Customer Perception Driven Product Evolution
Facilitation of Structured Feedback Collection
Oleksiy Khriyenko
Industrial Ontologies Group, Agora Center and Department of Mathematical Information Technology,
University of Jyväskylä, P.O. Box 35, FIN-40014, Jyväskylä, Finland
Keywords: Structured Feedback Collection, Triple Generation Support, Semantic Customer Feedback, Semantic
Personalization, Product Customization, Customer Analytics.
Abstract: Competitive environment not only requires effective advertising strategies from the product producers and
service providers, but also to do comprehensive and sufficient analysis of their customers to understand
their needs and expectations. Successfully involving customers into a product/service co-creation process,
companies more likely increase their future revenue. Customer feedback analysis is widely applied in
marketing and product development. Among other challenges (e.g. customer engagement, feedback
collection, etc.) automation of customer feedback analysis becomes very demanding task and requires
advance intelligent tools to understand customers’ product perception and preferences. Since, mining of free
text feedbacks (which is still the most representing form of the real voice of the customer) is challenging,
this work presents an approach towards customer-supported transformation of feedback into structured data.
Further analysis and manipulation with semantically enhanced customer feedback and product/service
description makes possible to automatically generate useful changes in existing products or even a new
product description that takes into account actual needs and preferences of customers.
1 INTRODUCTION
For years, many companies have been implementing
data mining tools and methods to discover various
indicators to be predictive about and proactive with
their customers. Along with data mining tools,
organizations implement a variety of technologies to
analyse customer data, including: business
intelligence (BI) tools, data discovery and
warehouses, predictive analytics, data and text
mining, social media analytics, customer data and
behaviour management, and more. Becoming
“customer centric” is a top priority for companies in
highly competitive markets nowadays. Companies
experience customer analytics technologies to find
best practices to improve customer intelligence,
reach customer-centric goals and increase their
revenue through customer satisfaction growth.
Customers do much more than simply buy products
or services. With digitalization of communication
channels, they are able to discuss and share
information about a product through social media
networks, to write reviews and leave comments to
reviews of others.
Nowadays, many of product developers and
service providers, as well as retailers, are pretty
much focusing at pushing their products to the
customers, making them willing to buy. Consumers
will not buy a product under uncertainty: whether
product matches their preferences or whether quality
is good as it is advertised. Since, consumers are
more likely to rely on other users than on seller-
generated information or even third-party experts,
consumer review systems have become extremely
popular, and have been found to affect the sales of a
variety of products (Chen and Xie, 2008; Chevalier
and Mayzlin, 2006; Ghose and Ipeirotis, 2011;
Ehrmann and Schmale, 2008). Definitely, intelligent
strategies towards proper selection of user’s reviews
to be shown for customer to increase its willingness
to buy may bring expected benefit. But, many
customers more likely would like to buy a product or
service that satisfies customer’s needs and
expectations rather than just simply choose
something suitable from what is offered. Advanced
analytics technologies, modern data management
and integration systems are enabling organizations
to gain far greater depth and breadth of knowledge
196
Khriyenko, O.
Customer Perception Driven Product Evolution - Facilitation of Structured Feedback Collection.
In Proceedings of the 12th International Conference on Web Information Systems and Technologies (WEBIST 2016) - Volume 2, pages 196-203
ISBN: 978-989-758-186-1
Copyright
c
2016 by SCITEPRESS – Science and Technology Publications, Lda. All r ights reserved
about customers; to understand sentiment drivers; to
identify features for the best segmentation; to
measure brand reputation; to make faster strides in
predicting retention, attrition, and return rates; to
discover patterns and anticipate customers’ problems
with products or services. Outcome of such
analytical tools is further used by domain experts to
develop new marketing and advertisement strategies,
to improve existing and to create new products and
services.
Product and service providers always aim to
attract new customers. At the same time, reducing
customer churn is a very crucial problem for most of
the companies who are doing business in highly
competitive environment. Customers are different.
