ABSTRACT: uses of multifaceted search is in the domain

ABSTRACT:
The price play a role when a consumer decides to choose where to buy a product
online .Therefore, online retailers pay special attention to the usability and
efficiency of their Web shop user interfaces. Now a day, many Web shops make
use of the so-called faceted navigation user interface, which is in literature
also sometimes referred to as ‘faceted search’. facets are usually grouped by
their property in user interfaces, in order to prevent them from being
scattered around, and, thereby, confusing the user. Multifaceted search is a
commonly used interaction paradigm in e-commerce applications, such as Web
shops. Because of the large amount of possible product attributes, Web shops
usually make use of static information to determine which facets should be
displayed. Unfortunately, this approach does not take into account the user
query, leading to a non-optimal facet drill down process. In this paper, propose
on automatic facet selection, with the goal of minimizing the number of steps
needed to find the desired product. This article proposes several algorithms
for facet selection, which we evaluate against the state-of-the-art algorithms
from the literature.

Keywords
–Faceted Search,
collaborative recommendation, Facet Selection,
personalization.

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!


order now

I.      
INTRODUCTION

 

Online product search has nowadays
become more important than ever, as consumers purchase more often on the Web. One
explanation for this is that the Web facilitates the user finding products that
better match their.ing channels. Not only do the users have
access to more information (e.g., user reviews, exact product information),
they also found it easier to shop from their homes. On the other hand, because
of the many options, users are often overwhelmed and found it difficult to
browse through the available products. Multifaceted search, also sometimes
referred to as `guided navigation’, is a popular interaction paradigm that
allows users to navigate through multidimensional data. One of the main uses
of multifaceted search is in the domain of e-commerce, i.e., Web shops. It is
being employed to solve the parametric product search problem for Web shops
that have collected local openings and product information. For example, in a
Web shop the user might enter a query like `Samsung, gps’ in order to search for
a Samsung phone that has built-in GPS capabilities. After showing the initial
result set, most Web shopping interfaces display the facets of the products in
the result set, which can be used to further drill down into the results set.
The facets in this case are product attribute/value combinations.

An important problem of
multifaceted search is the, Selection of facets that should be displayed for
each query. Because products have so many attributes that could be displayed as
facets, Web shops usually have some static business logic to display certain
facets for each result set. Although this works for local Web shops that do not
have many product categories, the creation of this business logic is a time
consuming process and is not appropriate for Web wide product search. One
solution to this problem is to employ an optimized facet selection process. The
goal of such an optimization process is to show facets that effectively
partition the product search space so that the user can easily drill down and found
its desired product. In literature, this is referred to as the facet selection
problem, which can be expressed as the optimization of a hyperactive media link
generation process

 

       
I.           
FACET SELECTION

 

The goal of this paper is to reduce
the search effort of a user that is searching for a product that meets his
needs. In this section, we first give the formal problem formulation and then
we present the considered facet selection algorithms. We assume that the number
of results scanned by a user (before
finding his desired product) represents the search effort. The main use case
that we consider is that a user submits a query 
to the product search engine 
Next, the search engine computes a 
ranked list of products  and a set
of facets  of size  that are to be displayed. Furthermore, the
set represents all facets that belong to all products Similar to , we
incorporate four assumptions about the user in the considered simulation
strategies for the evaluation. First, we assume that for each user , who has
submitted query q, there exists a single target product that  full fills the user’s needs. This target
product is assumed to always be present in the initial result set , but can be
ranked very low. Second, the user ends the search session if it founds in the
top results. Third, a user exactly knows which facets (if any) from  are associated to. Last, if there are any
facets that are associated with, we assume that the user then also selects
these. we incorporate four assumptions about the user in the considered
simulation strategies for the evaluation. First, we assume that for each user
u, who has submitted query, there exists a single target product that full
fills the user’s needs. This target product is assumed to always be present in
the initial result set, but can be ranked very low. Second, the user ends the
search session if it founds in the top results. Third, a user exactly knows which
facets (if any) from are associated to. Last, if there are any facets that are
associated with what, we assume that the user then also selects these. In this
paper, we also assume that multiple clicks (drill downs) can occur. More
specifically, we assume that the above described process can repeat itself a maximum
of times (iterations). If the user founds the desired product in the top
results in less than iterations, then the search session ends prematurely,
otherwise it ends after iterations. If we let it to remain unchanged, then the result
set at any iteration can be denoted as where represents the previously selected
facets. Similarly, the proposed facets by the search engine at any iteration
are denoted

