customers.

Where """
t 0 and t"""3 are m-dimensional vectors representing
two distinct Web Smart things over the term set T = {t1 , . . . ,
tm } associated with all Web Smart things.
IV.

DISUCSSION AND PRELIMINARY EVALUATION

The possibility of reaching out to all existing customers


with new products as suggested in our motivational scenario
would be difficult for human salesmen, because it requires
keeping track of various criteria like, the duration of
customer association, the credit standing of the customer,
and the service records. Moreover, these criteria changes
each day. A software application could manage these
requirements but the sales would be triggered only after the
details of new commodities are manually entered into the
system. In our approach, the commodities (things), group
themselves into communities and initiate a sales pitch. This
reduces the efforts for the salesmen and also increases the
possibility of a successful sale.

Existing Customers

100

Pure Clusters reach more Customers

80
60
Customers
Reached

40
20
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

With real-world commodities becoming active salesmen


and automatically reaching out to existing customers, the
possibility of retaining existing customers increases. If each
thing initiated a communication channel themselves on the
social platform this would create a viral phenomenon with
the plethora of things autonomously contacting potential
customers. The ability of things to cluster themselves into
communities would reduce the effect of such viral marketing
as only one member of a pure cluster needs to represent the
community of things. This is ideal if all members of a cluster
are strongly related to each other. Our similarity functions
and clustering algorithm enable the creation of clusters of
things, and therefore create pure communities.

Cluster Purity

Figure 4: Pure clusters reach out to more customers.

V.

CONCLUSION

Abstracting real-world things as Web Smart things and


creating communities of things enables business functions
like product sales. We presented here the creation of things
communities and the use of these communities to drive
commodities into populating the social Web to reach
relevant customer profiles. The friendships between things
and people could nurture future business transactions. These
social relationships between people and physical
commodities are exploited by Web Smart things to elevate
the value of commodities, promote adoption and induce an
opportunistic commercial transaction. The behavior of a
thing within the proximity of other things alters their group
dynamics and facilitates business functions like decisionmaking, selection, tracking and logistics. We presented the
creation of things semantic representation within AS using
our ontological structure and discussed the role of various
stakeholders in creating relevant content. We also presented
the process of creating things communities by clustering
things based on similarities. We evaluated the benefits of
creating pure communities to reduce viral marketing and
reach more customers thereby enhance the efficiency of the
sales team.

Since our research domain is still in its initial stages, it is


difficult to get a real-time sample data of Web Smart things
and therefore we decided to simulate the generation of the
WOM-Profiles. We manually created ten different clusters.
Each cluster had ten things of the same type, i.e. the <womstatic> parts were the same in each cluster. These things had
different owners defined in the <wom:owner> tag in the
<wom-dyamic> part of the WOM-Profiles. These clusters
represented the benchmark for pure clusters that we wanted
to test against.
Next, we created a corpus of distinct words for static
content (type of things). To represent new things, WOMProfiles were generated by randomly picking values from the
static corpus where each profile represented a Web Smart
thing. Next, we applied our algorithm to cluster these new
things into existing clusters i.e. by matching profiles that had
similar <wom-static> content. By varying the threshold we

We continue our research focus on studying the various


aspects of things behavior within different types of

383

[10] D. Guinard and V. Trifa, Towards the Web of Things: Web

communities. The benefits of these relationships and their


impact on industry and business are yet to be researched.
REFERENCES
[1]

[2]

[3]

[4]

[5]

[6]
[7]

[8]

[9]

[11]