They have own perception, own opinion and own
preferences. Therefore, not always actual product or
service meets personal expectations of a customer.
Big difference in ideal, actual and perceived
product/service may cause customer loss.
Automated distance measuring between ideal-actual-
perceived products will allow us to predict the level
of customer satisfaction and analyse a risk of that
loss. Using various methods and approaches,
companies are trying to engage customer to the
feedback provisioning process to be able to hear the
real voice of a customer. To apply intelligent
analytics on top of customer feedbacks, they should
be converted into a structured machine readable
form. Therefore, we have to elaborate techniques
and tools that automate customer feedback gathering
in structured form that allows its further automated
intelligent analysis. This will speed up
corresponding response towards customer
expectations from product/service providers and
facilitate customer involvement into development of
new products and services via shearing own
preferences and ideas.
Thus, the paper proposes a solution for
facilitation of structured customer feedback
collection process and automated generation of
preferable/ideal product and services from a
customers’ perspective. It will expand automated
part of a product adaptation and evolution lifecycle,
and allow system to perform responsive product
optimization. In other words, we try to facilitate
adaptation of products and services towards
customers’ expectations. The next section (Section
2) concerns the main contribution of the paper and
presents: motivation for structured feedback
collection; semantically enriched digital content as
enabler for customer-driven preference definition;
and the mechanism for triple-based feedback
collection. Section 3 presents the ways structured
feedback could be used to facilitate further product
development and targeted advertisement. Finally, the
paper shortly refers to the planned future
enhancement of proposed solution, as well as
concludes the proposed work.
2 STRUCTURED CUSTOMER
FEEDBACK COLLECTION
Customers are a wealth of information. To create
successful innovative business and succeed with
new product or service, company must listen to the
real voice of a customer based on analysis of
customer feedbacks. Collecting human knowledge
has been a common goal since ancient times. With
appearance of the Internet and the World Wide Web,
people have got a possibility to generate and share
almost unlimited amount of information. Therefore,
automated arrangement of user-generated datasets is
crucial. Since ordinary people mainly generate data
in text-based unstructured form, transformation of
this data into structured datasets enables further
automated processing by machines.
Many collaborative systems, having own internal
data models, insist users to create data accordingly.
For example, the Wikipedia articles are further
organized by contributors into structured
information, including keyword lists in the infobox,
classes, and directories. Their structured information
is entity based and follows a predefined schema.
With a purpose to not frighten users away with
possible complexity of internal data model and to
make system more user-friendly, automated text-
analysis techniques are used instead. But, since
customer feedback analysis domain is very sensitive
to possible mistakes and low efficiency of automated
natural language processing, fully automated text
analysis could become a bottleneck of a system. It is
very hard, if not impossible, for any automatic
technique to achieve perfect accuracy due to the
difficulty of natural language understanding.
Systems that need near-perfect solutions require
convenient user-friendly mechanism for human
involvement to correct errors made by automatic
techniques. Even, involvement of a domain expert
(which can be costly) into the process at the stage of
feedback analysis does not guarantee that a real
voice of a customer will not be distorted. It is much
more reasonable to ask user (customer) what (s)he
meant, rather than to ask some external expert about
the same later. Therefore, we still need to engage
customer into the process of structured feedback
Customer Perception Driven Product Evolution - Facilitation of Structured Feedback Collection
197
provisioning, and it is a real challenge to make it
unobtrusive and attractive for him/her.
According to the analysis of various methods and
strategies that help to make a customer willing to
provide a feedback (Khriyenko, 2015a), to get
comprehensive feedback or suggestion from
customers, company should target those customers
who are interested in product/service improvement
and who believe that his/her feedback (suggestion,
preferences, etc.) will be taken into account.
Customers are interested in co-creation of a new
product/service (that is about to meet their
expectations and preferences) as well as concerned
with improvement of existing services they use.