DATASET

2.1 Data Collection

We use a data set that is gathered
from Kieskeurig.nl,

the largest price comparison site
in the Netherlands. This service does not only provide price comparisons, but
also has very detailed information on products. For this evaluation, we focused
on consumer electronics and chose mobile phones to be the category of products
that we use in the experiments. The data set contains 980 products for which we
have key/value pairs, i.e., product attributes. All product information is in
Dutch, but should be understandable also for non-Dutch speaking people because
of the frequently used English terminology in the product attributes. Using the
product attributes, we created the facets using the following rules. A facet is
a combination of a product attribute and a value (or range of values. For
product properties that represent multivalued qualitative values, such as
`Supported Video formats’, we created a binary facet for each value. Similarly,
for single-valued qualitative product attributes, we created a single facet for
each value. For all the quantitative properties, we manually defined the ranges
that would represent the different facets. As a result of this facet creating
process, we obtained 487 facets for the 980 products. The size and variety of
this data set allows for a thorough evaluation of the facet selection algorithms

 

Personalized
facet search mechanism

 

An
efficient personalized faceted search mechanism can be used to: 1) solve millions
of e-commerce users’ immediate information needs. 2) Help users better understand
the data, especially the data space relevant to the user, And 3) help users better
understand how the engine works through the simple interactive interface, and
there by train users in how to make more effective use of the interface over
time.

.

 

 

 PROPOSED EVALUATION METHODOLOGY

We propose new algorithms for the
facet selection problem in product search. We evaluate several

approaches and compare our proposed
algorithms against several state-of-the-art facet selection algorithms from the
literature. Our proposed algorithms aim to partition the space in the most effective
manner and thus allow the user to drill down in the least amount of time. We
perform the evaluation on a large data set and analyze the results, differently
from previous works, across three different measures.

1.      
User Interface Utility

While
the specifics vary between individual faceted search interfaces, every
interface shares certain characteristics. In general, a faceted search
interface is divided into three parts. The first part is a list of the facets
in the document collection. Each facet has a list of available values associated
with it. When there area large number of values for a facet, interfaces tend to
display only a fraction of the available values, but allow the user to view the
complete list upon request. The user can restrict the current query by
selecting facet-value pairs from this list subset of the search space the user
is examining, but also allows the user to broaden the current query by removing
some previously selected facet-value pairs.

I.                  
PERSONALIZED
FACET-VALUE PAIR RECOMMENDATION

As stated in the
introduction, one of the keys to building an effective faceted search interface
is presenting the user with facet-values that that a relevant to the user’s
current search task. If the presented facet-values are not relevant to the
task, the user could be forced to spend extra effort to find his/her
document(s), or in the worst case not find
his/her document(s) at all. This section
describes several possible algorithms to select facet-value pairs.

Facet-Value Pair Suggestions After a user performs
an action in the middle of a session, the system needs present a list of facet value
pairs. We can view this as a feature selection ask, and the following is an incomplete
list of algorithms that can be used

 Most Frequent This is the
simplest suggestion method. In this method, the facet-value pairs that are
found in the currently selected documents are counted  ,and the most frequent values for each facet
are presented to the user for query refinement. This method is popular among
many commercially available faceted search interfaces, and thus provided an
appropriate baseline for comparison.

Most Probable In this method,
the facet-value pairs in the currently selected documents are ranked according to
their probability of being included in a document relevant to the user. These
probabilities can either be determined by the relevance judgments by the
community of users (Collaborative Prob.), or personalized for each
individual user (Personal Prob ).This method was examined as it can be
easily integrated into adaptive and personalized retrieval algorithms.

Mutual Information The point wise mutual information between the
presence of a facet-value pair appearing in a document and a document’s
relevanceiscalculated.The most informative values are then presented to the user
for query refinement. Mutual information was considered as a facet-value
suggestion metho nd since  it is a common
method used for feature selection.

2. Starting/Landing Page for
Faceted Search
A faceted search system needs to present a good starting/landing page for a
user. Since a user’s profile describes the qualities that define a relevant
document for that user, this information can be used by the system to initially
place each user nearer to his/her relevant documents even before
the user begins to formulate a query. This is accomplished by examining the
user’s profile and automatically constructing an initial query based on
facet-value pairs that are likely to be contained in the user’s relevant
documents. This automatically constructed query, along with the documents
returned by this query, creates the start state. The following
three methods to determine user’s start state are proposed.
Null Start State This is the simplest start state creation method. In
this method, each user begins in a state with no face-value pairs selected by
default and no pre-fetched documents. This method served as a baseline for
comparing the other start state creation methods as this is the most prevalent
method for beginning user initiated searches in information retrieval systems.