B. Christophe, M. Boussad, M. Lu, A. Pastor, and V.


Toubiana, The Web of Things Vision: Things as a Service
and Interaction Patterns, Bell Labs Technical Journal,
16(1):55-62, 2011
P. Olson, Googling Your Lost Keys, And The Coming
Revolution
In
Smart
Products,
Forbes
Tech,
http://www.forbes.com/sites/parmyolson/2012/09/14/googling
-your-lost-keys-and-the-coming-revolution-in-smartproducts/. Retrieved on 01/04/2013
S.S. Mathew, Managing Things in an Ambient Space, In
the Proceedings of the 9th International Conference on
Service Oriented Computing Workshops (ICSOC11),
Paphos, Cyprus, 2011.
S.S. Mathew, Y. Atif Y., Q. Sheng and Z. Maamar Z.,
"Building Sustainable Parking Lots with the Web of Things",
Personal and Ubiquitous Computing (PUC) Journal, Springer.
2013. Accepted, to appear.
S.S. Mathew, Y. Atif, Q.Z. Sheng, and Z. Maamar, The Web
of Things Challenges and Enabling Technologies, In
Internet of Things and Inter-Cooperative Computational
Technologies for Collective Intelligence, Studies in
Computational Intelligence book series, Springer 2013, 1-23.
M. Weiser. The Computer for the 21st Century. Scientific
American, 265(3):94104, 1991
D. Raggett, The Web of Things: Extending the Web into the
Real World, SOFSEM 2010: Theory and Practice of
Computer Science, 2010
D. Guinard, V. Trifa, S. Karnouskos, P. Spiess, and D. Savio,
Interacting with the SOA-Based Internet of Things:
Discovery, Query, Selection, and On-Demand Provisioning of
Web Services, IEEE Transactions on Services Computing. 3,
223235, 2010.
P. Spiess, S. Karnouskos, D. Guinard, D. Savio, O. Baecker,
L. M. S. d. Souza, and V. Trifa, SOA-based integration of
the internet of things in enterprise services, In the
Proceedings of the 2009 IEEE International Conference on
Web Services, (ICWS 09), IEEE Computer Society, pp. 968
975, 2009

[12]

[13]

[14]

[15]

[16]

[17]

[18]

[19]

384

Mashups for Embedded Devices, In the Proceedings of the


workshop Mashups, Enterprise Mashups and Lightweight
Composition on the Web (MEM09), 2009
N. B. Priyantha, A. Kansal, M. Goraczko, and F. Zhao, Tiny
web services: design and implementation of interoperable and
evolvable sensor networks, in the Proceedings of the 6th
ACM Conference on Embedded Network Sensor
Systems(SenSys08). New York, NY, USA: ACM, pp. 253
266, 2008
Z. Maamar, H. Hacid, and M.N. Huhns, "Why Web Services
Need Social Networks", Internet Computing, IEEE , vol.15,
no.2, pp.90-94, 2011
D. Guinard, M. Fischer, and V. Trifa, Sharing Using Social
Networks in a Composable Web of Things, In the
Proceedings of the IEEE International Workshop on the Web
of Things (WoT 2010), Mannheim, Germany, 2010.
W.I. Grosky, A. Kansal, S. Nath, J. Liu, and F. Zhao.
Senseweb: An infrastructure for shared sensing.
Multimedia, IEEE, 14(4):8-13, 2007.
J. Mitchell-Wong, R. Kowalczyk, A. Roshelova, B. Joy, and
H. Tsai. Opensocial: From social networks to social
ecosystem. In Digital EcoSystems and Technologies
Conference, 2007. DEST'07. Inaugural IEEE-IES, pages 361366. IEEE, 2007.
R. Xu, D. Wunsch, et al. Survey of clustering algorithms.
Neural Networks, IEEE Transactions on, 16(3):645-678,
2005.
Z. Huang. Extensions to the k-means algorithm for clustering
large data sets with categorical values. Data Mining and
Knowledge Discovery, 2(3):283-304, 1998.
T. Kanungo, D.M. Mount, N.S. Netanyahu, C.D. Piatko, R.
Silverman, and A.Y. Wu. An effcient k-means clustering
algorithm: Analysis and implementation. Pattern Analysis
and Machine Intelligence, IEEE Transactions, 24(7):881-892,
2002.
A. Huang, Similarity Measures for Text Document
Clustering, In the Proceedings of the Sixth New Zealand
Computer
Science
Research
Student
Conference NZCSRSC2008,
Christchurch,
New
Zealand (2008) , p. 4956, 2008

Вам также может понравиться