Therefore, it is preferable to allow customer to
provide a feedback/comment at the moment and the
place considered by customer as the most
appropriate one. Customer should be able to make it
for things (product/service parts, properties or
functionalities, etc.), which are considered as
important for him/her. There are a lot of customer
feedback support systems which are mostly based on
predefined feedback forms (Kampyle, OpinionLab,
PollDaddy, Feedbackify, Survey Monkey,
Zoomerang, Survey Gizmo, etc.). Since the forms
are created by (or according to) product/service
providers, they consist of issues that are important
for them, and only unstructured free text forms
become a place for actual concerns of a customer.
Therefore, talking about digital environment, we
have to provide a possibility to a customer, through
pointing at any part of visual representation of a
product or via selecting certain concept or piece of a
text related to it, to access feedback provisioning
tool with respect to associated feature/functionality
of the product or service.
Reference to product’s features and
functionalities implies existence of domain data
model and structured product description. Semantic
Web technology (Berners-Lee et. al., 2001;
Semantic Web, 2001) allows us to define domain
knowledge and data model in structured form via
Ontology. Then products’ descriptions, as well as
feedbacks provided by customers, could have
structured RDF based representations. Definitely, we
cannot expect average customer be a knowledge
modeling expert able to provide semantically
annotated content. Therefore, it is crucial to support
feedback providers with easy to use interfaces that
simplify the annotation process, placing annotation
in the context of their feedback provisioning
process. At the same time, all the digital content
should be enhanced with possibility to provide a
feedback on-the-fly. Talking about web content, we
facilitate it with extra JavaScript functionality that
implements semantically facilitated structural
feedback provisioning.
2.1 Structured Feedback Collection -
Enabled Digital Content
Semantic Feedback -enabled Digital Content is a
digital content which is facilitated by JavaScript
based functionality (SCF.js plus SCF.css) and
optionally contains already semantically annotated
pieces of texts and visual elements. The idea of such
annotation is not to just present the same
information in semantic form and put it in line with
an initial content. For this purpose we may use
RDFa notation (Adida et. al., 2015) for example.
The idea is to imbed entity annotations to provide
more intuitive and accurate guidance for customers
during feedback provisioning stage later. The
mechanism for customer driven structured feedback
provisioning allows user to select peace of a text or
visual element of the page, and access to the
associated with it product features (presented in
domain ontology and description of the product).
With a text fragment selection, tool performs
semantic annotation of the text. The annotation
module is built using Stanford NLP (Manning et. al.,
2014) and GATE (Cunningham et. al., 2002)
Enabled Java libraries that allow text annotation
using domain ontology and product/service
description in RDF as an annotation schema. As a
result, user sees a list of associated concepts/entities
from ontology and products descriptions in the order
of their relevance. To avoid errors of an automated
annotation, some annotations of digital content (text
or visual elements) could be imbedded to the
Semantic Feedback -enabled Digital Content by
product provider in advance. It could be done via
Semantic Feedback -enabled Digital Content
Creation tool (SFeDCC tool). In this case, pre-
annotated concepts (associated with corresponding
selection) will get higher priority in the list. At the
same time, imbed entity annotations are required in
case we associate them with visual elements of our
content, since automated extraction of any semantics
from visual content is very complicated task.
Therefore, visual content should be enriched with
corresponding annotations via SFeDCC tool at the
digital content preparation stage.
SFeDCC tool helps to make digital content
Semantic Feedback -enabled via enrichment with
semantic annotations. Using the tool, product or
service provider, who is versed in the product more
than any customer, can edit automatically created
WEBIST 2016 - 12th International Conference on Web Information Systems and Technologies
198
Figure 1: Semantic enrichment of textual content.
annotation more efficiently. Tool highlights
automatically recognized concepts/instances and
allows expert to check it and make necessary
corrections. At the same time, expert has a
possibility to manually annotate text fragment if
concepts/instances are not recognized automatically.