Collaborative Start State In this method the system automatically issues a query containing
the facet-value pairs that are the mostly likely to be contained in relevant
document as determined by the common
Bayesian prior. This query is then issued to the underlying retrieval algorithm
and the matching documents are initially suggested to the user. Since the start
state is created by the common prior, every userispresentedthesamestartstate.
Personalized Start State This method is similarly to the method above,
except that the default query is determined by each user’s profile.

3.      
EMPIRICALEVALUATION

In order to demonstrate these ideas, a set of experiments were
carried out using documents from the Internet Movie Database (IMDB) corpus along
with real user relevance
judgments for each document from the Movie Lens and Netflix Prize  corpora. These data sets were chosen since they
provided documents containing facets, such as director and actor, along
with real user relevancejudgmentsonthesedocuments. The IMDB corpus was trimmed
to contain only
documents found in either the Movie Lens or Netflix corpora. This led to
approximately 8,000 documents containing 367,417 facet-value pairs spread among
19facets.Both the Movie Lens and Netflix corpora were reduced to approximately
5,000 unique users through uniform sampling, giving742,036 and 633,257 user
judgments respectively. In both
data sets users expressed a preference for retrieved documents based on a 1 to
5 rating scale. Each rating was converted into Boolean relevance judgments by
assuming all movies that were rated 4 or greater were relevant, and all movies
rated 3 or less were non-relevant. These user judgments were randomly divided
90% for training and 10% for
testing. For simplicity, all facets were assumed to be nominal. Each user model
was a multivariate Bernoulli distribution, with the shared prior being a multi
variate Gamma distribution. Without loss of generality, a simple reward
mechanism shown in Table 4 is used for evaluation. The interface was configured
to return a maximum of 10 matching
documents per page, and a maximum of 5 values for each facet.

III.EXPERIMENTAL RESULTS

For the evaluation, we simulate a user
that is in a faceted search session. There are two aspects that are important
in this type of simulation. First, we need a way to generate queries that are
sufficiently realistic for the experiments. Second, we need one or more
simulation strategies of users in order to simulate the clicking on a facet.
Before we go into the details of these two aspects, let us first explain on a
high level how we have designed the simulation. Given a query, we submit it to
the product search engine, after which, for every product that we consider as a
possible target product, we simulate a faceted search session. The set of
possible target products consists of the 100 products after the top products.
We set which results in performing the simulation with each product ranked in
the range as a target product. The reason for this is that we want to measure
how the algorithms perform for many different target products. Next, the ranked
search results are obtained and a faceted search session is simulated, where a
user is aware of the target product, but is only able to recognize it when it
appears in the top-10 results. The user keeps clicking on a facet (described
shortly) until either the target product

Table:3 Search Efficiency for Netflix
Users

Action corresponds
to a bigger user utility. Four note worthy conclusions
can be made from these results. First, point wise mutual information (PMI)
significantly under Performed
when compared to the other facet value selection methods. Mutual information
measures the correlation between two random variables, in this case the
presence or absence of a facet-value pair and a document’s relevance. Point wise
mutual information rather than complete mutual information was used because only
facet-value pairs that
have a positive correlation should be suggested to the user. PMI breaks down when
there are facet-value pairs that are
strongly correlated with relevant documents, but occur in only tiny fraction of
all relevant documents. For example, if
all films containing the facet-value pair genre=film no ire ranked highly, then genre=film will be suggested early for query refinement,
even if genre=filmnoiris contain edibles than 1 percent   of all relevant documents. These results
show that correlation measure such
as mutual information  are not a good choice  for this type of selection problem, since the
probability of utilizing the suggested features is more important than how tightly
correlated they are with relevance. Finally, simply suggesting the most
frequent values for each
facet performed well when compared to the personalized suggestion methods.
There are two possible reasons for this. First, the frequency of facet value pairs
in the documents is correlated with users’ idea of what makes a document relevant.
In general this may not be the case, and thus frequency may not be a good
facet-value selection mechanism
for when the users’ expectations do not closely match what is contained in the
document repository. In this
case frequency would fail to provide good suggestions, and the personalized
probability models would be shown to be
superior. Second, the Personal
Prob algorithm
used to select facet-value pairs is not good enough and far from optimal. This
is not surprising since the probabilistic model proposed in this paper is based
on strong assumptions that may
not be true on the evaluation data sets.