Tool simply allows to select a visual element or text
fragment and provides graph based concept browser
to choose an appropriate instance from connected
semantic products descriptions or corresponding
entity from domain ontology in RDF format (these
sources are added to the script via addSource()
method). The browser is enhanced with text-based
search functionality to facilitate navigation inside a
graph. Corresponding annotation is imbedded into
the initial html page via html <span> element with
data-scfURI attribute. Figure 1 shows us an example
of initial text that describes a “kettle” product and
semantically annotated version of corresponding
Semantic Feedback -enabled Digital Content. Result
text is split up by <span> elements that bring
semantic annotation to automatically recognized or
manually selected entities.
In case of image annotation, tool gives a
possibility to select particular area of the image and
provides the same annotation tool with graph-based
browser as it was in the case of text annotation. As a
result, corresponding html image element is
extended to image map structure with data-scfURI
attribute that refers to annotation class or instance
from domain ontology or RDF based product
description (see Figure 2). If an image has a link, it
will be repeated in each image map area to be
visible.
2.2 Triple Based Feedback Collection
Depending on user’s selection, system creates a list
of entities (instances, classes, properties and their
values) associated with it. To be more intelligent and
user friendly, system range the list in the order of
entity relevance. Assuming that everything is
represented in RDF triple based structure “subject-
predicate-object”, system presents customer
feedback as a triple based description of an ideal
product. Since a customer feedback is a form of
expression of his/her own opinion about product
properties or functionalities (which are product
properties as well), in most of the cases, user
selection will cover “object” (the value of
property/functionality) and less it will cover
“predicate” as such. If we take a look at product
description in our example (see Figure 1) and decide
to provide a feedback about properties of the kettle
and its parts (their colors, materials they are made
of, the length of the kettle’s dock station’s wire,
etc.), most probably we select such fragments of
texts like: “blue”, “metal” or “made of metal”,
“plastic handle”, “black dock station”, “length” and
“0.5 meters”. In most of the cases they are “objects”
(values) or “subject-object” pairs, and only few of
them consist of “predicate” (e.g. “length”, “made
of”). It means that system should automatically
derive missing parts of triples using Ontology and
<p>Mainbodyofthekettleisblueandmadeofmetal. Kettlehasgrayplastichandleandblackdockstation.Thelength
ofthedockstationwireis0.5meters.</p>
<head>
…
<linkrel="stylesheet"type="text/css"href="http://scfDomain.com/scf/scf.css">
<scripttype="text/javascript"src="http://scfDomain.com/scf/scf.js"></script>
</script>
addSource(http://example.com/domainOntology.n3);
addSource(http://example.com/productDescription.rdf);
</script>
</head>
…
<p><span
data‐scfURI="pD:mainBodyOfKettle1">Main body of the kettle</span> is <span data‐
scfURI="dOnt:blue">blue</span> and made of <span data‐scfURI="dOnt:metal">metal</span>. <span data‐
scfURI="pD:kettle1">Kettle</span> has <span data‐scfURI="dOnt:gray">gray</span> <span data‐
scfURI="dOnt:plastic">plastic</span> <span data‐scfURI="pD:handleOfKettle1">handle</span> and <span data‐
scfURI="dOnt:black">black</span> <span data‐scfURI="pD:dockStationOfKettle1">dock station</span>. The <span
data‐
scfURI="dOnt:hasLength">length</span> of the <span data‐scfURI="pD:wireOfDockStationOfKettle1">dock
stationwire</span>is0.5meters.</p>
Initial Di
g
ital Conten
t
E
nriched Di
g
ital Conten
t
Customer Perception Driven Product Evolution - Facilitation of Structured Feedback Collection
199
Figure 2: Semantic enrichment of visual content.
RDF based descriptions of the products, as well as
take into account availability of potential candidates
in the text. In our example, we have a text with
detailed description of the product, and of course, it
contains a lot of “objects” and “predicates”. In case,
when page does not present product description, but
just has a mention of a product, selection brings us
only a “subject”. Then, it become much challenging
to guess user’s intention, the issue (s)he would like
to comment on. We cannot expect that user has any
experience with knowledge engineering and is able
to compose triple based feedback through browsing
domain ontology and product descriptions as it could
be done by expert via SFeDCC tool. Thus, system
tries to derive triples on-the-fly while user provides
a feedback in a free text form.
Figure 3 presents triple based feedback provision
supporting tool. The purpose is to help user to
generate triple based feedback by presenting it in a
human readable form. For example, if we select
“plastic handle” fragment of the text, system
recognize “handle” as a part of the kettle product
and derive corresponding URI of the resource
(pD:handleOfKettle1) as a subject for the triple. At
the same time, associating word “plastic” with
corresponding instance (dOnt:plastic) of the class
dOnt:Material in ontology, system derives most
suitable predicate for the triple (dOnt:madeOf),
based on domain and range definitions of the
property and available records in RDF product
description (pD: and dOnt: prefixes refer to the
namespace of product related definition and domain
ontology correspondently). To present a triple in
human readable form, system use text based
representation associated with corresponding entities
in ontology via scf:hrForm property (scf: prefix is
used to represent semantic customer feedback
ontology). It is not the same as rdfs:label property,
even though it might also have the same value. The
value of this property is not just a meaningful and
human readable part of entity’s URI (e.g “color”,
“made of”, etc.), but is the most suitable textual
representation of entity to make textual version of a
triple more natural for human (e.g. triple like
“handle - is - gray” instead of “handle - color -
gray”). If all three components of a triple are
derived, then system represents the triple and
expects changes of “object” from the user. If only
subject component was recognized by system, then
it is assumed that user will define predicate and
assign the value (object). Search of appropriate
components of a triple is done via analysis of textual
input from the user. Depending on what user is
typing in the text field, system modifies the order of
suitable entities to be chosen from the list. Any time,
when user would like to change the current status of
the triple’s component and deletes it from the triple,
system shows him/her a list of suitable options and
uses user’s input from the text field as a context to
order the list. If user would like to reset whole triple
and remove all three components, then text field is
associated with the whole triple rather than with
particular parts of it. Very often feedback consists of
relative evaluation rather than absolute meaning.
<ahref="http://www.someLink.com"target="_blank">
<imgid="img_kettle"src="kettle.png"width="150"height="200"alt="Kettle">
</a>
<ahref="http://www.someLink.com"target="_blank">
<imgid="img_kettle"src="kettle.png"width="150"height="200"alt="Kettle"
usemap="#kettleMap"data‐scfURI="pD:kettle1">
</a>
<mapname="kettleMap">
<areaid="img_kettle_area1"shape="rect"coords="0,170,150,200"alt=""
onmouseover="scfShowMenu(this);"onmouseout="scfHideMenu(this);"
data‐scfURI="pD:dockStationOfKettle1"href="http://www.someLink.com"target="_blank">
<areaid="img_kettle_area2"shape="poly"coords="30,40,100,40,100,170,30,170"alt=""
onmouseover="scfShowMenu(this);"onmouseout="scfHideMenu(this);"
data‐scfURI="pD:mainBodyOfKettle1"href="http://www.someLink.com"target="_blank">
<areaid="img_kettle_area3"shape="poly"coords="30,10,150,10,150,170,100,170,100,40,30,40"alt=""
onmouseover="scfShowMenu(this);"onmouseout="scfHideMenu(this);"
data‐scfURI="pD:handleOfKettle1"href="http://www.someLink.com"target="_blank">
…
</map>
Initial Di
g
ital Conten
t
E
nriched Di
g
ital Conten
t
WEBIST 2016 - 12th International Conference on Web Information Systems and Technologies
200
Figure 3: Structured feedback collection interface.