CONCLUSIONS

This article focused on automatic
facet selection in the domain of e-commerce, for the purpose of minimizing the
number of steps required by the user in order to found its desired product. We
proposed several facet selection algorithms, which we evaluated against the
state-of-the-art algorithms from literature. Furthermore, we implemented all
considered facet selection algorithms in a freely available Web application
called faccy.net.The evaluation was performed with simulations employing 1000 queries,
980 products, 487 facets, and three drill down strategies. We used three
different evaluation metrics. The experimental environment is repeatable and
controllable, which makes it a bench mark able evaluation environment.
Although the simulated users differ from real users, the evaluation methodology
does provide insight into understanding how various faceted interface design
algorithms perform. This paper does not intend to claim whether this
evaluation method is
better or worse than user studies. Instead, the outlined approach serves to complement
user studies by being cheap, repeatable, and controllable. How to select a set of
facet-value pairs at each step of
the interaction process to optimize a user utility is a more fundamental that
requires future research. This paper serves a first step towards personalized
faceted search. The facet value pair selection algorithms examined in this
paper are far from optimal.

 

 REFERENCES

1. AWS, “Amazon Web Services. Large
cloud computing provider from Amazon.com,” http://aws.amazon.com, 2014.

2 H.-J. Kim, Y. Zhu, W. Kim, and T.
Sun, “Dynamic Faceted Navigation in Decision Making using Semantic Web
Technology,” Decision Support Systems, vol. 61, pp. 59–68, 2014.

.3. S. Johnson Lim, Y. Liu, and W. Lee.
Multi-facet

product information search and
retrieval using semantically annotated product family ontology.

Information Processing &
Management,

46(4):479{493,2010.
4. J. Koren,
Y. Zhang, and X. Liu. Personalized

Interactive faceted search. In
Proceedings of the 17th

International conference on World
Wide Web (WWW 2008), pages 477{486. ACM, 2008.
6. B. Kules,
R. Capra, M. Banta, and T. Sierra. What do exploratory searchers look at in a
faceted search interface? In Proceedings of the 9th ACM/IEEE-CS Joint
Conference on Digital Libraries (JCDL 2009), pages 313{322. ACM, 2009.
7. S.
Liberman and R. Lempel. Approximately optimal facet selection. In Proceedings
of the 27th Annual  ACM Symposium on
Applied Computing (SAC 2012),pages 702{708. ACM, 2012.
8. S. Pandit
and C. Olston. Navigation-aided retrieval. In Proceedings of the 16th
International Conference on World Wide Web (WWW 2007), pages 391{400.ACM, 2007.
9. G. Sacco
and Y. Tzitzikas. Dynamic Taxonomies and Faceted Search, volume 25. Springer,
2009.
10 R. Li, S. Bao, Y. Yu, B. Fei, and Z. Su. Towards
effective browsing of large scale social annotations. In  WWW ’07:
Proceedings of the 16th international
conference on World Wide Web,
pages 943–952, NewYork,NY,USA,2007.ACM.
11. D.
Tunkelang. Faceted search. Synthesis Lectures on Information Concepts,
Retrieval, and Services,1(1):1{80,2009.
12 Y. Zhu, D.
Jeon, W. Kim, J. Hong, M. Lee, Z. Wen, and Y. Cai”The Dynamic Generation of
Refining Categories in Ontology- Based Search,” in Semantic Technology, ser.
Lecture Notes in Computer Science, 2013, vol. 7774, pp. 146–158.

14
S. Pandit and C. Olston. Navigation-aided retrieval.In WWW
’07: Proceedings of the 16th international
conference on World Wide Web,
2007

15. Kieskeurig.nl, “Major Dutch price
comparison engine with detailed product descriptions,”
http://www.kieskeurig.nl,2014.
16. Q. Liu, E.
Chen, H. Xiong, C. H. Ding, and J. Chen, “Enhancing Collaborative Filtering by
User Interest Expansion via Personalized Ranking,” IEEE Transactions on
Systems, Man, and Cybernetics, Part B: Cybernetics, vol. 42, no. 1, pp.
218–233, 2012
17
Tweakers.net, “Dutch IT-community with a dedicated price comparison
department,” http://www.tweakers.net, 2014.  .
18 S.
Liberman and R. Lempel, “Approximately Optimal Facet Value Selection,” Science
of Computer Programming, vol. 94, pp.18–31, 2014