Customer may say that the wire of the dock station
is too short for him/her rather to define concrete
length of it. Therefore, ontology should consider not
just absolute values (e.g numerical value, color’s or
material’s names, etc.), but also define relative
values or values defined via its property (e.g.
“longer”, ”too small”, “stronger”, “more sticky”
“softer”, etc. Thus, customer may express his/her
filling about the product. For example, customer
may say that “handle is too dark” meaning that (s)he
would prefer more lighter color of it, or say that “the
main body of the kettle is too hot” meaning that
(s)he prefers it made of some other material with
smaller heat conduction. Based on such feedbacks
system can build customer perceived and preferable
(ideal) versions of a product description.
3 CUSTOMER FEEDBACK
DRIVEN DESIRED/IDEAL
PRODUCT GENERATION
Having new preferable values for product properties
or new set of useful features based on customer
feedbacks, theoretically we may use them to
substitute initial ones in the product description.
Unfortunately, we cannot just simply populate initial
product description with new values (perceived or
preferable) of the properties, since such values might
not exist in the domain ontology and their use might
be restricted by property constrains. Therefore, we
introduce scf:CustomerPreference class (as a class
similar to rdf:Statement) to represent triple based
feedback of the customer. Similarly to RDF
Statement, each instance of this class has three main
properties: scf:subject, scf:predicat and scf:object.
Additionally, it might have optional property
scf:domain that refers to either particular instance
(product or its part) or more general class of things
(in case we represent aggregated customer
preference with respect to a class of products or a
concept). This extra property may simplify further
aggregation of preferences. scf:subject and
scf:predicat properties refer to corresponding
instance of a product description and its property. In
turn, scf:object may refer to an entity from a domain
ontology and represent preferable for customer
value, or to an entity of semantic customer feedback
ontology that extends a set of usual values with such
uncertain, relative values as: scf:longer,
scf:stronger, scf:hotter, scf:lighter, etc.
As you noticed, transformation from customer
perception (that could be expressed via free text
during feedback provisioning (see Figure 3) to
customer preference (which is actually submitted by
customer through his/her feedback) is done
automatically by system and only preference related
value is suggested for “object” value to be assigned.
In Figure 3 user expressed his/her perception about
the wire’s length of the dock station, which is too
short in his/her opinion. System makes
corresponding transformation and suggests the most
suitable value (“longer”) to describe user’s
IPREFER
1m
1.5m
2m
longer
Thelengthofthedockstation’swireis
tooshort
kettlebody
ismadeof
handle’s
coloris
lengthis
wire’s
Customer Perception Driven Product Evolution - Facilitation of Structured Feedback Collection
201
preference - “wire’s - length is - longer”. Thus,
when user submits this preference, corresponding
instance will be created (see
cp:customerPreference_03 in Figure 4).
Further, based on such customer preferences and
initial description of the product, we are able to
generate corresponding description of a desired
product. Knowing initial values for the properties,
system may reason about relative values in the
customer preference description and select
reasonable values from existing options. Uncertainty
of the values makes an advantage when we perform
collaborative generation of potentially the best
product configuration in the context of preferences
of all the customers who belongs to the selected
target group. In this case, if some customers
expressed the preference to have lighter color and
some customers specified exact values, system is
able to calculate average value using color
distribution palette. With respect to the materials’
properties (e.g. hardness, durability, etc.), the
absolute value (the name of material) could be
calculated based on suggested values (relative and
absolute) and corresponding properties from the
ontology of materials.
Assuming that RDF based description of the
product/service is used for software-driven
automation process and could not be considered as a
proper final result used by human (decision maker),
decision support system should present results in
more human readable form. Therefore, we need
visualization tools that are able to build visual model
of a product based on its RDF description. Some
tools just would need to transform our RDF based
product description into internal format of the tool.
In more challenging cases, when full transformation
Figure 4: RDF based customer preferences representation.
is not possible due to limitation of the tool’s data
model, we might need to apply more advance
resource visualization techniques. For example, one
of the relevant researches (Khriyenko, 2015b)
presents an approach towards automated creation of
semantically personalized user interface. Something
similar might be applied to build domain
independent product visualization tool.
With respect to analytical support, we are able to
calculate similarity distance between actual and
desired products based on their descriptions
presented in unified form. Therefore, having
statistical measurement of a threshold for customers
churn and predicting a level of customers’
satisfaction, system will warn product/service
provider and suggest appropriate tailoring
(customization) of the offering to what customer(s)
want. Moreover, collected and aggregated
customer’s preferences can become a part of a
personal customer profile. Matching preferences
from the customer profile with descriptions of
available products and services, provider can push
personalized advertisements to potential customers.
4 FUTURE WORK
Elaboration of supporting tool for customer-driven
extension of the ontology could be considered as a
logical extension of current solution. Currently,
system works with property values (“objects”)
available in existing domain ontology(ies) and some
initial set of extra “relative” values (presented in
semantic customer feedback ontology). Assuming
that customers may express preferences, which are
not covered by mentioned ontologies, and even
suggest new part(s) and property(ies)/feature(s) of a
product, we have to provide an appropriate support
for ontology extension (Witte et. al., 2010). Such
extension has dual side effect. From one hand, it
requires more sophisticated and intelligent tool to
support it, taking care of similarity matching
(Shvaiko and Euzenat, 2012; Jain et. al., 2010) to
avoid an ambiguity and appearance of several
entities for the same notion. From the other hand, it
becomes very important for customer feedback
consumers, since it might bring something new (new
functionality or property of a product) that had not
been considered by them before.
Another planned direction of further research is
to apply results towards customer reviews
management. Usually, readers of reviews try to infer
from review content to which extent their
preferences overlap with the reviewer’s preferences
@prefix cp: <www.example.com/customer_preferemce/>.
@prefix rdf:<http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
@prefix pD: <www.example.com/productDescription/>.
@prefix dOnt: <www.example.com/domainOntology/>.
@prefix scf: <www.example.com/semCustomerFeedback/>.
cp:customerPreference_03
rdf:type scf:CustomerPreference;
scf:domain pD:kettle1;
scf:subject pD:wireOfDockStationOfKettle1;
scf:predicate dOnt:hasLength;
scf:object scf:longer.
cp:customerPreference_02
rdf:type scf:CustomerPreference;
scf:domain pD:kettle1;
scf:subject pD:mainBodyOfKettle1;
scf:predicate dOnt:hasMaterial;
scf:object dOnt:plastic.
cp:customerPreference_01
rdf:type scf:CustomerPreference;
scf:domain pD:kettle1;
scf:subject pD:handleOfKettle1;
scf:predicate dOnt:hasColor;
scf:object scf:lighter.
…
WEBIST 2016 - 12th International Conference on Web Information Systems and Technologies
202
(Bailey, 2005). Thus, having structured
representation of preferences in the users’ profiles, it
becomes possible to automate review clustering and
facilitate proper review selection.
5 CONCLUSIONS
Since, individual creativity and personal experiences
will always be critical components of marketing
decisions. The role of customer analytics is not
necessarily to replace these, but to help decision
makers to come to the fact-based conclusions
through better knowledge of the organization’s
customers and markets. One of the challenges for the
product/service providers is customer feedback
collection and analysis, since it is associated with a
real voice of a customer. Among other challenges
(e.g. fruitful customer engagement to feedback
provisioning process), processing of unstructured
text based feedbacks becomes very challenging and
does not provide sufficient result. Therefor current
research presents an approach towards structured
customer feedback gathering that further facilitates
automated generation of preferable/desired product
description. The main achievements of the proposed
solution are: enrichment of digital content (web-
based product or service description) with semantic
annotations; mechanism for customer driven
structured feedback provisioning; free text based
feedback transformation into RDF based structured
data; automated creation of a new or improved
product/service description with respect to
expectations and preferences of a customer.
ACKNOWLEDGEMENTS
Research is done in Agora Center and MIT
departments (University of Jyvaskyla, Finland)
under the DIGILE Need4Speed program funded by
TEKES and consortium of industrial partners.
REFERENCES
Adida, B., Birbeck, M., McCarron, S., Herman, I., 2015.
RDFa Core 1.1 - Third Edition. W3C
Recommendation. 17 March 2015, URL:
http://www.w3.org/TR/rdfa-core/
Bailey, A.A., 2005. Consumer Awareness and Use of
Product Review Websites. Journal of Interactive
Advertising. 6, 68–81.
Berners-Lee, T., Hendler, J., Lassila, O., 2001. The
Semantic Web, Scientific American 284(5), pp.34-43.
Chen, Y., Xie, J., 2008. Online Consumer Review: Word-
of-Mouth as a New Element of Marketing
Communication Mix. Management Science. 54, 477–
491.
Chevalier, J.A., Mayzlin, D., 2006. The Effect of Word of
Mouth on Sales: Online Book Reviews. Journal of
Marketing Research. 43, 345–354.
Cunningham, H., Maynard, D., Bontcheva, K., Tablan, V.,
2002. GATE: an Architecture for Development of
Robust HLT Applications. Proceedings of the 40th
Anniversary Meeting of the Association for
Computational Linguistics (ACL).
Ehrmann, T., Schmale, H., 2008. The Hitchhiker’s Guide
to the Long Tail: The Influence of Online-Reviews
and Product Recommendations on Book Sales -
Evidence from Ger- man Online Retailing. ICIS 2008
Proceedings.
Ghose, A., Ipeirotis, P.G., 2011. Estimating the
Helpfulness and Economic Impact of Prod- uct
Reviews: Mining Text and Reviewer Characteristics.
IEEE Transactions on Knowledge and Data
Engineering. 23, 1498–1512.
Jain, P., Hitzler, P., Sheth, A.P., Verma, K., Yeh, P.Z.,
2010. Ontology Alignment for Linked Open Data. In:
Proceedings of the 9th International SemanticWeb
Conference, ISWC 2010, Shanghai, China, November
7-11, 2010, Springer-Verlag, 402–417.
Khriyenko, O., 2015a. Customer Feedback System:
Evolution towards semantically-enhanced systems, In:
Proceedings of the 11th International Conference on
Web Information Systems and Technologies, WEBIST
2015, 20-22 May, 2015, Lisbon, Portugal, 518-525.
Khriyenko, O., 2015b. Semantic UI: Automated Creation
of Semantically Personalized User Interface, In: GSTF
International Journal on Computing (JoC), ISSN
2251-3043 (E-periodical: 2010-2283), October 2015,
Vol. 4, No. 3, pp. 42-50, DOI: 10.5176/2251-
3043_4.3.330.
Manning, C.D., Surdeanu, M., Bauer, J., Finkel, J.,
Bethard, S.J., McClosky, D., 2014. The Stanford
CoreNLP Natural Language Processing Toolkit. In
Proceedings of 52nd Annual Meeting of the
Association for Computational Linguistics: System
Demonstrations, pp. 55-60.
Semantic Web, 2001. URL: http://www.w3.org/2001/sw/
Shvaiko, P., Euzenat, J., 2012. Ontology matching: state
of the art and future challenges. IEEE Transactions on
Knowledge and Data Engineering, 2012.
Witte, R., Khamis, N., Rilling, J., 2010. Flexible Ontology
Population from Text: The OwlExporter. In: Int. Conf.
on Language Resources and Evaluation (LREC),
Valletta, Malta, ELRA (05/2010 2010).
Customer Perception Driven Product Evolution - Facilitation of Structured Feedback Collection
203
